1 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3 * target.def (preferred_else_value): New target hook.
4 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
5 * doc/tm.texi: Regenerate.
6 * targhooks.h (default_preferred_else_value): Declare.
7 * targhooks.c (default_preferred_else_value): New function.
8 * internal-fn.h (conditional_internal_fn_code): Declare.
9 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10 (get_conditional_internal_fn): Use it.
11 (conditional_internal_fn_code): New function.
12 * gimple-match.h (gimple_match_cond): New struct.
13 (gimple_match_op): Add a cond member function.
14 (gimple_match_op::gimple_match_op): Update all forms to take a
16 * genmatch.c (expr::gen_transform): Use the same condition as res_op
17 for the suboperation, but don't specify a particular else_value.
18 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
19 (visit_nary_op, visit_reference_op_load): Pass
20 gimple_match_cond::UNCOND to the gimple_match_op constructor.
21 * gimple-match-head.c: Include tree-eh.h
22 (convert_conditional_op): New function.
23 (maybe_resimplify_conditional_op): Likewise.
24 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
25 (gimple_resimplify2): Likewise.
26 (gimple_resimplify3): Likewise.
27 (gimple_resimplify4): Likewise.
28 (maybe_push_res_to_seq): Return null for conditional operations.
29 (try_conditional_simplification): New function.
30 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
32 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
34 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
36 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
38 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
40 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
42 (hash_tree): Do not hash DECL_FCONTEXT
43 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
44 Do not stream DECL_FCONTEXT.
45 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
46 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
48 2018-07-12 Richard Biener <rguenther@suse.de>
51 * dwarf2out.c (gen_block_die): Only output blocks when they have
52 at least one !DECL_IGNORED_P variable.
54 2018-07-12 Richard Biener <rguenther@suse.de>
57 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
58 Remove -mieee-fp handling.
60 2018-07-12 Richard Biener <rguenther@suse.de>
62 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
63 left-over from last patch.
65 2018-07-12 Jakub Jelinek <jakub@redhat.com>
67 PR tree-optimization/86492
68 * gimple-ssa-store-merging.c
69 (imm_store_chain_info::coalesce_immediate_stores): Call
70 check_no_overlap even for the merge_overlapping case. Formatting fix.
72 2018-07-12 Richard Biener <rguenther@suse.de>
75 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
76 move possibly trapping operations into the conditional.
78 2018-07-12 Richard Biener <rguenther@suse.de>
80 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
81 (vn_lookup_simplify_result): Remove recursion limit applied
83 (vn_nary_build_or_lookup_1): Adjust.
84 (try_to_simplify): Likewise.
85 * gimple-match-head.c (gimple_resimplify1): Instead apply one
87 (gimple_resimplify2): Likewise.
88 (gimple_resimplify3): Likewise.
89 (gimple_resimplify4): Likewise.
91 2018-07-11 Jakub Jelinek <jakub@redhat.com>
93 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
94 Use __mmask64 type instead of __mmask8 for __M argument.
95 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
96 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
98 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
99 __mmask16 for __M argument.
100 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
101 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
102 to __mmask16 instead of __mmask8.
103 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
104 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
105 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
106 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
107 instead of __mmask16 for __U argument.
108 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
109 __mmask16 instead of __mmask8 for __U argument.
110 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
112 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
114 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
116 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
118 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
120 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
122 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
124 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
126 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
128 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
130 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
131 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
132 return type as well as __M argument type and all casts from __mmask8
134 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
135 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
136 return type as well as __M argument type and all casts from __mmask8
138 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
139 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
140 return type as well as __M argument type and all casts from __mmask8
142 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
143 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
144 return type as well as __M argument type and all casts from __mmask8
146 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
147 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
150 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
152 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
153 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
156 2018-07-11 Paul Koning <ni1d@arrl.net>
158 * doc/md.texi (define_subst): Document how multiple occurrences of
159 the same argument in the replacement pattern are handled.
161 2018-07-11 Paul Koning <ni1d@arrl.net>
163 * doc/extend.texi (Common Variable Attributes): Move "mode" into
165 (Common Type Attributes): Add "mode" attribute.
167 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
169 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
170 stream DECL_ORIGINAL_TYPE.
171 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
172 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
173 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
174 Do not walk original type.
175 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
177 (write_ts_decl_non_common_tree_pointers): Do not stream
179 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
180 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
182 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
184 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
185 thread twice from the same starting edge.
187 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
189 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
190 * gimple.c (gimple_call_nonnull_result_p): ...here...
191 (gimple_call_nonnull_arg): ...and here.
192 * gimple.h (gimple_call_nonnull_result_p): New.
193 (gimple_call_nonnull_arg): New.
195 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
197 * config/arm/arm-cpus.in: Move information from fpu field of each
198 cpu definition to the isa field.
199 * config/arm/parsecpu.awk (fpu): Delete match rule.
200 (gen_comm_data): Don't add bits from the CPU's FPU entry.
202 2018-07-11 Richard Biener <rguenther@suse.de>
205 * dwarf2out.c (init_sections_and_labels): Use
206 output_asm_line_debug_info consistently.
207 (dwarf2out_early_finish): Likewise.
208 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
211 2018-07-11 Richard Biener <rguenther@suse.de>
213 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
214 Rework father_bb setting in a way to avoid propagating constants
215 multiple times on a loop body.
217 2018-07-10 Mark Wielaard <mark@klomp.org>
220 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
222 2018-07-10 Richard Biener <rguenther@suse.de>
224 * hash-map.h (hash_map::iterator::operator*): Return
225 references to key and value.
227 2018-07-10 Jakub Jelinek <jakub@redhat.com>
230 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
231 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
232 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
233 (gimplify_omp_for): For composite loops, move outer
234 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
235 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
236 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
237 TREE_LIST for both the original class iterator and the "last" helper
238 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
239 loop, remember has_decl_expr from outer composite loops for the
240 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
242 2018-07-09 Martin Sebor <msebor@redhat.com>
246 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
247 accessing implicitly initialized array elements.
248 * expr.c (string_constant): Handle string initializers of
249 character arrays within aggregates.
250 * gimple-fold.c (fold_array_ctor_reference): Add argument.
251 Store element offset. As a special case, handle zero size.
252 (fold_nonarray_ctor_reference): Same.
253 (fold_ctor_reference): Add argument. Store subobject offset.
254 * gimple-fold.h (fold_ctor_reference): Add argument.
256 2018-07-09 Paul Koning <ni1d@arrl.net>
258 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
259 (pdp11_insn_cost): New function.
260 (pdp11_md_asm_adjust): New function.
261 (TARGET_INVALID_WITHIN_DOLOOP): Define.
262 (pdp11_rtx_costs): Update to match machine better.
263 (output_addr_const_pdp11): Correct format mismatch warnings.
264 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
265 * config/pdp11/pdp11.md: General change to add base_cost and/or
266 length attributes for use by new pdp11_insn_cost function.
267 (MIN_BRANCH): Correct definition.
269 (doloop_end): Use standard pattern name for looping pattern.
270 (doloop_end_nocc): New.
271 (movsf): Add another constraint alternative.
272 (zero_extendqihi2): Add constraint alternatives for not in place
274 (zero_extendhisi2): Remove.
275 (shift patterns): Add CC handling variants.
279 (define_peephole2): New peephole to recognize mov that sets CC for
282 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
284 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
285 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
288 2018-07-09 Jakub Jelinek <jakub@redhat.com>
291 * real.c (real_nextafter): Return true if result is denormal.
293 2018-07-09 Martin Liska <mliska@suse.cz>
295 * common.opt: Add back wrongly removed attribute.
297 2018-07-09 Richard Biener <rguenther@suse.de>
300 * dwarf2out.c (gen_block_die): For an early generated DIE
301 always output high/low PC attributes.
303 2018-07-09 Tom de Vries <tdevries@suse.de>
305 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
306 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
307 onto VAR_DECL with abstract origin.
309 2018-07-07 Jim Wilson <jimw@sifive.com>
311 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
313 2018-07-07 Tom de Vries <tdevries@suse.de>
315 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
317 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
319 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
321 (combine_bound): Use wide-int overflow calculation instead of
323 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
325 * fold-const.c (int_const_binop_2): Same.
326 (extract_muldiv_1): Same.
327 (fold_div_compare): Same.
328 (fold_abs_const): Same.
330 * poly-int.h (add): Same.
334 * predict.c (predict_iv_comparison): Same.
335 * profile-count.c (slow_safe_scale_64bit): Same.
336 * simplify-rtx.c (simplify_const_binary_operation): Same.
337 * tree-chrec.c (tree_fold_binomial): Same.
338 * tree-data-ref.c (split_constant_offset_1): Same.
339 * tree-if-conv.c (idx_within_array_bound): Same.
340 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
341 * tree-ssa-phiopt.c (minmax_replacement): Same.
342 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
343 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
344 * vr-values.c (vr_values::adjust_range_with_scev): Same.
345 * wide-int.cc (wi::add_large): Same.
346 (wi::mul_internal): Same.
347 (wi::sub_large): Same.
348 (wi::divmod_internal): Same.
349 * wide-int.h: Change overflow type to overflow_type for neg, add,
350 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
351 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
352 mul_internal, divmod_internal.
353 (overflow_type): New enum.
354 (accumulate_overflow): New.
356 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
358 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
359 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
361 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
363 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
364 argument is checked for zero before entering loop, avoid checking again.
366 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
368 * gimplify.h (generic_expr_could_trap_p): Set as global function.
369 * gimplify.h (generic_expr_could_trap_p): Likwise.
370 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
372 2018-07-06 Jakub Jelinek <jakub@redhat.com>
374 PR tree-optimization/86401
375 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
376 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
377 (fold_bit_and_mask): ... here. New helper function for match.pd.
378 * fold-const.h (fold_bit_and_mask): Declare.
379 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
381 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
384 * target.def (translate_mode_attribute): New hook.
385 * targhooks.h (default_translate_mode_attribute): Declare.
386 * targhooks.c (default_translate_mode_attribute): New function.
387 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
388 * doc/tm.texi: Regenerate.
389 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
390 (rs6000_translate_mode_attribute): New function.
392 2018-07-06 Paul Koning <ni1d@arrl.net>
394 * doc/md.texi (define_split): Document DONE and FAIL.
395 (define_peephole2): Ditto.
397 2018-07-05 Jeff Law <law@redhat.com>
399 PR tree-optimization/86010
400 * tree-ssa-dse.c (compute_trims): More aggressively trim at
401 both the head and tail of mem* and str* calls.
403 2018-07-05 Jim Wilson <jimw@sifive.com>
405 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
407 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
409 * config/aarch64/aarch64-simd.md: correct flags text for
410 MIN_EXPR replacement.
412 2018-07-05 James Clarke <jrtc27@jrtc27.com>
414 * configure: Regenerated.
416 2018-07-05 Carl Love <cel@us.ibm.com>
418 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
419 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
420 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
421 VSX_BUILTIN_DOUBLEL_V4SF.
423 2018-07-05 Martin Sebor <msebor@redhat.com>
426 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
427 than its domain to compute its the upper bound of a char array.
429 2018-07-05 Nathan Sidwell <nathan@acm.org>
431 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
432 * doc/cpp.texi: Update comment.
433 * doc/tm.texi: Rebuilt.
434 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
435 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
436 * doc/extend.texi (Backwards Compatibility): Clarify it is system
437 headers affected by extern "C".
438 * system.h: Poison NO_IMPLICIT_EXTERN_C.
439 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
440 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
441 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
442 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
443 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
444 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
445 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
446 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
447 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
448 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
449 config/sparc/sp64-elf.h, config/spu/spu.h,
450 config/stormy16/stormy16.h, config/v850/v850.h,
451 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
452 define NO_IMPLICIT_EXTERN_C.
453 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
455 2018-07-05 Tamar Christina <tamar.christina@arm.com>
458 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
459 instead of GET_MODE_SIZE when comparing Units.
461 2018-07-05 Tamar Christina <tamar.christina@arm.com>
464 * rtlanal.c (set_noop_p): Constrain on mode change,
465 include hard-reg-set.h
467 2018-07-05 Tamar Christina <tamar.christina@arm.com>
469 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
471 2018-07-05 Jakub Jelinek <jakub@redhat.com>
474 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
477 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
479 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
481 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
484 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
486 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
488 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
491 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
493 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
495 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
498 2018-07-04 Martin Liska <mliska@suse.cz>
500 * tree-switch-conversion.c: Define
501 max_ratio_for_speed and max_ratio_for_size constants.
503 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
504 Martin Liska <mliska@suse.cz>
509 * common.opt: Rename align options with 'str_' prefix.
510 * common/config/i386/i386-common.c (set_malign_value): New
512 (ix86_handle_option): Use it to set -falign-* options/
513 * config/aarch64/aarch64-protos.h (struct tune_params): Change
514 type from int to string.
515 * config/aarch64/aarch64.c: Update default values from int
517 * config/alpha/alpha.c (alpha_override_options_after_change):
519 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
520 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
521 max skip conditionally.
522 * config/i386/freebsd.h (SUBALIGN_LOG): New.
523 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
524 max skip conditionally.
525 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
526 max skip conditionally.
527 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
528 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
529 max skip conditionally.
530 * config/i386/i386.c (struct ptt): Change type from int to
532 (ix86_default_align): Set default values.
533 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
534 max skip conditionally.
535 * config/i386/iamcu.h (SUBALIGN_LOG): New.
536 (ASM_OUTPUT_MAX_SKIP_ALIGN):
537 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
538 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
539 max skip conditionally.
540 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
541 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
542 * config/i386/x86-64.h (SUBALIGN_LOG): New.
543 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
544 max skip conditionally.
545 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
546 * config/ia64/ia64.c (ia64_option_override): Set default values
547 for alignment options.
548 * config/m68k/m68k.c: Handle new str_align_* options.
549 * config/mips/mips.c (mips_set_compression_mode): Change
551 (mips_option_override): Set default values for options.
552 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
554 * config/rs6000/rs6000.c (rs6000_option_override_internal):
556 * config/rx/rx.c (rx_option_override): Likewise.
557 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
558 (LABEL_ALIGN): Use align_labels_log.
559 (LOOP_ALIGN): Use align_loops_align.
560 * config/s390/s390.c (s390_asm_output_function_label): Use new
562 * config/sh/sh.c (sh_override_options_after_change):
563 Change type of constants.
564 * config/spu/spu.c (spu_sched_init): Likewise.
565 * config/sparc/sparc.c (sparc_option_override): Set default
567 * config/visium/visium.c (visium_option_override): Likewise.
568 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
569 emit p2align format with last argument if it's not needed.
570 * doc/invoke.texi: Document extended format of -falign-*.
571 * final.c: Use align_labels alignment.
572 * flags.h (struct target_flag_state): Change type to use
574 (struct align_flags_tuple): New.
575 (struct align_flags): Likewise.
576 (align_loops_log): Redefine macro to use new types.
577 (align_loops_max_skip): Redefine macro to use new types.
578 (align_jumps_log): Redefine macro to use new types.
579 (align_jumps_max_skip): Redefine macro to use new types.
580 (align_labels_log): Redefine macro to use new types.
581 (align_labels_max_skip): Redefine macro to use new types.
582 (align_functions_log): Redefine macro to use new types.
583 (align_loops): Redefine macro to use new types.
584 (align_jumps): Redefine macro to use new types.
585 (align_labels): Redefine macro to use new types.
586 (align_functions): Redefine macro to use new types.
587 (align_functions_max_skip): Redefine macro to use new types.
588 (align_loops_value): New macro.
589 (align_jumps_value): New macro.
590 (align_labels_value): New macro.
591 (align_functions_value): New macro.
592 * function.c (invoke_set_current_function_hook): Propagate
593 alignment values from flags to global variables default in
595 * ipa-icf.c (sem_function::equals_wpa): Use
596 cl_optimization_option_eq instead of memcmp.
597 * lto-streamer.h (cl_optimization_stream_out): Support streaming
599 (cl_optimization_stream_in): Likewise.
600 * optc-save-gen.awk: Support strings in cl_optimization.
601 * opth-gen.awk: Likewise.
602 * opts.c (finish_options): Remove error checking of invalid
604 (MAX_CODE_ALIGN): Remove.
605 (MAX_CODE_ALIGN_VALUE): Likewise.
606 (parse_and_check_align_values): New function.
607 (check_alignment_argument): Likewise.
608 (common_handle_option): Use check_alignment_argument.
609 * opts.h (parse_and_check_align_values): Declare.
610 * toplev.c (init_alignments): Remove.
611 (read_log_maxskip): New.
612 (parse_N_M): Likewise.
613 (parse_alignment_opts): Likewise.
614 (backend_init_target): Remove usage of init_alignments.
615 * toplev.h (parse_alignment_opts): Declare.
616 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
618 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
619 * tree.c (cl_option_hasher::equal): New.
620 * varasm.c: Use new global macros.
622 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
624 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
625 Use a simpler align directive also if MAXSKIP = ALIGN-1.
626 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
627 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
628 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
629 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
630 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
631 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
632 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
633 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
634 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
635 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
636 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
638 2018-07-04 Martin Liska <mliska@suse.cz>
639 Jonathan Wakely <jwakely@redhat.com>
641 * coverage.c: Use correct type.
642 * doc/invoke.texi: Language correction.
644 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
647 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
648 ENDBRANCH for non-tail call which may return via indirect branch.
649 * doc/extend.texi: Document indirect_return attribute.
651 2018-07-03 Martin Sebor <msebor@redhat.com>
653 PR tree-optimization/86274
654 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
656 (format_floating): Correct handling of infinities and NaNs.
658 2018-07-03 Martin Sebor <msebor@redhat.com>
660 * print-tree.c (print_real_cst): New function.
661 (print_node_brief): Call it.
664 2018-07-03 Jeff Law <law@redhat.com>
666 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
667 into a single pattern.
669 * config/h8300/h8300.md (ors code_iterator): New.
670 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
671 a single pattern and single splitter.
672 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
673 (iorqi3_1, xorqi3_1): Likewise.
674 (iorqi3, xorqi3 expanders): Similarly.
676 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
677 (movmd_internal) into a single pattern using the P mode iterator.
678 (movmd splitters): Similarly.
679 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
680 (movsd splitters): Similarly.
682 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
683 ADDB, ADDW and ADDL into a single ADD attribute which selects the
684 right table based on the size of the operand.
685 * config/h8300/h8300.md (length_table): Corresponding changes. All
686 references to "addb", "addw" and "addl" changed to "add".
687 (btst patterns): Merge two variants into a single pattern.
688 (tstqi, tsthi): Likewise.
689 (addhi3_incdec, addsi3_incdec): Likewise.
690 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
691 (mulhi3, mulsi3): Likewise.
692 (udivhi3, udivsi3): Likewise.
693 (divhi3, divsi3): Likewise.
694 (andorqi3, andorhi3, andorsi3): Likewise.
696 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
699 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
700 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
702 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
704 PR tree-optimization/85694
705 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
706 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
708 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
709 (<u>avg<mode>3_ceil): New patterns.
711 2018-07-03 David Malcolm <dmalcolm@redhat.com>
713 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
714 scan-tree-dump directive.
715 * gcc.dg/vect/slp-perm-2.c: Likewise.
716 * gcc.dg/vect/slp-perm-3.c: Likewise.
717 * gcc.dg/vect/slp-perm-5.c: Likewise.
718 * gcc.dg/vect/slp-perm-6.c: Likewise.
719 * gcc.dg/vect/slp-perm-7.c: Likewise.
720 * gcc.dg/vect/slp-perm-8.c: Likewise.
722 2018-07-03 Marek Polacek <polacek@redhat.com>
725 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
727 2018-07-03 Richard Biener <rguenther@suse.de>
730 * tree-ssa-structalias.c (find_func_clobbers): Properly
731 handle indirect calls.
733 2018-07-03 Jeff Law <law@redhat.com>
735 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
736 (shifts): New code iterator.
737 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
738 expander. Fix HImode handling on H8/SX.
739 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
740 (subqi3, subhi3, subsi3 expanders): Likewise.
741 (andqi3, andhi3, andsi3 expanders): Likewise.
742 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
743 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
744 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
745 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
746 (zero_extendqihi2, zero_extendqisi2): Likewise.
747 (extendqihi2, extendqisi2): Likewise.
748 (rotlqi3, rotlhi3, rotlsi3): Likewise.
749 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
750 (rotlqi3_1, rotlhi3_1): Likewise.
751 (logicalhi3_sn, logicalsi3_sn): Likewise.
752 (logicalhi3, logicalsi3): Likewise.
754 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
756 * tree-vect-patterns.c (vect_recog_rotate_pattern)
757 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
758 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
759 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
760 type to append_pattern_def_seq instead of creating a stmt_vec_info
762 (build_mask_conversion): Likewise. Remove vinfo argument.
763 (vect_add_conversion_to_patterm): Likewise, renaming to...
764 (vect_add_conversion_to_pattern): ...this.
765 (vect_recog_mask_conversion_pattern): Update call to
766 build_mask_conversion. Pass the vector type to
767 append_pattern_def_seq here too.
768 (vect_recog_gather_scatter_pattern): Update call to
769 vect_add_conversion_to_pattern.
771 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
773 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
774 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
775 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
776 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
777 STMT_VINFO_PATTERN_DEF_SEQ to null here.
778 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
779 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
780 append_pattern_def_seq instead of new_pattern_def_seq.
781 (vect_recog_divmod_pattern): Do both of the above.
782 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
785 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
787 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
788 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
789 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
790 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
791 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
792 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
793 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
794 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
795 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
796 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
797 parameter with a single stmt_vec_info.
798 (vect_recog_func_ptr): Likewise.
799 (vect_recog_gather_scatter_pattern): Likewise, folding in...
800 (vect_try_gather_scatter_pattern): ...this.
801 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
802 the stmt_vec_info of the statement to be matched. Don't clear
803 STMT_VINFO_RELATED_STMT.
804 (vect_pattern_recog): Update call accordingly.
806 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
808 PR tree-optimization/85694
809 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
810 (uavgM3_ceil): Document new optabs.
811 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
812 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
814 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
815 (savg_ceil_optab): New optabs.
816 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
817 (vect_vect_recog_func_ptrs): Add it.
818 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
819 constant directly from the associated lhs.
821 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
823 * tree-vect-patterns.c (vect_split_statement): New function.
824 (vect_convert_input): Use it to try to split an existing cast.
826 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
828 * poly-int.h (print_hex): New function.
829 * dumpfile.h (dump_dec, dump_hex): Declare.
830 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
831 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
832 min_input_precision, operation_precision and operation_sign.
833 * tree-vect-patterns.c (vect_get_range_info): New function.
834 (vect_same_loop_or_bb_p, vect_single_imm_use)
835 (vect_operation_fits_smaller_type): Delete.
836 (vect_look_through_possible_promotion): Add an optional
837 single_use_p parameter.
838 (vect_recog_over_widening_pattern): Rewrite to use new
839 stmt_vec_info infomration. Handle one operation at a time.
840 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
841 (vect_truncatable_operation_p, vect_set_operation_type)
842 (vect_set_min_input_precision): New functions.
843 (vect_determine_min_output_precision_1): Likewise.
844 (vect_determine_min_output_precision): Likewise.
845 (vect_determine_precisions_from_range): Likewise.
846 (vect_determine_precisions_from_users): Likewise.
847 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
848 (vect_vect_recog_func_ptrs): Put over_widening first.
850 (vect_pattern_recog): Call vect_determine_precisions.
852 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
854 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
855 statements that have been replaced by further pattern statements.
856 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
858 2018-07-03 Richard Biener <rguenther@suse.de>
860 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
861 always set *dt. Dump vectype in vectype overload.
862 * dumpfile.h (dump_gimple_expr): New function.
863 (dump_gimple_expr_loc): Likewise.
864 * dumpfile.c (dump_gimple_expr): New function.
865 (dump_gimple_expr_loc): Likewise.
867 2018-07-02 Jeff Law <law@redhat.com>
869 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
870 the H8/300, H8/300H and H8/S variants into a single pattern.
871 (movhi_h8300, movqi_h8300hs): Similarly.
872 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
873 (QHI mode iterator): New.
875 * config/h8300/h8300.md: Remove trailing whitespace.
877 2018-07-02 Jim Wilson <jimw@sifive.com>
879 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
880 instead of emit_insn for interrupt returns.
881 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
882 (riscv_sret, riscv_uret): Likewise.
884 2018-07-02 David Malcolm <dmalcolm@redhat.com>
886 * pretty-print.c (selftest::test_pp_format): Move save and restore
887 of quotes to class auto_fix_quotes, and add an instance.
888 * selftest.c: Include "intl.h".
889 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
890 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
891 * selftest.h (selftest::auto_fix_quotes): New class.
893 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
895 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
896 (aarch64_sve_prepare_conditional_op): Remove.
897 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
898 Allow aarch64_simd_reg_or_zero as select operand; remove
899 the aarch64_sve_prepare_conditional_op call.
900 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
901 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
902 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
903 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
904 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
905 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
906 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
907 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
908 and a splitters to match all of the *_any patterns.
909 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
911 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
912 (SVE_COND_FP_BINARY_REV): Remove.
913 (sve_int_op_rev, sve_fp_op_rev): New.
914 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
915 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
916 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
917 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
918 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
919 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
920 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
921 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
922 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
924 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
925 Remove match_dup 1 from the inner unspec.
926 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
928 * config/aarch64/aarch64.md (movprfx): New attr.
929 (length): Default movprfx to 8.
930 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
931 (*madd<SVE_I>, *msub<SVE_I): Likewise.
932 (*<su>mul<SVE_I>3_highpart): Likewise.
933 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
934 (*v<ASHIFT><SVE_I>3): Likewise.
935 (*<su><MAXMIN><SVE_I>3): Likewise.
936 (*<su><MAXMIN><SVE_F>3): Likewise.
937 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
938 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
939 (*div<SVE_F>4): Likewise.
941 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
943 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
946 2018-07-02 Richard Biener <rguenther@suse.de>
948 PR tree-optimization/86363
949 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
950 memset argument refers to a non-variable address.
952 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
954 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
955 {PLUS,MINUS}_EXPR code to...
956 (adjust_symbolic_bound): ...here,
957 (combine_bound): ...here,
958 (set_value_range_with_overflow): ...and here.
960 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
962 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
964 (extract_range_from_abs_expr): ...here.
966 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
968 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
969 -fno-omit-frame-pointer when not optimizing.
971 2018-07-02 Martin Liska <mliska@suse.cz>
974 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
975 (propagate_nothrow): Likewise.
977 2018-07-02 Martin Liska <mliska@suse.cz>
980 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
982 2018-07-02 David Malcolm <dmalcolm@redhat.com>
984 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
985 function in r262149, changing "loc" param from source_location to
986 const dump_location_t &.
987 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
988 declaration, as above.
990 2018-07-01 Paul Koning <ni1d@arrl.net>
992 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
993 -munit-asm, -mgnu-asm, -mdec-asm.
994 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
995 (pdp11_output_labelref): New.
996 (pdp11_output_def): New.
997 (pdp11_output_addr_vec_elt): New.
998 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
999 %# and %@ format codes.
1000 (pdp11_option_override): New.
1001 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1002 (pdp11_output_ident): New.
1003 (pdp11_asm_named_section): New.
1004 (pdp11_asm_init_sections): New.
1005 (pdp11_file_start): New.
1006 (pdp11_file_end): New.
1007 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1008 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1009 %o, like %c but octal.
1010 (pdp11_option_override): New.
1011 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1013 (DATA_SECTION_ASM_OP): Ditto.
1014 (READONLY_DATA_SECTION_ASM_OP): New.
1015 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1016 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1017 (ASM_OUTPUT_LABELREF): Ditto.
1018 (ASM_OUTPUT_DEF): Ditto.
1019 (ASM_OUTPUT_EXTERNAL): New.
1020 (ASM_OUTPUT_SOURCE_FILENAME): New.
1021 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1022 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1023 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1024 %# and %@ format codes.
1025 * config/pdp11/pdp11.opt (mgnu-asm): New.
1026 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1027 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1028 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1030 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1032 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1033 dereferencing path[] beyond its length.
1035 (debug_all_paths): New.
1036 (rewire_first_differing_edge): New.
1037 (adjust_paths_after_duplication): New.
1038 (duplicate_thread_path): Call adjust_paths_after_duplication.
1040 (thread_through_all_blocks): Add new argument to
1041 duplicate_thread_path.
1043 2018-06-30 Jim Wilson <jimw@sifive.com>
1045 * config/riscv/predicates.md (p2m1_shift_operand): New.
1046 (high_mask_shift_operand): New.
1047 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1048 pattern using p2m1_shift_operand.
1049 (lshsi3_zero_extend_3+2): New combiner pattern using
1050 high_mask_shift_operand.
1052 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1054 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1056 (vect_recog_rotate_pattern): ...here.
1057 (vect_convert_input): Try to insert casts of invariants in the
1059 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1060 preheader to be empty.
1062 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1064 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1065 vector type. If given, install it in the new statement's
1067 (vect_element_precision): New function.
1068 (vect_unpromoted_value): New struct.
1069 (vect_unpromoted_value::vect_unpromoted_value): New function.
1070 (vect_unpromoted_value::set_op): Likewise.
1071 (vect_look_through_possible_promotion): Likewise.
1072 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1073 (vect_widened_op_tree, vect_convert_input): Likewise.
1074 (vect_convert_inputs, vect_convert_output): Likewise.
1075 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1076 to handle the optional cast of the multiplication result and
1077 vect_widened_op_tree to detect the widened multiplication itself.
1078 Do not require the input and output of promotion casts to have
1079 the same sign, but base the signedness of the operation on the
1080 input rather than the result. If the pattern includes two
1081 promotions, check that those promotions have the same sign.
1082 Do not restrict the MULT_EXPR handling to a double-width result;
1083 handle quadruple-width results and wider. Use vect_convert_inputs
1084 to convert the inputs to the common type.
1085 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1086 to handle the optional cast of the ABS result. Also allow a sign
1087 change or a sign extension between the ABS and MINUS.
1088 Use vect_widened_op_tree to detect the widened subtraction and use
1089 vect_convert_inputs to convert the inputs to the common type.
1090 (vect_handle_widen_op_by_const): Delete.
1091 (vect_recog_widen_op_pattern): New function.
1092 (vect_recog_widen_mult_pattern): Use it.
1093 (vect_recog_widen_shift_pattern): Likewise.
1094 (vect_recog_widen_sum_pattern): Use
1095 vect_look_through_possible_promotion to handle the promoted
1098 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1100 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1101 the containing gimple_seq *.
1102 * gimple-iterator.h (gsi_for_stmt): Declare it.
1103 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1104 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1105 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1106 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1107 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1109 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1111 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1112 statement being replaced is part of an existing pattern
1113 definition sequence, inserting the new pattern statements before
1115 (vect_pattern_recog_1): Don't return a bool. If the statement
1116 is already part of a pattern, instead apply pattern matching
1117 to the pattern definition statements. Don't clear the
1118 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1119 (vect_pattern_recog): Don't break after the first match;
1120 continue processing the pattern definition statements instead.
1121 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1123 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1125 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1126 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1127 (vect_recog_widen_sum_pattern): Use it.
1129 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1131 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1132 phi is not a pattern statement and has not been replaced by
1133 a pattern statement.
1134 * tree-vect-patterns.c (type_conversion_p): Don't check
1135 STMT_VINFO_IN_PATTERN_P.
1136 (vect_recog_vector_vector_shift_pattern): Likewise.
1137 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1138 the pattern statement rather than the original statement; check
1139 directly for a WIDEN_MULT_EXPR here.
1140 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1141 vect_is_simple_use to return the pattern statement rather
1142 than the original statement; use is_pattern_stmt_p to check
1143 for such a pattern statement.
1144 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1145 to return the pattern statement rather than the original statement;
1146 don't do the same transformation here.
1147 (vect_is_simple_use): If the defining statement has been replaced
1148 by a pattern statement, return the pattern statement instead.
1149 Remove the corresponding (local) transformation from the vectype
1152 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1154 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1155 end and default to null.
1156 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1157 (vectorizable_reduction): Update calls accordingly, dropping the
1158 gimple ** argument if the passed-back statement isn't needed.
1159 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1160 (vect_recog_rotate_pattern): Likewise.
1161 (vect_recog_mask_conversion_pattern): Likewise.
1162 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1163 (vect_mask_constant_operand_p): Likewise.
1164 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1165 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1166 (get_group_load_store_type, get_load_store_type): Likewise.
1167 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1168 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1169 (vectorizable_conversion, vectorizable_assignment): Likewise.
1170 (vectorizable_shift, vectorizable_operation): Likewise.
1171 (vectorizable_store, vect_is_simple_cond): Likewise.
1172 (vectorizable_condition, vectorizable_comparison): Likewise.
1173 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1174 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1175 and move it to the end. Cope with null def_stmt_outs.
1177 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1179 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1181 2018-06-29 Jeff Law <law@redhat.com>
1183 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1184 displacements for TARGET_V850E2V3 and newer.
1185 (TARGET_LRA_P): Remove. Defaults to LRA now.
1186 * config/v850/v850.md (sign23byte_load): Remove.
1187 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1188 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1190 2018-06-29 Martin Liska <mliska@suse.cz>
1193 * coverage.c (coverage_init): Mangle full path name.
1194 * doc/invoke.texi: Document the change.
1195 * gcov-io.c (mangle_path): New.
1196 * gcov-io.h (mangle_path): Likewise.
1197 * gcov.c (mangle_name): Use mangle_path for path mangling.
1199 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1201 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1202 if starting source register is not even.
1204 2018-06-29 Martin Liska <mliska@suse.cz>
1206 PR tree-optimization/86263
1207 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1208 Make edge redirection.
1210 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1212 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1213 (dump_scope_depth): New variable.
1214 (get_dump_scope_depth): New function.
1215 (dump_begin_scope): New function.
1216 (dump_end_scope): New function.
1217 * dumpfile.h (get_dump_scope_depth): New declaration.
1218 (dump_begin_scope): New declaration.
1219 (dump_end_scope): New declaration.
1220 (class auto_dump_scope): New class.
1221 (AUTO_DUMP_SCOPE): New macro.
1222 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1225 2018-06-29 Richard Biener <rguenther@suse.de>
1227 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1228 compute_all_dependences succeeds.
1229 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1230 exceed --param loop-max-datarefs-for-datadeps.
1232 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1234 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1236 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1239 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1240 alternative 0 in preferred_for_speed attribute.
1242 2018-06-28 Paul Koning <ni1d@arrl.net>
1244 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1245 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1246 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1247 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1248 for shift insn patterns.
1250 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1252 * cgraph.c (cgraph_node::get_body): Replace assignments to
1253 "dump_file" with calls to set_dump_file.
1254 * dumpfile.c (alt_dump_file): Make static, and group with...
1255 (alt_flags): ...this definition.
1256 (dumps_are_enabled): New variable.
1257 (refresh_dumps_are_enabled): New function.
1258 (set_dump_file): New function.
1259 (set_alt_dump_file): New function.
1260 (gcc::dump_manager::dump_start): Replace assignments to
1261 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1263 (gcc::dump_manager::dump_finish): Likewise.
1264 * dumpfile.h (alt_dump_file): Delete decl.
1265 (dumps_are_enabled): New variable decl.
1266 (set_dump_file): New function decl.
1267 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1269 * tree-nested.c (lower_nested_functions): Replace assignments to
1270 "dump_file" with calls to set_dump_file.
1272 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1274 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1275 goto_locus of each outgoing edge of each basic block.
1277 2018-06-28 Richard Biener <rguenther@suse.de>
1279 * dwarf2out.c (decl_scope_table): Remove.
1280 (push_decl_scope): Likewise.
1281 (pop_decl_scope): Likewise.
1282 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1283 (gen_struct_or_union_type_die): Likewise.
1284 (gen_tagged_type_die): Likewise.
1285 (dwarf2out_init): Do not initialize decl_scope_table.
1286 (dwarf2out_c_finalize): Do not free it.
1288 2018-06-28 Richard Biener <rguenther@suse.de>
1290 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1291 deciding whether to not re-use a DIE.
1293 2018-06-28 Richard Biener <rguenther@suse.de>
1295 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1296 DW_AT_abstract_origin attribute.
1298 2018-06-28 Martin Liska <mliska@suse.cz>
1300 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1301 Use newly introduced constants.
1302 * tree-switch-conversion.h (struct jump_table_cluster):
1303 Define max_ratio_for_size and max_ratio_for_speed.
1305 2018-06-28 Martin Liska <mliska@suse.cz>
1307 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1308 Add new checking assert to catch invalid state.
1309 (jump_table_cluster::can_be_handled): Handle single case
1311 (jump_table_cluster::is_beneficial): Bail out for such case.
1312 (bit_test_cluster::find_bit_tests):
1313 Add new checking assert to catch invalid state.
1314 (bit_test_cluster::can_be_handled): Handle single case
1316 (bit_test_cluster::is_beneficial): Bail out for such case.
1317 (switch_decision_tree::analyze_switch_statement):
1320 2018-06-28 Martin Liska <mliska@suse.cz>
1322 * common.opt: Introduce -completion option.
1323 * gcc.c (driver_handle_option): Handle it.
1324 (driver::main): Print completions if completion
1326 * opt-suggestions.c (option_proposer::get_completions):
1328 (option_proposer::suggest_completion): Likewise.
1329 (option_proposer::find_param_completions): Likewise.
1330 (verify_autocompletions): Likewise.
1331 (test_completion_valid_options): Likewise.
1332 (test_completion_valid_params): Likewise.
1333 (in_completion_p): Likewise.
1334 (empty_completion_p): Likewise.
1335 (test_completion_partial_match): Likewise.
1336 (test_completion_garbage): Likewise.
1337 (opt_proposer_c_tests): Likewise.
1338 * opt-suggestions.h: Declare new functions.
1339 * opts.c (common_handle_option): Handle OPT__completion_.
1340 * selftest-run-tests.c (selftest::run_tests): Add
1341 opt_proposer_c_tests.
1342 * selftest.c (assert_str_startswith): New.
1343 * selftest.h (assert_str_startswith): Likewise.
1344 (opt_proposer_c_tests): New.
1345 (ASSERT_STR_STARTSWITH): Likewise.
1347 2018-06-28 Martin Liska <mliska@suse.cz>
1349 * Makefile.in: Add opt-suggestions.o.
1350 * gcc-main.c: Include opt-suggestions.h.
1351 * gcc.c (driver::driver): Likewise.
1352 (driver::~driver): Remove m_option_suggestions.
1353 (driver::build_option_suggestions): Moved to option_proposer.
1354 (driver::suggest_option): Likewise.
1355 (driver::handle_unrecognized_options): Use option_proposer.
1356 * gcc.h (class driver): Add new memver m_option_proposer.
1357 * opt-suggestions.c: New file.
1358 * opt-suggestions.h: New file.
1360 2018-06-28 Martin Liska <mliska@suse.cz>
1362 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1363 (auto_string_vec::~auto_string_vec): Likewise.
1365 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1367 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1368 prevent_decl_creation_for_types fields up and add reset_location field.
1369 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1370 statement if id->reset_location is true.
1371 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1372 id->reset_location is true.
1373 (copy_phis_for_bb): Force input_location on the arguments if
1374 id->reset_location is true.
1375 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1376 is set on the function to be inlined.
1378 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1380 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1382 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1384 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1385 registers for Pmode.
1386 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1387 hard registers for the clobbered pseudo.
1389 2018-06-27 Paul Koning <ni1d@arrl.net>
1391 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1392 mutually exclusive options.
1393 * config/pdp11/constraints.md (h): New constraint.
1394 (O): Update definition to match shift code generation.
1395 (D): New constraint.
1396 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1398 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1400 (output_jump): Change arguments.
1401 (pdp11_fixed_cc_regs): New function.
1402 (pdp11_cc_mode): Ditto.
1403 (pdp11_expand_shift): Ditto.
1404 (pdp11_assemble_shift): Ditto.
1405 (pdp11_small_shift): Ditto.
1406 (pdp11_branch_cost): Remove.
1407 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1409 (pdp11_register_move_cost): Update for CC registers.
1410 (pdp11_rtx_costs): Add case for LSHIFTRT.
1411 (pdp11_output_jump): Add CCNZ mode conditional branches.
1412 (notice_update_cc_on_set): Remove.
1413 (pdp11_cc_mode): New function.
1414 (simple_memory_operand): Correct pre/post decrement case.
1415 (no_side_effect_operand): New function.
1416 (pdp11_regno_reg_class): Add CC_REGS class.
1417 (pdp11_fixed_cc_regs): New function.
1418 (pdp11_small_shift): New function.
1419 (pdp11_expand_shift): New function to expand shift insns.
1420 (pdp11_assemble_shift): New function to output shifts.
1421 (pdp11_branch_cost): Remove.
1422 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1423 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1424 (WCHAR_TYPE): Ditto.
1425 (PTRDIFF_TYPE): Ditto.
1426 (ADJUST_INSN_LENGTH): New macro.
1427 (FIXED_REGISTERS): Add CC registers.
1428 (CALL_USED_REGISTERS): Ditto.
1430 (REG_CLASS_NAMES): Ditto.
1431 (REG_CLASS_CONTENTS): Ditto.
1432 (SELECT_CC_MODE): Use new function.
1433 (TARGET_FLAGS_REGNUM): New macro.
1434 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1435 (cc0_reg_rtx): Remove.
1436 (CC_STATUS_MDEP): Remove.
1437 (CC_STATUS_MDEFP_INIT): Remove.
1438 (CC_IN_FPU): Remove.
1439 (NOTICE_UPDATE_CC): Remove.
1440 (REGISTER_NAMES): Add CC registers.
1441 (BRANCH_COST): Change to constant 1.
1442 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1444 * config/pdp11/pdp11.opt (mbcopy): Remove.
1445 (mbcopy-builtin): Remove.
1446 (mbranch-cheap): Remove.
1447 (mbranch-expensive): Remove.
1448 * config/pdp11/predicates.md (expand_shift_operand): Update to
1449 match shift code generation.
1450 (ccnz_operator): New predicate.
1451 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1452 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1453 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1455 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1456 description of O constraint.
1458 2018-06-27 Jeff Law <law@redhat.com>
1459 Austin Law <austinklaw@gmail.com>
1461 * config/v850/v850.md (addsi3_set_flags): New pattern.
1462 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1463 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1464 (zero_extendhisi2_v850_set_flags): Likewise.
1465 (zero_extendqisi2_v850_set_flags): Likewise.
1466 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1467 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1468 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1470 * config/v850/v850-protos.h (notice_update_cc): Remove.
1471 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1472 (v850_print_operand): Handle 'D' and "d".
1473 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1474 Add handling of arithmetic/logical operations compared against zero.
1475 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1476 Do not look at v850_compare_op, instead get mode from last argument.
1477 (v850_gen_compare): Remove
1478 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1479 after reload for prologue insns.
1480 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1482 (construct_save_jarl): Likewise.
1483 (TARGET_FLAGS_REGNUM): Define.
1484 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1485 (NOTICE_UPDATE_CC): Remove.
1486 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1487 than cc0. Conditionalize on reload_completed.
1488 (cmpsi_insn, setfcc_insn): Likewise.
1489 (tst1 splitter): Turn into define_and_split which sets the flags
1491 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1492 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1493 (cstoresf4, cstoredf4): Clobber the flags.
1494 (cmpsi, cmpsf, cmpdf): Remove expanders.
1495 (setf_insn): Remove pattern.
1496 (addsi3): Turn into define_and_split which clobbers the flags after
1497 reload and a suitable pattern (addsi3_clobber_flags) for use after
1499 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1500 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1501 (ashrsi3, ashrsi3_v850e2): Likewise.
1502 (bins): Clobber the flags.
1503 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1504 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1505 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1506 (call_value_internal_short, call_value_internal_long): Likewise.
1507 (callt_save_interrupt, callt_return_interrupt): Likewise.
1508 (save_interrupt, return_interrupt): Likewise.
1509 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1510 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1511 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1512 (All FP comparisons): Only allow after reload has completed.
1514 (divh, divhu): Tweak output template.
1515 (branch_z_normal, branch_z_invert): Remove
1516 (branch_nz_normal, branch_nz_invert): Likewise.
1517 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1519 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1520 * config/v850/v850.c (notice_update_cc): Remove.
1521 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1522 (CC_NO_CARRY): Likewise.
1523 (NOTICE_UPDATE_CC): Define to nothing.
1524 * config/v850/v850.md: Remove block comment on cc0 handling
1525 Remove "cc" attribute from all patterns. Remove cc_status handling
1526 from all patterns. Minor formatting fixes.
1528 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1530 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1531 (cortex-a76.cortex-a55): Likewise.
1532 * config/aarch64/aarch64-tune.md: Regenerate.
1533 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1534 cortex-a76.cortex-a55.
1536 2018-06-27 Jeff Law <law@redhat.com>
1538 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1539 (MULTILIB_DIRNAMES): Similarly.
1541 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1543 * gimple.h (gimple_return_retbnd): Delete.
1544 (gimple_return_set_retbnd): Likewise.
1545 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1546 gimple_return_set_retbnd.
1547 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1548 gimple_return_retbnd and adjust.
1549 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1550 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1551 Explicitly return NULL in a couple more cases. Move assertion
1552 on debug statements and remove unreachable code.
1553 (reset_debug_binding): Do not test id->retbnd.
1554 (expand_call_inline): Do not set it.
1556 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1558 * configure.ac: Add --disable-gcov option.
1559 * configure: Regenerate.
1560 * Makefile.in: Honour @enable_gcov@.
1561 * doc/install.texi: Document --disable-gcov.
1563 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1565 * config/arm/arm-cpus.in (cortex-a76): New entry.
1566 (cortex-a76.cortex-a55): Likewise.
1567 * config/arm/arm-tables.opt: Regenerate.
1568 * config/arm/arm-tune.md: Likewise.
1569 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1570 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1571 cortex-a76.cortex-a55.
1573 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1576 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1578 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1580 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1582 (EPILOGUE_USES): Likewise.
1584 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1586 * tree-inline.c (remap_location): New function extracted from...
1587 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1588 (copy_phis_for_bb): ...here. Call remap_location.
1589 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1591 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1593 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1594 unaligned vsx for 16B memset.
1596 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1599 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1600 ieee128_float_type_node to long_double_type_node unless
1601 TARGET_LONG_DOUBLE_128 is set.
1603 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1605 * cfgloop.c (get_loop_location): Convert return type from
1606 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1607 by implicit construction from rtx_insn *, and using
1608 dump_user_location_t::from_function_decl for the fallback case.
1609 * cfgloop.h (get_loop_location): Convert return type from
1610 location_t to dump_user_location_t.
1611 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1612 dump_printf_loc to pass in a dump_location_t rather than a
1613 location_t, via the gimple stmt.
1614 * coverage.c (get_coverage_counts): Update calls to
1615 dump_printf_loc to pass in dump_location_t rather than a
1617 * doc/optinfo.texi (Dump types): Convert example of
1618 dump_printf_loc from taking "locus" to taking "insn". Update
1619 description of the "_loc" calls to cover dump_location_t.
1620 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1622 (dump_user_location_t::dump_user_location_t): New constructors,
1623 from gimple *stmt and rtx_insn *.
1624 (dump_user_location_t::from_function_decl): New function.
1625 (dump_loc): Make static.
1626 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1627 const dump_location_t &.
1628 (dump_generic_expr_loc): Delete.
1629 (dump_printf_loc): Convert param "loc" from location_t to
1630 const dump_location_t &.
1631 (selftest::test_impl_location): New function.
1632 (selftest::dumpfile_c_tests): New function.
1633 * dumpfile.h: Include "profile-count.h".
1634 (class dump_user_location_t): New class.
1635 (struct dump_impl_location_t): New struct.
1636 (class dump_location_t): New class.
1637 (dump_printf_loc): Convert 2nd param from source_location to
1638 const dump_location_t &.
1639 (dump_generic_expr_loc): Delete.
1640 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1641 const dump_location_t &.
1642 * gimple-fold.c (fold_gimple_assign): Update call to
1643 dump_printf_loc to pass in a dump_location_t rather than a
1644 location_t, via the gimple stmt.
1645 (gimple_fold_call): Likewise.
1646 * gimple-loop-interchange.cc
1647 (loop_cand::analyze_iloop_reduction_var): Update for change to
1648 check_reduction_path.
1649 (tree_loop_interchange::interchange): Update for change to
1651 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1652 change in return-type of find_loop_location.
1653 (graphite_regenerate_ast_isl): Likewise.
1654 * graphite-optimize-isl.c (optimize_isl): Likewise.
1655 * graphite.c (graphite_transform_loops): Likewise.
1656 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1657 pass in a dump_location_t rather than a location_t, via the
1659 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1660 * ipa.c (walk_polymorphic_call_targets): Likewise.
1661 * loop-unroll.c (report_unroll): Convert "locus" param from
1662 location_t to dump_location_t.
1663 (decide_unrolling): Update for change to get_loop_location's
1665 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1666 location_t to dump_user_location_t.
1667 (grid_find_single_omp_among_assignments_1): Updates calls to
1668 dump_printf_loc to pass in a dump_location_t rather than a
1669 location_t, via the gimple stmt.
1670 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1671 location_t to dump_location_t. Updates calls to dump_printf_loc
1672 to pass in a dump_location_t rather than a location_t, via the
1674 (grid_inner_loop_gridifiable_p): Likewise.
1675 (grid_dist_follows_simple_pattern): Likewise.
1676 (grid_gfor_follows_tiling_pattern): Likewise.
1677 (grid_target_follows_gridifiable_pattern): Likewise.
1678 (grid_attempt_target_gridification): Convert initialization
1679 of local "grid" from memset to zero-initialization; FIXME: does
1680 this require C++11? Update call to dump_printf_loc to pass in a
1681 optinfo_location rather than a location_t, via the gimple stmt.
1682 * profile.c (read_profile_edge_counts): Updates call to
1683 dump_printf_loc to pass in a dump_location_t rather than a
1685 (compute_branch_probabilities): Likewise.
1686 * selftest-run-tests.c (selftest::run_tests): Call
1688 * selftest.h (dumpfile_c_tests): New decl.
1689 * tree-loop-distribution.c (pass_loop_distribution::execute):
1690 Update for change in return type of find_loop_location.
1691 * tree-parloops.c (parallelize_loops): Likewise.
1692 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1693 "locus" from location_t to dump_user_location_t.
1694 (canonicalize_loop_induction_variables): Likewise.
1695 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1696 for change in return type of find_loop_location.
1697 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1698 to dump_printf_loc to pass in a dump_location_t rather than a
1699 location_t, via the stmt.
1700 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1702 * tree-vect-loop-manip.c (find_loop_location): Convert return
1703 type from source_location to dump_user_location_t.
1704 (vect_do_peeling): Update for above change.
1705 (vect_loop_versioning): Update for change in type of
1707 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1708 from location_t to dump_user_location_t.
1709 (vect_estimate_min_profitable_iters): Update for change in type
1711 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1712 location_t to dump_location_t.
1713 (vect_slp_bb): Update for change in type of vect_location.
1714 * tree-vectorizer.c (vect_location): Convert from source_location
1715 to dump_user_location_t.
1716 (try_vectorize_loop_1): Update for change in vect_location's type.
1717 (vectorize_loops): Likewise.
1718 (increase_alignment): Likewise.
1719 * tree-vectorizer.h (vect_location): Convert from source_location
1720 to dump_user_location_t.
1721 (find_loop_location): Convert return type from source_location to
1722 dump_user_location_t.
1723 (check_reduction_path): Convert 1st param from location_t to
1724 dump_user_location_t.
1725 * value-prof.c (check_counter): Update call to dump_printf_loc to
1726 pass in a dump_user_location_t rather than a location_t; update
1727 call to error_at for change in type of "locus".
1728 (check_ic_target): Update call to dump_printf_loc to
1729 pass in a dump_user_location_t rather than a location_t, via the
1732 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1734 * config/s390/s390.h (enum processor_flags): Do not use
1735 default tune parameter when -march was specified.
1737 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1740 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1741 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1744 2018-06-26 Richard Biener <rguenther@suse.de>
1746 PR tree-optimization/86287
1748 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1749 (vect_analyze_loop): Initialize n_stmts.
1751 2018-06-26 Richard Biener <rguenther@suse.de>
1754 * fold-const.c (fold_convertible_p): Pointer extension
1757 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1760 * dwarf2out.c (loc_list_has_views): Adjust comments.
1761 (dw_loc_list): Split single cross-partition range with
1764 2018-06-25 Jeff Law <law@redhat.com>
1766 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1767 on -mbig-switch by default.
1769 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1771 (const_float_0_operand): Remove unused predicate.
1772 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1773 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1774 (recipsf2): New expander. Original pattern now called
1776 (recipdf2, recipdf2_insn): Similarly.
1777 (rsqrtsf2, rsqrtsf2_insn): Similarly
1778 (rsqrtdf2, rsqrtdf2_insn): Similarly
1780 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1782 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1783 Simplify logic for FreeBSD (twice).
1785 2018-06-25 Martin Sebor <msebor@redhat.com>
1787 PR tree-optimization/86204
1788 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1789 a strnlen result if it's less than the length of the string.
1791 2018-06-25 Martin Sebor <msebor@redhat.com>
1793 PR tree-optimization/85700
1794 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1795 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1796 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1798 2018-06-25 Martin Sebor <msebor@redhat.com>
1800 * doc/extend.texi (Zero-length arrays): Update and clarify.
1802 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1804 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1805 added IEEE/IBM long double multilib support on PowerPC little
1806 endian Linux systems.
1807 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1808 (MULTILIB_DEFAULTS): Likewise.
1809 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1811 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1812 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1813 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1815 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1818 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1819 (REORDER_45): Likewise.
1821 2018-06-25 Jeff Law <law@redhat.com>
1823 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1824 dividend to 32 bits. Adjust length.
1825 (udivmodhi4): Cleanup output template. Fix length.
1827 2018-06-25 Carl Love <cel@us.ibm.com>
1829 * config/rs6000/vsx.md: Change word selector to prefered location.
1831 2018-06-25 Richard Biener <rguenther@suse.de>
1833 PR tree-optimization/86304
1834 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1835 epilogue-if-converted loops as well.
1837 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1839 * lto-section-out.c (lto_begin_section): Do not print section
1840 name for noaddr and unnumbered dumps.
1842 2018-06-25 Richard Biener <rguenther@suse.de>
1844 * tree-vectorizer.h (struct vec_info_shared): New structure
1845 with parts split out from struct vec_info and loop_nest from
1846 struct _loop_vec_info.
1847 (struct vec_info): Adjust accordingly.
1848 (struct _loop_vec_info): Likewise.
1849 (LOOP_VINFO_LOOP_NEST): Adjust.
1850 (LOOP_VINFO_DATAREFS): Likewise.
1851 (LOOP_VINFO_DDRS): Likewise.
1852 (struct _bb_vec_info): Likewise.
1853 (BB_VINFO_DATAREFS): Likewise.
1854 (BB_VINFO_DDRS): Likewise.
1855 (struct _stmt_vec_info): Add dr_aux member.
1856 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1857 (DR_MISALIGNMENT_UNINITIALIZED): New.
1858 (set_dr_misalignment): Adjust.
1859 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1860 (vect_analyze_loop): Adjust prototype.
1861 (vect_analyze_loop_form): Likewise.
1862 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1863 Compute dependences lazily.
1864 (vect_record_base_alignments): Use shared datarefs/ddrs.
1865 (vect_verify_datarefs_alignment): Likewise.
1866 (vect_analyze_data_refs_alignment): Likewise.
1867 (vect_analyze_data_ref_accesses): Likewise.
1868 (vect_analyze_data_refs): Likewise.
1869 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1870 constructor parameter for shared part.
1871 (vect_analyze_loop_form): Pass in shared part and adjust.
1872 (vect_analyze_loop_2): Pass in storage for the number of
1873 stmts. Move loop nest finding to the caller. Compute
1875 (vect_analyze_loop): Pass in shared part.
1876 (vect_transform_loop): Verify shared datarefs are unchanged.
1877 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1878 constructor parameter for shared part.
1879 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1880 (vect_slp_bb): Verify shared datarefs are unchanged before
1882 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1884 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1885 DR_MISALIGNMENT_UNINITIALIZED.
1886 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1887 parameter for shared part.
1888 (vec_info::~vec_info): Adjust.
1889 (vec_info_shared::vec_info_shared): New.
1890 (vec_info_shared::~vec_info_shared): Likewise.
1891 (vec_info_shared::save_datarefs): Likewise.
1892 (vec_info_shared::check_datarefs): Likewise.
1893 (try_vectorize_loop_1): Construct shared part live for analyses
1894 of a single loop for multiple vector sizes.
1895 * tree-parloops.c (gather_scalar_reductions): Adjust.
1897 2018-06-25 Richard Biener <rguenther@suse.de>
1899 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1900 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1901 (vect_analyze_data_refs): Remove similar code from here and
1902 simplify accordingly.
1904 2018-06-25 Richard Biener <rguenther@suse.de>
1906 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1907 for reverse storage order accesses rather than asserting
1908 they cannot happen here.
1910 2018-06-25 Tom de Vries <tdevries@suse.de>
1913 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1914 Use data16 instead of .byte for insn prefix.
1916 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1919 * parser.c (make_char_string_pack): Pass this literal chars
1920 through cpp_interpret_string.
1921 (cp_parser_userdef_numeric_literal): Check the result of
1922 make_char_string_pack.
1924 2018-06-24 Maya Rashish <coypu@sdf.org>
1926 * ginclude/stddef.h: Simplify conditions around avoiding
1927 re-definition of __size_t.
1929 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1931 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1932 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1934 2018-06-22 Maya Rashish <coypu@sdf.org>
1936 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1937 made to OpenBSD and NetBSD through the years.
1939 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1941 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1942 behavior of vec_pack (vector double, vector double) to match
1943 behavior of vec_float2 (vector double, vector double).
1945 2018-06-22 Olivier Hainque <hainque@adacore.com>
1947 * gimplify.c (gimplify_function_tree): Prevent creation
1948 of a trampoline for the address of the current function
1949 passed to entry/exit instrumentation hooks.
1951 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1954 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1957 2018-06-22 Martin Liska <mliska@suse.cz>
1959 PR tree-optimization/86263
1960 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1961 Bail out if is_enabled is false.
1962 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1964 (jump_table_cluster::is_enabled): New function.
1966 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1968 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1969 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1970 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1971 (lto_input_ts_binfo_tree_pointers): Likewise.
1972 * tree-streamer-out.c (streamer_write_tree_bitfields,
1973 write_ts_binfo_tree_pointers): Likewise.
1974 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1976 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1978 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1980 2018-06-22 Martin Liska <mliska@suse.cz>
1982 * symbol-summary.h (get): Make it pure and inline move
1983 functionality from ::get function.
1984 (get): Remove and inline into ::get and ::get_create.
1985 (get_create): Move code from ::get function.
1987 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1990 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1991 -x assembler-with-cpp.
1993 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1995 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1996 _FILE_OFFSET_BITS=64 for C++.
1998 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2000 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2001 conversion insn that shows up when pr85657-3.c is compiled using
2002 IEEE 128-bit long double.
2003 (neg<mode>2_internal): Use the correct mode to check whether the
2004 mode is IBM extended.
2005 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2006 multiply and divide external functions from being created more
2009 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2011 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2013 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2014 the edge can be forwarded.
2015 (cfg_layout_merge_blocks): Likewise.
2017 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2019 * except.c (finish_eh_generation): Commit edge insertions only after
2020 the EH edges have been redirected from post-landing to landing pads.
2022 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2024 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2025 create_tmp_var_for to create the FRAME decl.
2026 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2028 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2030 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2031 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2032 debug statement when resetting its value.
2033 (expand_call_inline): Copy the locus of the call onto the assignment
2034 of the return value, if any. Use local variable in more cases.
2036 2018-06-21 Martin Liska <mliska@suse.cz>
2038 * ipa-pure-const.c (propagate_nothrow): Use
2039 funct_state_summaries->get.
2040 (dump_malloc_lattice): Likewise.
2041 (propagate_malloc): Likewise.
2043 2018-06-21 Richard Biener <rguenther@suse.de>
2045 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2046 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2047 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2049 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2050 BLOCK_ABSTRACT_ORIGIN unconditionally.
2052 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2054 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2056 * ipa-reference.c (ipa_reference_c_finalize): Delete
2057 ipa_ref_opt_sum_summaries and set it to NULL.
2059 2018-06-21 Tom de Vries <tdevries@suse.de>
2061 PR tree-optimization/85859
2062 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2063 test with comment from bb_no_side_effects_p.
2065 2018-06-21 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/86232
2068 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2069 max for constant niter.
2071 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2073 * config/aarch64/aarch64-simd.md
2074 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2076 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2078 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2079 Make opernads of the unspec commutative.
2081 2018-06-21 Richard Biener <rguenther@suse.de>
2083 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2084 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2085 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2086 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2088 (vect_record_base_alignments): Do not record base alignment
2089 for gathers or scatters.
2090 (vect_compute_data_ref_alignment): Drop return value that is always
2091 true. Bail out early for gathers or scatters.
2092 (vect_enhance_data_refs_alignment): Bail out early for gathers
2094 (vect_find_same_alignment_drs): Likewise.
2095 (vect_analyze_data_refs_alignment): Remove dead code.
2096 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2097 (vect_analyze_data_refs): For possible gathers or scatters do
2098 not create an alternate DR, just check their possible validity
2099 and mark them. Adjust DECL_NONALIASED handling to not rely
2101 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2102 update inits of gathers or scatters.
2103 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2104 Also copy gather/scatter flag to pattern vinfo.
2106 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2108 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2109 behavior of vec_packsu (vector unsigned long long, vector unsigned
2110 long long) to match behavior of vec_packs with same signature.
2112 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2113 Thomas Schwinge <thomas@codesourcery.com>
2114 Cesar Philippidis <cesar@codesourcery.com>
2116 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2117 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2118 (gimplify_adjust_omp_clauses): Likewise.
2119 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2120 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2121 (gimplify_omp_target_update): Update handling of acc update and
2123 * omp-low.c (install_var_field): Remove unused parameter
2124 base_pointers_restrict.
2125 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2126 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2128 (omp_target_base_pointers_restrict_p): Delete.
2129 (scan_omp_target): Update call to scan_sharing_clauses.
2130 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2132 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2133 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2134 (convert_local_omp_clauses): Likewise.
2135 * tree-pretty-print.c (dump_omp_clause): Likewise.
2136 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2138 (omp_clause_code_name): Likewise.
2140 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2143 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2146 PR tree-optimization/86231
2147 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2148 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2150 2018-06-20 Tom de Vries <tdevries@suse.de>
2152 PR tree-optimization/86097
2153 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2154 iv type if signedness of iv type is not the same as that of *nit.
2156 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2158 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2159 EDGE_EH edges, verify they are all EDGE_EH.
2161 2018-06-20 Maya Rashish <coypu@sdf.org>
2163 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2165 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2167 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2168 * config/aarch64/aarch64.c (xgene1_tunings): Add
2169 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2170 (aarch64_mode_valid_for_sched_fusion_p):
2171 Allow 16-byte modes.
2172 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2173 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2175 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2177 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2178 * config/aarch64/iterators.md (VQ2): New mode iterator.
2180 2018-06-20 Martin Liska <mliska@suse.cz>
2182 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2183 Change default ratio from 10 to 8.
2185 2018-06-20 Martin Liska <mliska@suse.cz>
2187 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2189 (bit_test_cluster::find_bit_tests): Likewise.
2190 (switch_decision_tree::analyze_switch_statement): Find clusters.
2191 * tree-switch-conversion.h (struct jump_table_cluster): Document
2194 2018-06-20 Martin Liska <mliska@suse.cz>
2196 * tree-switch-conversion.c (switch_conversion::collect):
2197 Record m_uniq property.
2198 (switch_conversion::expand): Bail out for special conditions.
2199 (group_cluster::~group_cluster): New.
2200 (group_cluster::group_cluster): Likewise.
2201 (group_cluster::dump): Likewise.
2202 (jump_table_cluster::emit): New.
2203 (switch_decision_tree::fix_phi_operands_for_edges): New.
2204 (struct case_node): Remove struct.
2205 (jump_table_cluster::can_be_handled): New.
2206 (case_values_threshold): Moved to header.
2207 (reset_out_edges_aux): Likewise.
2208 (jump_table_cluster::is_beneficial): New.
2209 (bit_test_cluster::can_be_handled): Likewise.
2210 (add_case_node): Remove.
2211 (bit_test_cluster::is_beneficial): New.
2212 (case_bit_test::cmp): New.
2213 (bit_test_cluster::emit): New.
2214 (expand_switch_as_decision_tree_p): Remove.
2215 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2216 (fix_phi_operands_for_edge): Likewise.
2217 (switch_decision_tree::analyze_switch_statement): New.
2218 (compute_cases_per_edge): Move ...
2219 (switch_decision_tree::compute_cases_per_edge): ... here.
2220 (try_switch_expansion): Likewise.
2221 (switch_decision_tree::try_switch_expansion): Likewise.
2222 (record_phi_operand_mapping): Likewise.
2223 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2224 (emit_case_decision_tree): Likewise.
2225 (switch_decision_tree::emit): Likewise.
2226 (balance_case_nodes): Likewise.
2227 (switch_decision_tree::balance_case_nodes): Likewise.
2228 (dump_case_nodes): Likewise.
2229 (switch_decision_tree::dump_case_nodes): Likewise.
2230 (emit_jump): Likewise.
2231 (switch_decision_tree::emit_jump): Likewise.
2232 (emit_cmp_and_jump_insns): Likewise.
2233 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2234 (emit_case_nodes): Likewise.
2235 (switch_decision_tree::emit_case_nodes): Likewise.
2236 (conditional_probability): Remove.
2237 * tree-switch-conversion.h (enum cluster_type): New.
2239 (struct cluster): Likewise.
2240 (cluster::cluster): Likewise.
2241 (struct simple_cluster): Likewise.
2242 (simple_cluster::simple_cluster): Likewise.
2243 (struct group_cluster): Likewise.
2244 (struct jump_table_cluster): Likewise.
2245 (struct bit_test_cluster): Likewise.
2246 (struct min_cluster_item): Likewise.
2247 (struct case_tree_node): Likewise.
2248 (case_tree_node::case_tree_node): Likewise.
2249 (jump_table_cluster::case_values_threshold): Likewise.
2250 (struct case_bit_test): Likewise.
2251 (struct switch_decision_tree): Likewise.
2252 (struct switch_conversion): Likewise.
2253 (switch_decision_tree::reset_out_edges_aux): Likewise.
2255 2018-06-20 Martin Liska <mliska@suse.cz>
2257 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2258 (hoist_edge_and_branch_if_true): Likewise.
2259 (expand_switch_using_bit_tests_p): Likewise.
2260 (struct case_bit_test): Likewise.
2261 (case_bit_test_cmp): Likewise.
2262 (emit_case_bit_tests): Likewise.
2263 (switch_conversion::switch_conversion): New class.
2264 (struct switch_conv_info): Remove old struct.
2265 (collect_switch_conv_info): More to ...
2266 (switch_conversion::collect): ... this.
2267 (check_range): Likewise.
2268 (switch_conversion::check_range): Likewise.
2269 (check_all_empty_except_final): Likewise.
2270 (switch_conversion::check_all_empty_except_final): Likewise.
2271 (check_final_bb): Likewise.
2272 (switch_conversion::check_final_bb): Likewise.
2273 (create_temp_arrays): Likewise.
2274 (switch_conversion::create_temp_arrays): Likewise.
2275 (free_temp_arrays): Likewise.
2276 (gather_default_values): Likewise.
2277 (switch_conversion::gather_default_values): Likewise.
2278 (build_constructors): Likewise.
2279 (switch_conversion::build_constructors): Likewise.
2280 (constructor_contains_same_values_p): Likewise.
2281 (switch_conversion::contains_same_values_p): Likewise.
2282 (array_value_type): Likewise.
2283 (switch_conversion::array_value_type): Likewise.
2284 (build_one_array): Likewise.
2285 (switch_conversion::build_one_array): Likewise.
2286 (build_arrays): Likewise.
2287 (switch_conversion::build_arrays): Likewise.
2288 (gen_def_assigns): Likewise.
2289 (switch_conversion::gen_def_assigns): Likewise.
2290 (prune_bbs): Likewise.
2291 (switch_conversion::prune_bbs): Likewise.
2292 (fix_phi_nodes): Likewise.
2293 (switch_conversion::fix_phi_nodes): Likewise.
2294 (gen_inbound_check): Likewise.
2295 (switch_conversion::gen_inbound_check): Likewise.
2296 (process_switch): Use the newly created class.
2297 (switch_conversion::expand): New.
2298 (switch_conversion::~switch_conversion): New.
2299 * tree-switch-conversion.h: New file.
2301 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2303 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2304 tree-vect-patterns.c.
2305 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2306 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2307 (vect_recog_sad_pattern): Likewise.
2308 (vect_recog_widen_sum_pattern): Likewise.
2309 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2310 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2311 (vect_recog_rotate_pattern): Likewise.
2312 (vect_recog_mult_pattern): Likewise.
2313 (vect_recog_vector_vector_shift_pattern): Likewise.
2314 (vect_recog_divmod_pattern): Likewise.
2315 (vect_recog_mixed_size_cond_pattern): Likewise.
2316 (vect_recog_bool_pattern): Likewise.
2317 (vect_recog_mask_conversion_pattern): Likewise.
2318 (vect_try_gather_scatter_pattern): Likewise.
2319 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2320 (vect_recog_over_widening_pattern): Likewise.
2321 (vect_recog_gather_scatter_pattern): Likewise.
2322 (vect_recog_func_ptr): Move from tree-vectorizer.h
2323 (vect_vect_recog_func_ptrs): Move further down the file.
2324 (vect_recog_func): Likewise. Remove the third argument.
2325 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2326 (vect_pattern_recog_1): Expect the pattern function to do any
2327 necessary target tests. Also expect it to provide a vector type.
2328 Remove the type_in handling.
2330 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2332 * tree-vect-patterns.c (vect_pattern_detected): New function.
2333 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2334 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2335 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2336 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2337 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2338 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2339 (vect_recog_mask_conversion_pattern)
2340 (vect_try_gather_scatter_pattern): Likewise.
2342 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2344 * tree-vect-patterns.c (vect_get_internal_def): New function.
2345 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2346 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2347 (search_type_for_mask_1): Use it.
2349 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2351 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2352 redundant WIDEN_SUM_EXPR handling.
2353 (vect_recog_sad_pattern): Likewise.
2355 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2357 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2358 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2359 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2360 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2361 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2363 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2365 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2366 use the stmt_vec_info of the original bb statement for the
2367 new zero assignment, even if the call is part of a pattern.
2369 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2371 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2372 that the sequence is attached to the original statement rather
2373 than the pattern statement.
2374 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2375 PATTERN_DEF_SEQ from the original statement rather than
2376 the main pattern statement.
2377 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2378 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2379 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2381 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2383 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2384 definition statements before the early exit for statements that aren't
2386 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2388 (vect_transform_loop): ...here. Process pattern definition
2389 statements without first checking whether the main pattern
2390 statement is live or relevant.
2392 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2394 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2395 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2397 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2399 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2400 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2401 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2402 (expand_block_compare): Change select_block_compare_mode call.
2403 (expand_strncmp_align_check): Use new functions, fix comment.
2404 (emit_final_str_compare_gpr): New function.
2405 (expand_strn_compare): Refactor and clean up code.
2406 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2408 2018-06-19 Tony Reix <tony.reix@atos.com>
2409 Damien Bergamini <damien.bergamini@atos.com>
2410 David Edelsohn <dje.gcc@gmail.com>
2412 * collect2.c (static_obj): New variable.
2413 (static_libs): New variable.
2414 (is_in_list): Uncomment declaration.
2415 (main): Track AIX libraries linked statically.
2416 (is_in_list): Uncomment definition.
2417 (scan_prog_file): Don't add AIX shared libraries initializer
2418 to constructor list if linking statically.
2420 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2422 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2424 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2426 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2428 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2431 2018-06-19 Martin Liska <mliska@suse.cz>
2433 * config/i386/i386.c (ix86_can_inline_p): Do not use
2434 ipa_fn_summaries::get_create.
2435 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2437 (devirtualization_time_bonus): Likewise.
2438 (ipcp_propagate_stage): Likewise.
2439 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2440 (edge_set_predicate): Likewise.
2441 (evaluate_conditions_for_known_args): Likewise.
2442 (evaluate_properties_for_edge): Likewise.
2443 (ipa_call_summary::reset): Tranform to ...
2444 (ipa_call_summary::~ipa_call_summary): ... this.
2445 (ipa_fn_summary::reset): Transform to ...
2446 (ipa_fn_summary::~ipa_fn_summary): ... this.
2447 (ipa_fn_summary_t::remove): Rename to ...
2448 (ipa_fn_summary_t::remove_callees): ... this.
2449 (ipa_fn_summary_t::duplicate): Use placement new
2450 instead of memory copy.
2451 (ipa_call_summary_t::duplicate): Likewise.
2452 (ipa_call_summary_t::remove): Remove.
2453 (dump_ipa_call_summary): Change get_create to get.
2454 (ipa_dump_fn_summary): Dump only when summary exists.
2455 (analyze_function_body): Use symbol_summary::get instead
2457 (compute_fn_summary): Likewise.
2458 (estimate_edge_devirt_benefit): Likewise.
2459 (estimate_edge_size_and_time): Likewise.
2460 (inline_update_callee_summaries): Likewise.
2461 (remap_edge_change_prob): Likewise.
2462 (remap_edge_summaries): Likewise.
2463 (ipa_merge_fn_summary_after_inlining): Likewise.
2464 (write_ipa_call_summary): Likewise.
2465 (ipa_fn_summary_write): Likewise.
2466 (ipa_free_fn_summary): Likewise.
2467 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2468 (struct ipa_call_summary): Likewise.
2469 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2471 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2472 (estimate_size_after_inlining): Likewise.
2473 (estimate_growth): Likewise.
2474 (growth_likely_positive): Likewise.
2475 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2476 (inline_call): Likewise.
2477 * ipa-inline.c (caller_growth_limits): Likewise.
2478 (can_inline_edge_p): Likewise.
2479 (can_inline_edge_by_limits_p): Likewise.
2480 (compute_uninlined_call_time): Likewise.
2481 (compute_inlined_call_time): Likewise.
2482 (want_inline_small_function_p): Likewise.
2483 (edge_badness): Likewise.
2484 (update_caller_keys): Likewise.
2485 (update_callee_keys): Likewise.
2486 (inline_small_functions): Likewise.
2487 (inline_to_all_callers_1): Likewise.
2488 (dump_overall_stats): Likewise.
2489 (early_inline_small_functions): Likewise.
2490 (early_inliner): Likewise.
2491 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2492 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2493 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2494 * ipa-split.c (execute_split_functions): Likewise.
2495 * symbol-summary.h: Likewise.
2496 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2498 2018-06-19 Richard Biener <rguenther@suse.de>
2500 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2501 (vectorize_loops): ... here. Fix dbgcnt handling.
2502 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2504 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2507 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2508 ieee128 argument takes up only one (vector) register, not two (floating
2511 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2513 * gimplify.c (gimplify_init_constructor): Really never clear for an
2514 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2516 2018-06-19 Richard Biener <rguenther@suse.de>
2518 PR tree-optimization/86179
2519 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2520 after failed recognition.
2522 2018-06-18 Martin Sebor <msebor@redhat.com>
2525 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2526 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2527 Handle integer subtraction.
2528 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2529 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2531 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2533 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2534 param from rtx to rtx_insn *.
2535 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2537 (frv_ifcvt_modify_insn): Likwise.
2538 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2539 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2540 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2541 as_a <rtx_insn *> cast to local "unprotected_region" once
2542 it's been established that it's not NULL or pc_rtx.
2543 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2544 param "sethi" from rtx to rtx_insn *.
2545 (nds32_group_float_insns): Likewise for param "insn".
2546 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2548 (vax_output_int_subtract): Likewise.
2549 * config/vax/vax.c (vax_output_int_add): Likewise for param
2551 (vax_output_int_subtract): Likewise.
2552 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2553 (emit_pattern_after): Likewise for param "after".
2554 (emit_insn_after): Likewise.
2555 (emit_jump_insn_after): Likewise.
2556 (emit_call_insn_after): Likewise.
2557 (emit_debug_insn_after): Likewise.
2558 (emit_pattern_before): Likewise for param "before".
2559 (emit_insn_before): Likewise.
2560 (emit_jump_insn_before): Likewise.
2561 * final.c (get_insn_template): Likewise for param "insn", removing
2563 * output.h (get_insn_template): Likewise for 2nd param.
2564 * rtl.h (emit_insn_before): Likewise.
2565 (emit_jump_insn_before): Likewise.
2566 (emit_debug_insn_before_noloc): Likewise.
2567 (emit_insn_after): Likewise.
2568 (emit_jump_insn_after): Likewise.
2569 (emit_call_insn_after): Likewise.
2570 (emit_debug_insn_after): Likewise.
2571 (set_insn_deleted): Likewise for param.
2573 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2576 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2577 floating point modes, so that IFmode is numerically greater than
2578 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2579 to declare the ordering. This prevents IFmode from being
2580 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2581 machine. Include rs6000-modes.h to share the fractional values
2582 between genmodes* and the rest of the compiler.
2586 * config/rs6000/rs6000-modes.h: New file.
2587 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2588 meaning of rs6000_long_double_size so that 126..128 selects an
2589 appropriate 128-bit floating point type.
2590 (rs6000_option_override_internal): Likewise.
2591 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2592 (TARGET_LONG_DOUBLE_128): Change the meaning of
2593 rs6000_long_double_size so that 126..128 selects an appropriate
2594 128-bit floating point type.
2595 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2596 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2597 source and destination to match the standard usage.
2598 (truncifkf2): Likewise.
2599 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2600 ISA 2.07 to use an explicit clobber, instead of passing in a
2602 (copysign<mode>3_soft): Likewise.
2604 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2606 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2607 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2608 (vect_slp_analyze_instance_dependence): Likewise.
2609 (vect_enhance_data_refs_alignment): Likewise.
2610 (vect_analyze_data_refs_alignment): Likewise.
2611 (vect_slp_analyze_and_verify_instance_alignment
2612 (vect_analyze_data_ref_accesses): Likewise.
2613 (vect_prune_runtime_alias_test_list): Likewise.
2614 (vect_analyze_data_refs): Likewise.
2615 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2616 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2617 (vect_analyze_scalar_cycles_1): Likewise.
2618 (vect_get_loop_niters): Likewise.
2619 (vect_analyze_loop_form_1): Likewise.
2620 (vect_update_vf_for_slp): Likewise.
2621 (vect_analyze_loop_operations): Likewise.
2622 (vect_analyze_loop): Likewise.
2623 (vectorizable_induction): Likewise.
2624 (vect_transform_loop): Likewise.
2625 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2626 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2627 (vect_make_slp_decision): Likewise.
2628 (vect_detect_hybrid_slp): Likewise.
2629 (vect_slp_analyze_operations): Likewise.
2630 (vect_slp_bb): Likewise.
2631 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2632 (vectorizable_bswap): Likewise.
2633 (vectorizable_call): Likewise.
2634 (vectorizable_simd_clone_call): Likewise.
2635 (vectorizable_conversion): Likewise.
2636 (vectorizable_assignment): Likewise.
2637 (vectorizable_shift): Likewise.
2638 (vectorizable_operation): Likewise.
2639 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2641 2018-06-18 Martin Sebor <msebor@redhat.com>
2643 PR tree-optimization/81384
2644 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2645 * builtins.c (expand_builtin_strnlen): New function.
2646 (expand_builtin): Call it.
2647 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2648 * builtins.def (BUILT_IN_STRNLEN): New.
2649 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2650 Warn for bounds in excess of maximum object size.
2651 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2652 single-value ranges. Handle strnlen.
2653 (handle_builtin_strlen): Handle strnlen.
2654 (strlen_check_and_optimize_stmt): Same.
2655 * doc/extend.texi (Other Builtins): Document strnlen.
2657 2018-06-18 Maya Rashish <coypu@sdf.org>
2659 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2660 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2661 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2663 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2665 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2667 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2669 * tree.c (escaped_string::escape): Replace cast to char * by
2670 const_cast<char *> (unescaped).
2672 2018-06-18 Nick Clifton <nickc@redhat.com>
2675 * tree.c (escaped_string): New class. Converts an unescaped
2676 string into its escaped equivalent.
2677 (warn_deprecated_use): Use the new class to convert the
2678 deprecation message, if present.
2679 (test_escaped_strings): New self test.
2680 (test_c_tests): Add test_escaped_strings.
2681 * doc/extend.texi (deprecated): Add a note that the
2682 deprecation message is affected by the -fmessage-length
2683 option, and that control characters will be escaped.
2684 (#pragma GCC error): Document this pragma.
2685 (#pragma GCC warning): Likewise.
2686 * doc/invoke.texi (-fmessage-length): Document this option's
2687 effect on the #warning and #error preprocessor directives and
2688 the deprecated attribute.
2690 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2692 * tree.c (decl_value_expr_lookup): Revert latest change.
2693 (decl_value_expr_insert): Likewise.
2695 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2697 * gimplify.c (nonlocal_vlas): Delete.
2698 (nonlocal_vla_vars): Likewise.
2699 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2701 (gimplify_body): Do not create and destroy nonlocal_vlas.
2702 * tree-nested.c: Include diagnostic.h.
2703 (use_pointer_in_frame): Tweak.
2704 (lookup_field_for_decl): Add assertion and declare the transformation.
2705 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2706 internal error when the reference is in a wrong context. Do not
2707 create a debug decl by default.
2708 (note_nonlocal_block_vlas): Delete.
2709 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2710 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2711 create a debug decl by default.
2712 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2713 call is in a wrong context.
2714 (fixup_vla_decls): New function.
2715 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2716 debug variables were created.
2717 * tree.c (decl_value_expr_lookup): Add checking assertion.
2718 (decl_value_expr_insert): Likewise.
2720 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2723 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2724 * tree-scalar-evolution.c (interpret_expr): Likewise.
2725 (expression_expensive_p): Likewise.
2726 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2727 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2728 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2729 (ssa_defined_by_minus_one_stmt_p): New.
2731 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2734 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2735 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2736 * dojump.c (do_jump): Likewise.
2737 * expr.c (expand_expr_real_2): Check operand type's sign.
2738 * fold-const.c (const_unop): Handle ABSU_EXPR.
2739 (fold_abs_const): Likewise.
2740 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2741 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2742 (strip_sign_op_1): Likesise.
2743 * match.pd: Add new pattern to generate ABSU_EXPR.
2744 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2745 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2746 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2747 * tree-inline.c (estimate_operator_cost): Likewise.
2748 * tree-pretty-print.c (dump_generic_node): Likewise.
2749 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2750 * tree.def (ABSU_EXPR): New.
2752 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2755 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2756 documented as preserved for backward compatibility only.
2757 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2759 PR rtl-optimization/86108
2760 * bb-reorder.c (create_forwarder_block): Renamed to ...
2761 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2762 jump from new landing pad to the second part.
2763 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2766 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2769 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2770 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2771 Only call store_expr for halves if the mode is the same.
2774 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2775 Fix up comment formatting.
2777 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2779 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2780 (typed_splay_tree::closure,
2781 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2782 (typed_splay_tree::typed_splay_tree,
2783 typed_splay_tree::operator =): Declared private.
2784 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2785 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2786 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2787 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2788 typed_splay_tree::splay_tree_splay,
2789 typed_splay_tree::splay_tree_foreach_helper,
2790 typed_splay_tree::splay_tree_insert,
2791 typed_splay_tree::splay_tree_remove,
2792 typed_splay_tree::splay_tree_lookup,
2793 typed_splay_tree::splay_tree_predecessor,
2794 typed_splay_tree::splay_tree_successor,
2795 typed_splay_tree::splay_tree_min,
2796 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2797 (typed_splay_tree::root, typed_splay_tree::comp,
2798 typed_splay_tree::delete_key,
2799 typed_splay_tree::delete_value): New data members.
2800 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2801 typed_splay_tree::remove.
2803 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2805 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2806 -mginv and -mno-ginv to the assembler.
2807 * config/mips/mips.opt (-mcrc): New option.
2809 * doc/invoke.text (-mcrc): Document.
2812 2018-06-15 Nick Clifton <nickc@redhat.com>
2815 * tree.c (escaped_string): New class. Converts an unescaped
2816 string into its escaped equivalent.
2817 (warn_deprecated_use): Use the new class to convert the
2818 deprecation message, if present.
2819 (test_escaped_strings): New self test.
2820 (test_c_tests): Add test_escaped_strings.
2821 * doc/extend.texi (deprecated): Add a note that the
2822 deprecation message is affected by the -fmessage-length
2823 option, and that control characters will be escaped.
2824 (#pragma GCC error): Document this pragma.
2825 (#pragma GCC warning): Likewise.
2826 * doc/invoke.texi (-fmessage-length): Document this option's
2827 effect on the #warning and #error preprocessor directives and
2828 the deprecated attribute.
2830 2018-06-15 Richard Biener <rguenther@suse.de>
2832 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2833 here, also noting vector size used.
2834 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2835 size used in MSG_OPTIMIZED_LOCATIONS dump.
2836 (pass_slp_vectorize::execute): Adjust.
2838 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2841 * config/arc/arc.c (arc_return_address_register): Fix
2844 2018-06-15 Richard Biener <rguenther@suse.de>
2847 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2848 leave useless conversion stripping to force_gimple_operand_gsi.
2849 (gimplify_build2): Likewise.
2850 (gimplify_build1): Likewise.
2852 2018-06-15 Richard Biener <rguenther@suse.de>
2855 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2856 before adjusting their block.
2858 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2860 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2861 multilibs for *-*-rtems*.
2862 * config/riscv/t-rtems: New file.
2864 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2867 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2868 unsigned_type_for returns NULL.
2871 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2872 subregs of multi-word pseudos unless the float mode has word size.
2874 2018-06-14 Richard Biener <rguenther@suse.de>
2877 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2878 duplicate and harmful type_hash_canon.
2879 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2881 2018-06-14 Richard Biener <rguenther@suse.de>
2884 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2887 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2889 * config/rtems.h (STDINT_LONG32): Define.
2891 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2892 Prachi Godbole <prachi.godbole@imgtec.com>
2894 * config/mips/mips-cpus.def: Define P6600.
2895 * config/mips/mips-tables.opt: Regenerate.
2896 * config/mips/mips.c (mips_ucbranch_type): New enum.
2897 (mips_rtx_cost_data): Add support for P6600.
2898 (mips_issue_rate): Likewise.
2899 (mips_multipass_dfa_lookahead): Likewise.
2900 (mips_avoid_hazard): Likewise.
2901 (mips_reorg_process_insns): Likewise.
2902 (mips_classify_branch_p6600): New function.
2903 * config/mips/mips.h (TUNE_P6600): New define.
2904 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2905 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2906 * config/mips/mips.md: Include p6600.md.
2907 (processor): Add p6600.
2908 * config/mips/p6600.md: New file.
2909 * doc/invoke.texi: Add p6600 to supported architectures.
2911 2018-06-13 Martin Sebor <msebor@redhat.com>
2913 PR tree-optimization/86114
2914 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2916 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2918 2018-06-13 Richard Biener <rguenther@suse.de>
2920 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2921 Properly set vector type of the intermediate stmt.
2922 * tree-vect-stmts.c (vectorizable_operation): The destination
2923 var always has vectype_out type.
2925 2018-06-13 Jeff Law <law@redhat.com>
2927 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2928 integer 0 for argument to print_rtl_with_bb.
2929 (rl78_reorg): Likewise.
2931 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2933 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2934 from rtx to rtx_insn *.
2935 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2937 (add_sched_insns_for_speculation): Likewise for local "target",
2938 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2939 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2940 from rtx_insn ** to rtx_code_label **.
2941 (reorg_emit_nops): Likewise.
2942 (c6x_reorg): Likewise for local "call_labels".
2943 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2945 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2946 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2947 the loops over LABEL_REFS.
2948 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2950 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2951 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2952 (split_branches): Strengthen local "olabel" from rtx to
2953 rtx_insn *, adding a safe_as_a cast.
2954 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2956 (add_insn_after): Likewise for first two params.
2957 (add_insn_before): Likewise.
2958 (remove_insn): Likewise for param.
2959 (emit_pattern_before_noloc): Likewise for second and third params.
2960 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2961 (emit_call_insn_before_noloc): Likewise.
2962 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2964 (emit_barrier_before): Likewise.
2965 (emit_label_before): Strengthen "label" param from "rtx" to
2966 "rtx_code_label *". Strengthen "before" param from "rtx" to
2968 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2970 (emit_pattern_after_noloc): Likewise.
2971 (emit_insn_after_noloc): Likewise.
2972 (emit_jump_insn_after_noloc): Likewise.
2973 (emit_call_insn_after_noloc): Likewise.
2974 (emit_debug_insn_after_noloc): Likewise.
2975 (emit_barrier_after): Likewise.
2976 (emit_label_after): Likewise for both params.
2977 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2978 "loc" param from "int" to "location_t".
2979 (emit_insn_after_setloc): Likewise.
2980 (emit_jump_insn_after_setloc): Likewise.
2981 (emit_call_insn_after_setloc): Likewise.
2982 (emit_debug_insn_after_setloc): Likewise.
2983 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2984 "loc" param from "int" to "location_t".
2985 (emit_pattern_before): Convert NULL_RTX to NULL.
2986 (emit_insn_before_setloc): Convert "loc" param from "int" to
2988 (emit_jump_insn_before_setloc): Likewise.
2989 (emit_call_insn_before_setloc): Likewise.
2990 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2991 rtx_insn *. Convert "loc" param from "int" to "location_t".
2992 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2993 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2994 Convert 3rd param from "int" to "location_t".
2995 (emit_barrier_before, emit_barrier_after, next_real_insn):
2996 Strengthen param from rtx to rtx_insn *.
2997 (emit_label_before): Strengthen 1st param from "rtx" to
2998 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3000 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3001 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3002 Strengthen 2nd param from "rtx" to "rtx_insn *".
3003 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3004 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3005 Likewise. Convert 3rd param from "int" to "location_t".
3006 (emit_label_after): Strengthen 1st param from "rtx" to
3008 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3010 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3011 from "rtx" to "rtx_insn *".
3013 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3015 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3016 bodies streamed in with -Q.
3017 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3018 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3019 * gimple-streamer-out.c: Include gimple-pretty-print.h
3020 (output_bb): Dump stmts streamed.
3021 * lto-section-out.c: Include print-tree.h
3022 (lto_begin_section): Dump sections created.
3023 (lto_output_decl_index): Dump decl encoded.
3024 * lto-streamer-out.c: Include print-tree.h
3025 (create_output_block): Dump output block created.
3026 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3027 (output_function): Dump function output.
3028 (output_constructor): Dump constructor streamed.
3029 (write_global_stream): Output indexes encoded.
3030 (produce_asm_for_decls): Dump streams encoded.
3031 * lto-streamer.c (streamer_dump_file): New global var.
3032 * lto-streamer.h (streamer_dump_file): Declare.
3033 * passes.c (ipa_write_summaries): Initialize streamer dump.
3034 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3037 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3040 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3041 offsets for register save directives. Emit a second batch of save
3042 directives, if need be, when the function accesses prior frames.
3044 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3046 * config/arc/fpu.md (fmasf4): Force operand to register.
3047 (fnmasf4): Likewise.
3049 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3051 * config/arc/arc-protos.h (arc_pad_return): Remove.
3052 * config/arc/arc.c (machine_function): Remove force_short_suffix
3054 (arc_print_operand): Adjust printing of '&'.
3055 (arc_verify_short): Remove conditional printing of short suffix.
3056 (arc_final_prescan_insn): Remove reference to size_reason.
3057 (pad_return): New function.
3058 (arc_reorg): Call pad_return.
3059 (arc_pad_return): Remove.
3060 (arc_init_machine_status): Remove reference to force_short_suffix.
3061 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3062 (attr length): When attribute iscompact is true force to 2
3063 regardless; in the case of maybe check if we want to force the
3064 instruction to have 4 bytes length.
3065 (nopv): Change it to generate 4 byte long nop as well.
3066 (blockage): New pattern.
3067 (simple_return): Remove call to arc_pad_return.
3068 (p_return_i): Likewise.
3070 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3072 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3074 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3076 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3079 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3081 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3082 for ARC700 and ARCv2.
3084 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3087 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3088 operands[2] instead of operands[1].
3091 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3093 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3094 case, check whether the outer register overlaps an unallocatable
3095 register, not just whether it fits the required class.
3097 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3099 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3101 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3102 (memrefs_conflict_p): Likewise.
3103 (init_alias_analysis): Likewise.
3104 * cfgexpand.c (expand_debug_expr): Likewise.
3105 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3106 * cse.c (fold_rtx): Likewise.
3107 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3108 * expr.c (emit_block_move_hints): Likewise.
3109 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3110 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3111 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3112 (emit_group_store): Likewise.
3113 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3114 to read the PRE/POST_MODIFY increment.
3115 * calls.c (store_one_arg): Use strip_offset.
3116 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3118 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3120 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3121 (simplify_binary_operation_1): Extend CONST_INT handling to
3123 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3124 than a HOST_WIDE_INT.
3125 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3127 (adjust_mems, add_stores): Update accodingly.
3128 (vt_canonicalize_addr): Track polynomial offsets.
3129 (emit_note_insn_var_location): Likewise.
3130 (vt_add_function_parameter): Likewise.
3131 (vt_initialize): Likewise.
3133 2018-06-12 Jeff Law <law@redhat.com>
3135 * config.gcc (alpha*-*-freebsd*): Remove.
3136 * config/alpha/freebsd.h: Remove.
3138 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3141 * spellcheck-tree.c (levenshtein_distance): Rename to...
3142 (get_edit_distance): ...this, and update for underlying renaming.
3143 * spellcheck-tree.h (levenshtein_distance): Rename to...
3144 (get_edit_distance): ...this.
3145 * spellcheck.c (levenshtein_distance): Rename to...
3146 (get_edit_distance): ...this. Convert from Levenshtein distance
3147 to Damerau-Levenshtein distance by supporting transpositions of
3148 adjacent characters. Rename "v1" to "v_next" and "v0" to
3150 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3151 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3152 update for underlying renaming.
3153 (selftest::levenshtein_distance_unit_test): Rename to...
3154 (selftest::test_get_edit_distance_unit): ...this, and update for
3155 underlying renaming.
3156 (selftest::test_find_closest_string): Add example from PR 69968
3157 where transposition helps
3158 (selftest::test_metric_conditions): Update for renaming.
3159 (selftest::test_metric_conditions): Likewise.
3160 (selftest::spellcheck_c_tests): Likewise.
3161 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3162 (get_edit_distance): ...this.
3163 (best_match::consider): Update for renaming.
3165 2018-06-12 Martin Sebor <msebor@redhat.com>
3167 PR tree-optimization/85259
3168 * builtins.c (compute_objsize): Handle constant offsets.
3169 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3170 true iff a warning has been issued.
3171 * gimple.h (gimple_nonartificial_location): New function.
3172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3173 gimple_nonartificial_location and handle -Wno-system-headers.
3174 (handle_builtin_stxncpy): Same.
3176 2018-06-12 Martin Sebor <msebor@redhat.com>
3179 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3181 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3183 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3184 BUILTIN_VEC_XST entries for pointer to double and long long.
3186 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3189 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3191 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3194 2018-06-12 Martin Liska <mliska@suse.cz>
3196 * doc/options.texi: Document IntegerRange.
3198 2018-06-12 Martin Liska <mliska@suse.cz>
3200 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3201 * opt-functions.awk: Handle Deprecated flag.
3202 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3204 (read_cmdline_option): Report warning for a deprecated option.
3205 * opts.h (struct cl_option): Add new field cl_deprecated.
3206 (CL_ERR_DEPRECATED): New.
3208 2018-06-12 Martin Liska <mliska@suse.cz>
3210 * doc/options.texi: Document Deprecated option flag.
3212 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3214 * config/arc/arc-arch.h (arc_extras): New enum.
3215 (arc_cpu_t):Add field extra.
3216 (arc_cpu_types): Consider the extras.
3217 * config/arc/arc-cpus.def: Add extras info.
3218 * config/arc/arc-opts.h (processor_type): Consider extra field.
3219 * config/arc/arc.c (arc_override_options): Handle extra field.
3221 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3223 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3224 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3226 * config/arc/arc.c (arc_init): Update pic warning.
3227 (irq_range): Update irq range parsing warnings.
3228 (arc_override_options): Update various warning messages.
3229 (arc_handle_aux_attribute): Likewise.
3231 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3233 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3235 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3237 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3238 compiler messages directives.
3240 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3242 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3243 * config/mips/mips-tables.opt: Regenerate.
3244 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3246 * doc/invoke.texi: Document -march=i6500.
3248 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3250 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3251 (i6400_gpmul): Add cpu_unit.
3252 (i6400_gpdiv): Likewise.
3253 (i6400_msa_add_d): Update reservations.
3254 (i6400_msa_int_add) Likewise.
3255 (i6400_msa_short_logic3) Likewise.
3256 (i6400_msa_short_logic2) Likewise.
3257 (i6400_msa_short_logic) Likewise.
3258 (i6400_msa_move) Likewise.
3259 (i6400_msa_cmp) Likewise.
3260 (i6400_msa_short_float2) Likewise.
3261 (i6400_msa_div_d) Likewise.
3262 (i6400_msa_long_logic1) Likewise.
3263 (i6400_msa_long_logic2) Likewise.
3264 (i6400_msa_mult) Likewise.
3265 (i6400_msa_long_float2) Likewise.
3266 (i6400_msa_long_float4) Likewise.
3267 (i6400_msa_long_float5) Likewise.
3268 (i6400_msa_long_float8) Likewise.
3269 (i6400_fpu_fadd): Include frint type.
3270 (i6400_fpu_store): New define_insn_reservation.
3271 (i6400_fpu_load): Likewise.
3272 (i6400_fpu_move): Likewise.
3273 (i6400_fpu_fcmp): Likewise.
3274 (i6400_fpu_fmadd): Likewise.
3275 (i6400_int_mult): Include imul3nc type and update reservation.
3276 (i6400_int_div): Include idiv3 type and update reservation.
3277 (i6400_int_load): Update to check type not move_type.
3278 (i6400_int_store): Likewise.
3279 (i6400_int_prefetch): Set zero latency.
3281 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3283 * gcc.c: Document new %@{...} sequence.
3284 (LINK_COMMAND_SPEC): Use it for the -L switches.
3285 (cpp_unique_options): Use it for the -I switches.
3286 (at_file_argbuf): New global variable.
3287 (in_at_file): Likewise.
3288 (alloc_args): Create at_file_argbuf.
3289 (clear_args): Truncate at_file_argbuf.
3290 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3291 (open_at_file): New function.
3292 (close_at_file): Likewise.
3293 (create_at_file): Delete.
3294 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3297 (validate_switches_from_spec): Deal with %@{...} sequence.
3298 (validate_switches): Likewise.
3299 (driver::finalize): Call clear_args.
3301 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3303 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3305 2018-06-11 Martin Sebor <msebor@redhat.com>
3307 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3308 (Wno-class-memaccess): Add @opindex.
3309 (Wno-templates, Wno-multiple-inheritance): Same.
3310 (Wno-virtual-inheritance, Wno-namespaces): Same.
3311 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3312 (Wno-format-overflow, Wno-format-truncation): Same.
3313 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3314 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3315 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3316 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3317 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3318 (Wno-misspelled-isr): Same.
3320 2018-06-11 Martin Sebor <msebor@redhat.com>
3322 * PR tree-optimization/86083
3323 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3325 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3327 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3329 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3332 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3333 on the correct operand.
3334 (*movdi_internal64): Ditto.
3336 2018-06-11 Martin Liska <mliska@suse.cz>
3338 PR tree-optimization/86089
3339 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3341 2018-06-11 Julia Koval <julia.koval@intel.com>
3343 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3344 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3345 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3347 2018-06-11 Olivier Hainque <hainque@adacore.com>
3349 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3350 for Ada with strict dwarf2.
3352 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3355 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3358 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3360 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3362 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3364 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3367 2018-06-08 Martin Liska <mliska@suse.cz>
3369 * tree-cfg.h (debug_function): Fix argument type to match
3372 2018-06-08 Martin Liska <mliska@suse.cz>
3374 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3375 Remove usage of MPX-related (and removed) fields.
3376 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3378 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3380 * cfg.c (debug): Use TDF_NONE rather than 0.
3381 * cfghooks.c (debug): Likewise.
3382 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3383 (struct dump_option_value_info): Convert to...
3384 (struct kv_pair): ...this template type.
3385 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3387 (optinfo_verbosity_options): Likewise.
3388 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3390 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3391 than int for "optgroup_flags" param.
3392 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3394 (dump_dec): Likewise.
3395 (dump_finish): Use TDF_NONE rather than 0.
3396 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3397 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3398 than 0. Update for change to option_ptr.
3399 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3400 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3401 0. Update for changes to optinfo_verbosity_options and
3403 (opt_info_switch_p): Convert optgroup_flags from int to
3405 (dump_basic_block): Use dump_flags_t rather than int
3406 for "dump_kind" param.
3407 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3408 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3409 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3410 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3411 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3412 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3413 TDF_NONE): Convert from macros to...
3414 (enum dump_flag): ...this new enum.
3415 (dump_flags_t): Update to use enum.
3416 (operator|, operator&, operator~, operator|=, operator&=):
3417 Implement for dump_flags_t.
3418 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3419 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3420 Convert from macros to...
3421 (enum optgroup_flag): ...this new enum.
3422 (optgroup_flags_t): New typedef.
3423 (operator|, operator|=): Implement for optgroup_flags_t.
3424 (struct dump_file_info): Convert field "alt_flags" to
3425 dump_flags_t. Convert field "optgroup_flags" to
3427 (dump_basic_block): Use dump_flags_t rather than int for param.
3428 (dump_generic_expr_loc): Likewise.
3429 (dump_dec): Likewise.
3430 (dump_register): Convert param "optgroup_flags" to
3432 (opt_info_enable_passes): Likewise.
3433 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3435 * gimple-pretty-print.c (debug): Likewise.
3436 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3437 (merged_store_group::apply_stores): Likewise.
3438 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3439 * gimple.c (verify_gimple_pp): Likewise.
3440 * graphite-poly.c (print_pbb_body): Likewise.
3441 * passes.c (pass_manager::register_one_dump_file): Convert
3442 local "optgroup_flags" to optgroup_flags_t.
3443 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3445 (debug_body): Likewise.
3446 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3447 to optgroup_flags_t.
3448 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3450 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3451 (convert_mult_to_fma): Likewise.
3452 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3453 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3454 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3455 "dump_kind" to dump_flags_t.
3457 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3459 * config/rs6000/rs6000.c (min, max): Delete.
3461 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3463 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3464 -mabi=spe and -mabi=no-spe.
3466 2018-06-08 Martin Liska <mliska@suse.cz>
3468 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3469 where we expect an existing summary.
3471 2018-06-08 Martin Liska <mliska@suse.cz>
3473 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3474 * ipa-inline.h (estimate_edge_growth): Likewise.
3476 2018-06-08 Martin Liska <mliska@suse.cz>
3478 * cgraph.c (function_version_hasher::hash): Use
3479 cgraph_node::get_uid ().
3480 (function_version_hasher::equal):
3481 * cgraph.h (cgraph_node::get_uid): New method.
3482 * ipa-inline.c (update_caller_keys): Use
3483 cgraph_node::get_uid ().
3484 (update_callee_keys): Likewise.
3485 * ipa-utils.c (searchc): Likewise.
3486 (ipa_reduced_postorder): Likewise.
3487 * lto-cgraph.c (input_node): Likewise.
3488 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3489 * symbol-summary.h (symtab_insertion): Likewise.
3490 (symtab_removal): Likewise.
3491 (symtab_duplication): Likewise.
3492 * tree-pretty-print.c (dump_function_header): Likewise.
3493 * tree-sra.c (convert_callers_for_node): Likewise.
3495 2018-06-08 Martin Liska <mliska@suse.cz>
3497 * cgraph.c (symbol_table::create_edge): Always assign a new
3499 (symbol_table::free_edge): Do not recycle numbers.
3500 * cgraph.h (cgraph_edge::get): New method.
3501 * symbol-summary.h (symtab_removal): Use it.
3502 (symtab_duplication): Likewise.
3503 (call_summary::hashable_uid): Remove.
3505 2018-06-08 Martin Liska <mliska@suse.cz>
3507 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3508 (initialize_growth_caches): Remove.
3509 (free_growth_caches): Likewise.
3510 (do_estimate_edge_time): Use edge_growth_cache.
3511 (do_estimate_edge_size): Likewise.
3512 (do_estimate_edge_hints): Likewise.
3513 * ipa-inline.c (reset_edge_caches): Likewise.
3514 (recursive_inlining): Likewise.
3515 (inline_small_functions): Likewise.
3516 * ipa-inline.h (initialize_growth_caches): Remove.
3517 (estimate_edge_size): Likewise.
3518 (estimate_edge_time): Likewise.
3519 (estimate_edge_hints): Likewise.
3520 (reset_edge_growth_cache): Likewise.
3521 * symbol-summary.h (call_summary::remove): New method.
3523 2018-06-08 Martin Liska <mliska@suse.cz>
3525 * ipa-cp.c (class edge_clone_summary): New summary.
3526 (grow_edge_clone_vectors): Remove.
3527 (ipcp_edge_duplication_hook): Remove.
3528 (class edge_clone_summary_t): New call_summary class.
3529 (ipcp_edge_removal_hook): Remove.
3530 (edge_clone_summary_t::duplicate): New function.
3531 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3532 (create_specialized_node): Likewise.
3533 (ipcp_driver): Initialize edge_clone_summaries and do not
3536 2018-06-08 Martin Liska <mliska@suse.cz>
3538 * symbol-summary.h (get): New function.
3539 (call_summary::m_initialize_when_cloning): New class member.
3541 2018-06-08 Martin Liska <mliska@suse.cz>
3543 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3544 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3545 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3546 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3547 (remove_cgraph_node_from_order): Use the removed_nodes set.
3548 (do_per_function_toporder): Likwise.
3549 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3550 instead of summary_uid.
3551 (symtab_removal): Likewise.
3552 (symtab_duplication): Likewise.
3554 2018-06-08 Martin Liska <mliska@suse.cz>
3556 * ipa-cp.c (ipcp_store_bits_results): Use
3557 ipcp_transformation_sum.
3558 (ipcp_store_vr_results): Likewise.
3559 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3561 (ipcp_transformation_initialize): ... this.
3562 (ipa_set_node_agg_value_chain):
3563 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3564 (write_ipcp_transformation_info): Likewise.
3565 (read_ipcp_transformation_info): Likewise.
3566 (ipcp_update_bits): Likewise.
3567 (ipcp_update_vr): Likewise.
3568 (ipcp_transform_function): Likewise.
3569 * ipa-prop.h: Rename ipcp_transformation_summary to
3570 ipcp_transformation.
3571 (class ipcp_transformation_t): New function summary.
3572 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3573 (ipa_get_agg_replacements_for_node): Likewise.
3575 2018-06-08 Martin Liska <mliska@suse.cz>
3577 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3579 (class funct_state_summary_t): New function_summary class.
3580 (has_function_state): Remove.
3581 (get_function_state): Likewise.
3582 (set_function_state): Likewise.
3583 (add_new_function): Likewise.
3584 (funct_state_summary_t::insert): New function.
3585 (duplicate_node_data): Remove.
3586 (remove_node_data): Remove.
3587 (funct_state_summary_t::duplicate): New function.
3588 (register_hooks): Create new funct_state_summaries.
3589 (pure_const_generate_summary): Use it.
3590 (pure_const_write_summary): Likewise.
3591 (pure_const_read_summary): Likewise.
3592 (propagate_pure_const): Likewise.
3593 (propagate_nothrow): Likewise.
3594 (dump_malloc_lattice): Likewise.
3595 (propagate_malloc): Likewise.
3596 (execute): Do not register hooks, just remove summary
3598 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3601 2018-06-08 Martin Liska <mliska@suse.cz>
3603 * ipa-reference.c (remove_node_data): Remove.
3604 (duplicate_node_data): Likewise.
3605 (class ipa_ref_var_info_summary_t): New class.
3606 (class ipa_ref_opt_summary_t): Likewise.
3607 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3608 (get_reference_optimization_summary): Use
3609 ipa_ref_opt_sum_summaries.
3610 (set_reference_vars_info): Remove.
3611 (set_reference_optimization_summary): Likewise.
3612 (ipa_init): Create summaries.
3613 (init_function_info): Use function summary.
3614 (ipa_ref_opt_summary_t::duplicate): New function.
3615 (ipa_ref_opt_summary_t::remove): New function.
3616 (get_read_write_all_from_node): Fix GNU coding style.
3617 (propagate): Use function summary.
3618 (write_node_summary_p): Fix GNU coding style.
3619 (stream_out_bitmap): Likewise.
3620 (ipa_reference_read_optimization_summary): Use function summary.
3621 (ipa_reference_c_finalize): Do not release hooks.
3623 2018-06-08 Martin Liska <mliska@suse.cz>
3625 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3626 (analyze_function_body): Extract multiple calls of get_create.
3627 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3628 * ipa-inline.c (recursive_inlining): Use ::get method.
3629 * ipa-inline.h (estimate_edge_growth): Likewise.
3631 2018-06-08 Martin Liska <mliska@suse.cz>
3633 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3635 (hsa_function_summary::hsa_function_summary): Use the new enum
3637 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3638 * hsa-gen.c (hsa_get_host_function): Likewise.
3639 (get_brig_function_name): Likewise.
3640 * ipa-hsa.c (process_hsa_functions): Likewise.
3641 (ipa_hsa_write_summary): Likewise.
3642 * symbol-summary.h (symtab_duplication): Use ::get function/
3643 (get): New function.
3645 2018-06-08 Martin Liska <mliska@suse.cz>
3647 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3649 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3650 (hsa_register_kernel): Likewise.
3651 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3652 * hsa-gen.c (hsa_get_host_function): Likewise.
3653 (get_brig_function_name): Likewise.
3654 (generate_hsa): Likewise.
3655 (pass_gen_hsail::execute): Likewise.
3656 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3657 (devirtualization_time_bonus): Likewise.
3658 (ipcp_propagate_stage): Likewise.
3659 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3660 (edge_set_predicate): Likewise.
3661 (evaluate_conditions_for_known_args): Likewise.
3662 (evaluate_properties_for_edge): Likewise.
3663 (ipa_fn_summary::reset): Likewise.
3664 (ipa_fn_summary_t::duplicate): Likewise.
3665 (dump_ipa_call_summary): Likewise.
3666 (ipa_dump_fn_summary): Likewise.
3667 (analyze_function_body): Likewise.
3668 (compute_fn_summary): Likewise.
3669 (estimate_edge_devirt_benefit): Likewise.
3670 (estimate_edge_size_and_time): Likewise.
3671 (estimate_calls_size_and_time): Likewise.
3672 (estimate_node_size_and_time): Likewise.
3673 (inline_update_callee_summaries): Likewise.
3674 (remap_edge_change_prob): Likewise.
3675 (remap_edge_summaries): Likewise.
3676 (ipa_merge_fn_summary_after_inlining): Likewise.
3677 (ipa_update_overall_fn_summary): Likewise.
3678 (read_ipa_call_summary): Likewise.
3679 (inline_read_section): Likewise.
3680 (write_ipa_call_summary): Likewise.
3681 (ipa_fn_summary_write): Likewise.
3682 (ipa_free_fn_summary): Likewise.
3683 * ipa-hsa.c (process_hsa_functions): Likewise.
3684 (ipa_hsa_write_summary): Likewise.
3685 (ipa_hsa_read_section): Likewise.
3686 * ipa-icf.c (sem_function::merge): Likewise.
3687 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3688 (do_estimate_edge_time): Likewise.
3689 (estimate_size_after_inlining): Likewise.
3690 (estimate_growth): Likewise.
3691 (growth_likely_positive): Likewise.
3692 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3693 (inline_call): Likewise.
3694 * ipa-inline.c (caller_growth_limits): Likewise.
3695 (can_inline_edge_p): Likewise.
3696 (can_inline_edge_by_limits_p): Likewise.
3697 (compute_uninlined_call_time): Likewise.
3698 (compute_inlined_call_time): Likewise.
3699 (want_inline_small_function_p): Likewise.
3700 (edge_badness): Likewise.
3701 (update_caller_keys): Likewise.
3702 (update_callee_keys): Likewise.
3703 (recursive_inlining): Likewise.
3704 (inline_small_functions): Likewise.
3705 (inline_to_all_callers_1): Likewise.
3706 (dump_overall_stats): Likewise.
3707 (early_inline_small_functions): Likewise.
3708 (early_inliner): Likewise.
3709 * ipa-inline.h (estimate_edge_growth): Likewise.
3710 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3711 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3712 * ipa-prop.h (IPA_NODE_REF): Likewise.
3713 (IPA_EDGE_REF): Likewise.
3714 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3715 (propagate_malloc): Likewise.
3716 * ipa-split.c (execute_split_functions): Likewise.
3717 * symbol-summary.h: Rename get to get_create.
3719 (get_create): Likewise.
3720 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3722 2018-06-08 Martin Liska <mliska@suse.cz>
3724 * symbol-summary.h (release): Move definition out of class
3726 (symtab_removal): Likewise.
3727 (symtab_duplication): Likewise.
3729 2018-06-08 Martin Liska <mliska@suse.cz>
3731 * symbol-summary.h (function_summary): Move constructor
3732 implementation out of class declaration.
3733 (release): Likewise.
3734 (symtab_insertion): Likewise.
3735 (symtab_removal): Likewise.
3736 (symtab_duplication): Likewise.
3739 2018-06-08 Martin Liska <mliska@suse.cz>
3741 * Makefile.in: Remove support for MPX (macros, related functions,
3742 fields in cgraph_node, ...).
3743 * builtin-types.def (BT_BND): Likewise.
3744 (BT_FN_BND_CONST_PTR): Likewise.
3745 (BT_FN_CONST_PTR_BND): Likewise.
3746 (BT_FN_VOID_PTR_BND): Likewise.
3747 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3748 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3749 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3750 (expand_builtin_mempcpy_with_bounds): Likewise.
3751 (expand_builtin_memset_with_bounds): Likewise.
3752 (expand_builtin_memset_args): Likewise.
3753 (std_expand_builtin_va_start): Likewise.
3754 (expand_builtin): Likewise.
3755 (expand_builtin_with_bounds): Likewise.
3756 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3757 (DEF_LIB_BUILTIN_CHKP): Likewise.
3758 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3759 (DEF_CHKP_BUILTIN): Likewise.
3760 (BUILT_IN_MEMCPY): Likewise.
3761 (BUILT_IN_MEMMOVE): Likewise.
3762 (BUILT_IN_MEMPCPY): Likewise.
3763 (BUILT_IN_MEMSET): Likewise.
3764 (BUILT_IN_STPCPY): Likewise.
3765 (BUILT_IN_STRCAT): Likewise.
3766 (BUILT_IN_STRCHR): Likewise.
3767 (BUILT_IN_STRCPY): Likewise.
3768 (BUILT_IN_STRLEN): Likewise.
3769 (BUILT_IN_MEMCPY_CHK): Likewise.
3770 (BUILT_IN_MEMMOVE_CHK): Likewise.
3771 (BUILT_IN_MEMPCPY_CHK): Likewise.
3772 (BUILT_IN_MEMSET_CHK): Likewise.
3773 (BUILT_IN_STPCPY_CHK): Likewise.
3774 (BUILT_IN_STRCAT_CHK): Likewise.
3775 (BUILT_IN_STRCPY_CHK): Likewise.
3776 * calls.c (store_bounds): Likewise.
3777 (emit_call_1): Likewise.
3778 (special_function_p): Likewise.
3779 (maybe_warn_nonstring_arg): Likewise.
3780 (initialize_argument_information): Likewise.
3781 (finalize_must_preallocate): Likewise.
3782 (compute_argument_addresses): Likewise.
3783 (expand_call): Likewise.
3784 * cfgexpand.c (expand_call_stmt): Likewise.
3785 (expand_return): Likewise.
3786 (expand_gimple_stmt_1): Likewise.
3787 (pass_expand::execute): Likewise.
3788 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3789 (cgraph_node::remove): Likewise.
3790 (cgraph_node::dump): Likewise.
3791 (cgraph_node::verify_node): Likewise.
3792 * cgraph.h (chkp_function_instrumented_p): Likewise.
3793 (symtab_node::get_alias_target): Likewise.
3794 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3795 (cgraph_local_p): Likewise.
3796 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3797 (cgraph_edge::rebuild_references): Likewise.
3798 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3799 (walk_polymorphic_call_targets): Likewise.
3800 (cgraph_node::expand_thunk): Likewise.
3801 (symbol_table::output_weakrefs): Likewise.
3802 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3803 (ix86_handle_option): Likewise.
3804 * config/i386/constraints.md: Likewise.
3805 * config/i386/i386-builtin-types.def (BND): Likewise.
3809 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3810 (BDESC_FIRST): Likewise.
3812 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3813 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3814 * config/i386/i386.c (enum reg_class): Likewise.
3815 (ix86_target_string): Likewise.
3816 (ix86_option_override_internal): Likewise.
3817 (ix86_conditional_register_usage): Likewise.
3818 (ix86_valid_target_attribute_inner_p): Likewise.
3819 (ix86_set_indirect_branch_type): Likewise.
3820 (ix86_set_current_function): Likewise.
3821 (ix86_function_arg_regno_p): Likewise.
3822 (init_cumulative_args): Likewise.
3823 (ix86_function_arg_advance): Likewise.
3824 (ix86_function_arg): Likewise.
3825 (ix86_pass_by_reference): Likewise.
3826 (ix86_function_value_regno_p): Likewise.
3827 (ix86_function_value_1): Likewise.
3828 (ix86_function_value_bounds): Likewise.
3829 (ix86_return_in_memory): Likewise.
3830 (ix86_setup_incoming_vararg_bounds): Likewise.
3831 (ix86_va_start): Likewise.
3832 (indirect_thunk_need_prefix): Likewise.
3833 (print_reg): Likewise.
3834 (ix86_print_operand): Likewise.
3835 (ix86_expand_call): Likewise.
3836 (ix86_output_function_return): Likewise.
3837 (reg_encoded_number): Likewise.
3838 (BDESC_VERIFYS): Likewise.
3839 (ix86_init_mpx_builtins): Likewise.
3840 (ix86_init_builtins): Likewise.
3841 (ix86_emit_cmove): Likewise.
3842 (ix86_emit_move_max): Likewise.
3843 (ix86_expand_builtin): Likewise.
3844 (ix86_builtin_mpx_function): Likewise.
3845 (ix86_get_arg_address_for_bt): Likewise.
3846 (ix86_load_bounds): Likewise.
3847 (ix86_store_bounds): Likewise.
3848 (ix86_load_returned_bounds): Likewise.
3849 (ix86_store_returned_bounds): Likewise.
3850 (ix86_class_likely_spilled_p): Likewise.
3851 (ix86_hard_regno_mode_ok): Likewise.
3852 (x86_order_regs_for_local_alloc): Likewise.
3853 (ix86_mitigate_rop): Likewise.
3854 (ix86_bnd_prefixed_insn_p): Likewise.
3855 (ix86_mpx_bound_mode): Likewise.
3856 (ix86_make_bounds_constant): Likewise.
3857 (ix86_initialize_bounds): Likewise.
3858 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3859 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3860 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3861 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3862 (TARGET_CHKP_BOUND_MODE): Likewise.
3863 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3864 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3865 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3866 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3867 * config/i386/i386.h (TARGET_MPX): Likewise.
3868 (TARGET_MPX_P): Likewise.
3869 (VALID_BND_REG_MODE): Likewise.
3870 (FIRST_BND_REG): Likewise.
3871 (LAST_BND_REG): Likewise.
3872 (enum reg_class): Likewise.
3873 (BND_REG_P): Likewise.
3874 (BND_REGNO_P): Likewise.
3875 (BNDmode): Likewise.
3876 (ADJUST_INSN_LENGTH): Likewise.
3877 * config/i386/i386.md: Likewise.
3878 * config/i386/i386.opt: Likewise.
3879 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3880 (defined): Likewise.
3881 (LINK_MPX): Likewise.
3882 (MPX_SPEC): Likewise.
3883 (LIBMPX_SPEC): Likewise.
3884 (LIBMPXWRAPPERS_SPEC): Likewise.
3885 (CHKP_SPEC): Likewise.
3886 * config/i386/predicates.md: Likewise.
3887 * dbxout.c (dbxout_type): Likewise.
3888 * doc/extend.texi: Likewise.
3889 * doc/invoke.texi: Likewise.
3890 * doc/md.texi: Likewise.
3891 * doc/tm.texi: Likewise.
3892 * doc/tm.texi.in: Likewise.
3893 * dwarf2out.c (is_base_type): Likewise.
3894 (gen_formal_types_die): Likewise.
3895 (gen_subprogram_die): Likewise.
3896 (gen_type_die_with_usage): Likewise.
3897 (gen_decl_die): Likewise.
3898 (dwarf2out_late_global_decl): Likewise.
3899 * expr.c (expand_assignment): Likewise.
3900 (emit_storent_insn): Likewise.
3901 (store_expr_with_bounds): Likewise.
3902 (store_expr): Likewise.
3903 (expand_expr_real_1): Likewise.
3904 * expr.h (store_expr_with_bounds): Likewise.
3905 * function.c (use_register_for_decl): Likewise.
3906 (struct bounds_parm_data): Likewise.
3907 (assign_parms_augmented_arg_list): Likewise.
3908 (assign_parm_find_entry_rtl): Likewise.
3909 (assign_parm_is_stack_parm): Likewise.
3910 (assign_parm_load_bounds): Likewise.
3911 (assign_bounds): Likewise.
3912 (assign_parms): Likewise.
3913 (expand_function_start): Likewise.
3914 * gcc.c (CHKP_SPEC): Likewise.
3915 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3916 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3917 (wrestrict_dom_walker::check_call): Likewise.
3918 * gimple.c (gimple_build_call_from_tree): Likewise.
3919 * gimple.h (enum gf_mask): Likewise.
3920 (gimple_call_with_bounds_p): Likewise.
3921 (gimple_call_set_with_bounds): Likewise.
3922 * gimplify.c (gimplify_init_constructor): Likewise.
3923 * ipa-cp.c (initialize_node_lattices): Likewise.
3924 (propagate_constants_across_call): Likewise.
3925 (find_more_scalar_values_for_callers_subset): Likewise.
3926 * ipa-hsa.c (process_hsa_functions): Likewise.
3927 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3928 * ipa-icf.c (sem_function::merge): Likewise.
3929 * ipa-inline.c (early_inliner): Likewise.
3930 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3931 (warn_function_cold): Likewise.
3932 (propagate_pure_const): Likewise.
3933 * ipa-ref.h (enum GTY): Likewise.
3934 * ipa-split.c (find_retbnd): Likewise.
3935 (consider_split): Likewise.
3936 (split_function): Likewise.
3937 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3938 * ipa.c (walk_polymorphic_call_targets): Likewise.
3939 (symbol_table::remove_unreachable_nodes): Likewise.
3940 (process_references): Likewise.
3941 (cgraph_build_static_cdtor_1): Likewise.
3942 * lto-cgraph.c (lto_output_node): Likewise.
3943 (output_refs): Likewise.
3944 (compute_ltrans_boundary): Likewise.
3945 (input_overwrite_node): Likewise.
3946 (input_node): Likewise.
3947 (input_cgraph_1): Likewise.
3948 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3949 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3950 (class pass_chkp_instrumentation_passes): Likewise.
3951 (make_pass_chkp_instrumentation_passes): Likewise.
3952 * passes.def: Likewise.
3953 * rtl.h (struct GTY): Likewise.
3954 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3955 * stor-layout.c (layout_type): Likewise.
3956 * symtab.c: Likewise.
3957 * target.def: Likewise.
3958 * targhooks.c (default_chkp_bound_type): Likewise.
3959 (default_chkp_bound_mode): Likewise.
3960 (default_builtin_chkp_function): Likewise.
3961 (default_chkp_function_value_bounds): Likewise.
3962 (default_chkp_make_bounds_constant): Likewise.
3963 (default_chkp_initialize_bounds): Likewise.
3964 * targhooks.h (default_chkp_bound_type): Likewise.
3965 (default_chkp_bound_mode): Likewise.
3966 (default_builtin_chkp_function): Likewise.
3967 (default_chkp_function_value_bounds): Likewise.
3968 (default_chkp_make_bounds_constant): Likewise.
3969 (default_chkp_initialize_bounds): Likewise.
3970 * toplev.c (compile_file): Likewise.
3971 (process_options): Likewise.
3972 * tree-core.h (DEF_BUILTIN): Likewise.
3973 (DEF_BUILTIN_CHKP): Likewise.
3974 * tree-inline.c (declare_return_variable): Likewise.
3975 (remap_gimple_stmt): Likewise.
3976 (copy_bb): Likewise.
3977 (initialize_inlined_parameters): Likewise.
3978 (expand_call_inline): Likewise.
3979 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3980 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3981 (make_pass_ipa_chkp_produce_thunks): Likewise.
3982 (make_pass_chkp): Likewise.
3983 (make_pass_chkp_opt): Likewise.
3984 (make_pass_chkp_instrumentation_passes): Likewise.
3985 * tree-pretty-print.c (dump_generic_node): Likewise.
3986 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3987 * tree-ssa-dce.c (propagate_necessity): Likewise.
3988 (eliminate_unnecessary_stmts): Likewise.
3989 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3990 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3991 * tree-ssa-sccvn.h: Likewise.
3992 * tree-ssa-strlen.c (get_string_length): Likewise.
3993 (valid_builtin_call): Likewise.
3994 (adjust_last_stmt): Likewise.
3995 (handle_builtin_strchr): Likewise.
3996 (handle_builtin_strcpy): Likewise.
3997 (handle_builtin_stxncpy): Likewise.
3998 (handle_builtin_memcpy): Likewise.
3999 (handle_builtin_strcat): Likewise.
4000 (strlen_check_and_optimize_stmt): Likewise.
4001 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4002 * tree-streamer-in.c: Likewise.
4003 * tree-streamer.c (record_common_node): Likewise.
4004 * tree.c (tree_code_size): Likewise.
4005 (wide_int_to_tree_1): Likewise.
4006 (type_contains_placeholder_1): Likewise.
4007 (build_common_tree_nodes): Likewise.
4008 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4009 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4010 (POINTER_BOUNDS_P): Likewise.
4011 (BOUNDED_TYPE_P): Likewise.
4012 (BOUNDED_P): Likewise.
4013 (CALL_WITH_BOUNDS_P): Likewise.
4014 (pointer_bounds_type_node): Likewise.
4015 * value-prof.c (gimple_ic): Likewise.
4016 * var-tracking.c (vt_add_function_parameters): Likewise.
4017 * varasm.c (make_decl_rtl): Likewise.
4018 (assemble_start_function): Likewise.
4019 (output_constant): Likewise.
4020 (maybe_assemble_visibility): Likewise.
4021 * varpool.c (ctor_for_folding): Likewise.
4022 * chkp-builtins.def: Remove.
4023 * ipa-chkp.c: Remove.
4024 * ipa-chkp.h: Remove.
4025 * rtl-chkp.c: Remove.
4026 * rtl-chkp.h: Remove.
4027 * tree-chkp-opt.c: Remove.
4028 * tree-chkp.c: Remove.
4029 * tree-chkp.h: Remove.
4031 2018-06-07 Carl Love <cel@us.ibm.com>
4033 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4034 vextract_fp_from_shortl): Add BE support.
4036 2018-06-07 Paul Koning <ni1d@arrl.net>
4038 * compare-elim.c (try_merge_compare): Don't merge compare if
4039 address contains a side effect.
4040 (try_eliminate_compare): Likewise.
4042 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4044 * config.gcc: Support "tremont".
4045 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4046 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4048 * config/i386/i386.c (m_TREMONT): Define.
4049 (processor_target_table): Add "tremont".
4050 (PTA_TREMONT): Define.
4051 (ix86_lea_outperforms): Add TARGET_TREMONT.
4052 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4053 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4054 and M_INTEL_GOLDMONT_PLUS.
4055 (fold_builtin_cpu): Add "tremont".
4056 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4057 (ix86_option_override_internal): Add "tremont".
4058 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4059 (processor_type): Add PROCESSOR_TREMONT.
4060 * config/i386/x86-tune.def: Add m_TREMONT.
4061 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4063 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4065 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4066 symbol defined for msp430i* devices to be lower case.
4068 gcc/testsuite/gcc.target/msp430/
4069 * msp430i-device-symbol.c: New test.
4070 * msp430f-device-symbol.c: New test.
4071 * msp430.h: New test header file.
4073 2018-06-07 Richard Biener <rguenther@suse.de>
4075 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4076 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4077 Properly wrap signed arithmetic if overflow wraps.
4079 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4081 PR tree-optimization/69615
4082 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4083 of a cast from a same precision integral SSA_NAME in a bb dominated
4084 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4085 cast to utype if rhs2 has already a compatible type.
4087 2018-06-07 Richard Biener <rguenther@suse.de>
4089 PR tree-optimization/85935
4090 * graphite-scop-detection.c (find_params_in_bb): Analyze
4091 condition operands with respect to the correct loop. Assert
4092 the analysis doesn't fail.
4094 2018-06-04 Carl Love <cel@us.ibm.com>
4096 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4097 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4098 as it is slightly cheaper.
4099 (first_match_or_eos_index_<mode>):
4100 Calculate index using natural element order.
4101 (first_match_index_<mode>):
4102 Calculate index using natural element order.
4103 (first_match_or_eos_index_<mode>):
4104 Calculate index using natural order.
4105 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4106 for BE and LE modes.
4107 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4108 P9V_BUILTIN_VCLZLSBB_V16QI.
4109 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4112 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4114 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4115 indentation and line wrap for many prototypes. Add missing
4116 @smallexample directives around block of prototypes for vec_xl and
4119 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4121 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4122 track if we pass or return IEEE 128-bit floating point.
4123 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4124 C++ mangling that is compatible with GCC 8.1.
4125 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4126 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4127 floating point types.
4128 (rs6000_function_arg_advance_1): Likewise.
4129 (rs6000_mangle_type): Optionally generate mangled names that match
4130 what GCC 8.1 generated for IEEE 128-bit floating point types.
4131 (rs6000_globalize_decl_name): If we have an external function that
4132 passes or returns IEEE 128-bit types, generate a weak reference
4133 from the mangled name used in GCC 8.1 to the current mangled
4135 (rs6000_init_builtins): Make __ibm128 use the long double type if
4136 long double is IBM extended double. Make __float128 use the long
4137 double type if long double is IEEE 128-bit.
4140 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4141 macro for __ibm128 built-in functions.
4142 (PACK_IF): Add __ibm128 pack/unpack functions.
4143 (UNPACK_IF): Likewise.
4144 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4145 enable long double built-in functions if long double is IEEE
4146 128-bit floating point.
4147 (rs6000_invalid_builtin): Update long double built-in function
4149 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4150 functions, adjust the built-in function to use the long double
4151 built-in function if __ibm128 and long double are the same type.
4152 * doc/extend.texi (PowerPC builtins): Update documention for
4153 __builtin_{,un}pack_longdouble. Add documentation for
4154 __builtin_{,un}pack_ibm128.
4156 2018-06-06 Jim Wilson <jimw@sifive.com>
4158 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4159 (struct machine_function): New field interrupt_mode.
4160 (riscv_handle_type_attribute): New function. Add forward declaration.
4161 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4162 (riscv_expand_epilogue): Check interrupt_mode field.
4163 (riscv_set_current_function): Check interrupt attribute args and
4164 set interrupt_mode field.
4165 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4166 (riscv_sret, riscv_uret): New.
4167 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4168 new arguments to interrupt attribute.
4170 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4173 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4174 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4176 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4178 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4179 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4180 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4181 several redundant entries.
4183 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4185 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4186 type from "rtx" to "rtx_insn *".
4187 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4188 for local "call_insn", removing cast.
4189 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4191 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4193 PR tree-optimization/86066
4194 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4195 for BIT_INSERT_EXPR stores.
4197 2018-06-06 Richard Biener <rguenther@suse.de>
4199 PR tree-optimization/86062
4200 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4201 component refs ontop
4202 of to be offsetted base.
4204 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4206 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4207 to be static and remove check on interrupt attribute name.
4209 gcc/testsuite/gcc.target/msp430/
4210 * function-attributes-4.c: New test.
4211 * static-interrupts.c: New test.
4213 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4215 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4216 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4218 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4221 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4223 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4224 Remove second argument, change how error is called.
4225 (aarch64_layout_arg): Remove second argument from
4226 aarch64_err_no_fpadvsimd call.
4227 (aarch64_init_cumulative_args): Ditto.
4228 (aarch64_gimplify_va_arg_expr): Ditto.
4229 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4231 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4233 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4234 (*simple_return_indirect_internal<mode>): Rename from
4235 simple_return_indirect_internal. Use W mode iterator.
4236 (rstorssp): New expander.
4237 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4238 (clrssbsy): New expander.
4239 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4241 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4243 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4245 (cmse_check_pointed_object): Likewise.
4247 2018-06-05 Martin Liska <mliska@suse.cz>
4249 PR gcov-profile/47618
4250 * doc/invoke.texi: Document how -fprofile-dir format
4253 2018-06-05 Richard Biener <rguenther@suse.de>
4255 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4256 removal pretend DOM info isn't available so we do not update
4257 it and only remove edges, not dominated blocks. Actually free
4258 DOM info in case we removed something. Remove unreachable blocks.
4259 (mfb_keep_latches): Work with either DOM info or marked backedges.
4260 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4261 first. Mark backedges if DOM info isn't available.
4262 (Re-)compute DOM info after cleanup_control_flow_pre.
4264 2018-06-05 Richard Biener <rguenther@suse.de>
4266 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4267 (locus_discrim_hasher::hash): Adjust.
4268 (locus_discrim_hasher::equal): Likewise.
4269 (next_discriminator_for_locus): Work on line directly.
4270 (same_line_p): Pass in expanded locus1 as well.
4271 (assign_discriminators): Avoid redundant location expansions.
4273 2018-06-05 Richard Biener <rguenther@suse.de>
4275 PR tree-optimization/86046
4276 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4277 if required after clearing TREE_ADDRESSABLE.
4279 2018-06-05 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/86047
4282 * tree-ssa-loop.c (for_each_index): Glob handling of all
4283 decls and constants and really handle all of them.
4285 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4288 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4289 qualifier_void_pointer and qualifier_const_void_pointer.
4290 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4291 (arm_init_builtins): Handle the above.
4292 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4293 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4294 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4297 2018-06-05 Martin Liska <mliska@suse.cz>
4299 * auto-profile.c (read_autofdo_file): Do not use
4300 gcov_ctr_summary struct.
4301 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4302 * coverage.c (struct counts_entry): Likewise.
4303 (read_counts_file): Read just single summary entry.
4304 (get_coverage_counts): Use gcov_summary struct.
4305 * coverage.h (get_coverage_counts): Likewise.
4306 * gcov-dump.c (dump_working_sets): Likewise.
4307 (tag_summary): Dump just single summary.
4308 * gcov-io.c (gcov_write_summary): Write just histogram
4310 (gcov_read_summary): Read just single summary.
4311 (compute_working_sets): Use gcov_summary struct.
4312 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4313 of GCOV_COUNTERS_SUMMABLE.
4314 (GCOV_COUNTERS_SUMMABLE): Remove.
4315 (GCOV_FIRST_VALUE_COUNTER): Replace with
4316 GCOV_COUNTER_V_INTERVAL.
4317 (struct gcov_ctr_summary): Remove.
4318 (struct gcov_summary): Directly use fields of former
4320 (compute_working_sets): Use gcov_summary struct.
4321 * gcov.c (read_count_file): Do not use ctrs fields.
4322 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4324 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4326 * profile.c: Likewise.
4327 * profile.h: Likewise.
4329 2018-06-05 Martin Liska <mliska@suse.cz>
4331 PR gcov-profile/84846
4332 * gcov.c (output_lines): Print working directory only
4333 in intermediate format.
4335 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4337 * config/s390/s390-builtin-types.def: Add void function type.
4338 * config/s390/s390-builtins.def: Use the function type for the
4341 2018-06-04 Jim Wilson <jimw@sifive.com>
4343 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4345 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4346 and maybe_eh_return. Change regno to unsigned int. Use new args to
4347 handle EH_RETURN_DATA_REGNO registers properly.
4348 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4349 (riscv_expand_epilogue): Update comment. Change argument name and
4350 type. Update code to use new name and type. Pass new args to
4351 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4353 * config/riscv/riscv.md (NORMAL_RETURN): New.
4354 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4355 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4356 (eh_return): Call gen_eh_return_internal and emit barrier.
4357 (eh_return_internal): Call riscv_expand_epilogue.
4359 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4361 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4362 bit_insertion field and declare can_be_merged_into method.
4363 (merged_store_group::can_be_merged_into): New method.
4364 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4365 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4366 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4368 2018-06-04 Richard Biener <rguenther@suse.de>
4370 PR tree-optimization/85955
4371 * builtins.c (fold_builtin_sincos): Convert pointers to
4372 destination to appropriate type before dereferencing.
4374 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4376 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4378 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4380 * expr.c (expand_expr_real_1): Force the operand into memory if
4381 its TYPE_MODE is BLKmode and if there is no integer mode for
4382 the number of bits being extracted.
4384 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4388 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4389 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4391 2018-06-04 Richard Biener <rguenther@suse.de>
4393 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4394 (cleanup_tree_cfg_noloop): ... single caller. Do
4395 start_recording_case_labels later.
4397 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4399 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4400 to _IMMINTRIN_H_INCLUDED.
4401 * config/i386/pconfigintrin.h: Ditto.
4402 * config/i386/waitpkgintrin.h: Ditto.
4403 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4404 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4405 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4406 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4407 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4408 waitpkgintrin.h and cldemoteintrin.h.
4410 2018-06-04 Richard Biener <rguenther@suse.de>
4412 PR tree-optimization/86038
4413 * tracer.c (find_best_successor): Check probability for
4414 being initialized, bail out if not.
4416 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4419 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4420 of bits to ignore when comparing architectures.
4422 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4424 PR tree-optimization/69615
4425 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4426 maximum or minimum of the type, try to merge it also as if
4427 range1 is + [-, x - 1] or + [x + 1, -].
4430 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4432 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4434 PR tree-optimization/86034
4435 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4436 the unsigned bitfield type in a bit insertion sequence if it does not
4437 have a larger precision than the bitfield size.
4438 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4440 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4442 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4444 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4447 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4448 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4449 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4451 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4453 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4454 Disable -fdelete-null-pointer-checks for ELF toolchain.
4456 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4457 Kito Cheng <kito.cheng@gmail.com>
4459 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4460 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4461 (nds32*-*-*): Add glibc and uclibc conditions.
4462 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4463 (TARGET_EXCEPT_UNWIND_INFO): Define.
4464 * config/nds32/elf.h: New file.
4465 * config/nds32/linux.h: New file.
4466 * config/nds32/nds32-elf.opt: New file.
4467 * config/nds32/nds32-linux.opt: New file.
4468 * config/nds32/nds32-fp-as-gp.c
4469 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4470 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4472 (nds32_asm_file_end): Ditto.
4473 (nds32_print_operand): Ditto.
4474 (nds32_insert_attributes): Ditto.
4475 (nds32_init_libfuncs): New function.
4476 (TARGET_HAVE_TLS): Define.
4477 (TARGET_INIT_LIBFUNCS): Define.
4478 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4480 (TARGET_ELF): Apply different mcmodel setting.
4481 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4482 been migrated into elf.h and linux.h files.
4483 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4484 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4485 (mcmodel): The content has been migrated into nds32-elf.opt and
4486 nds32-linux.opt files.
4487 * config/nds32/t-elf: New file.
4488 * config/nds32/t-linux: New file.
4490 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4491 Shiva Chen <shiva0217@gmail.com>
4493 * config/nds32/constants.md (unspec_volatile_element): Add
4494 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4495 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4497 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4498 (make_pass_nds32_fp_as_gp): Declare.
4499 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4501 (nds32_asm_function_end_prologue): Remove unused asm output.
4502 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4503 (nds32_asm_file_start): Output necessary fp_as_gp information.
4504 (nds32_option_override): Adjust register usage.
4505 (nds32_expand_prologue): Consider fp_as_gp situation.
4506 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4507 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4510 (simple_return): Ditto.
4511 (omit_fp_begin): Output special directive for fp_as_gp.
4512 (omit_fp_end): Output special directive for fp_as_gp.
4513 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4514 mforbid-fp-as-gp): New options.
4516 2018-06-01 Mark Wielaard <mark@klomp.org>
4518 * dwarf2out.c (dwarf2out_finish): Remove generation of
4519 DW_AT_loclists_base.
4521 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4523 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4524 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4525 (struct merged_store_group): Add bit_insertion field.
4526 (dump_char_array): Use standard hexadecimal format.
4527 (merged_store_group::merged_store_group): Set bit_insertion to false.
4528 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4529 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4530 also print the mask in the dump file.
4531 (pass_store_merging::gate): Minor tweak.
4532 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4533 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4534 stores with INTEGER_CST stores.
4535 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4536 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4537 and use it throughout. Generate bit insertion sequences if need be.
4538 (pass_store_merging::process_store): Remove redundant condition.
4539 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4541 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4543 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4544 the 128-bit floating point types. Fix function comment.
4546 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4548 * config/aarch64/aarch64-simd.md
4549 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4551 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4554 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4556 PR tree-optimization/85989
4557 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4559 (backprop::intersect_uses): Check it when deciding whether this
4560 is a backedge reference.
4561 (backprop::process_block): Add each phi to m_visited_phis
4562 after visiting it, then clear it at the end.
4564 2018-06-01 Richard Biener <rguenther@suse.de>
4566 * tree-vectorizer.h (vect_dr_stmt): New function.
4567 (vect_get_load_cost): Adjust.
4568 (vect_get_store_cost): Likewise.
4569 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4570 Use vect_dr_stmt instead of DR_SMTT.
4571 (vect_record_base_alignments): Likewise.
4572 (vect_calculate_target_alignment): Likewise.
4573 (vect_compute_data_ref_alignment): Likewise and make static.
4574 (vect_update_misalignment_for_peel): Likewise.
4575 (vect_verify_datarefs_alignment): Likewise.
4576 (vector_alignment_reachable_p): Likewise.
4577 (vect_get_data_access_cost): Likewise. Pass down
4578 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4579 (vect_get_peeling_costs_all_drs): Likewise.
4580 (vect_peeling_hash_get_lowest_cost): Likewise.
4581 (vect_enhance_data_refs_alignment): Likewise.
4582 (vect_find_same_alignment_drs): Likewise.
4583 (vect_analyze_data_refs_alignment): Likewise.
4584 (vect_analyze_group_access_1): Likewise.
4585 (vect_analyze_group_access): Likewise.
4586 (vect_analyze_data_ref_access): Likewise.
4587 (vect_analyze_data_ref_accesses): Likewise.
4588 (vect_vfa_segment_size): Likewise.
4589 (vect_small_gap_p): Likewise.
4590 (vectorizable_with_step_bound_p): Likewise.
4591 (vect_prune_runtime_alias_test_list): Likewise.
4592 (vect_analyze_data_refs): Likewise.
4593 (vect_supportable_dr_alignment): Likewise.
4594 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4595 (vect_gen_prolog_loop_niters): Likewise.
4596 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4597 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4599 (vect_recog_mask_conversion_pattern): Likewise.
4600 (vect_try_gather_scatter_pattern): Likewise.
4601 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4602 to vect_get_store_cost.
4603 (vect_get_store_cost): Get stmt_info instead of DR.
4604 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4605 (vect_get_load_cost): Get stmt_info instead of DR.
4607 2018-06-01 Richard Biener <rguenther@suse.de>
4610 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4611 vars wrapped in MEM_REFs.
4613 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4615 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4616 Fix subreg tests so that we only return a choice between
4617 GENERAL_REGS and FP_REGS if the original classes included both.
4619 2018-06-01 Richard Biener <rguenther@suse.de>
4622 * tree-ssa-structalias.c (get_function_part_constraint):
4623 Handle NULL fi->decl.
4624 (find_func_aliases_for_call): Properly handle indirect
4625 fi from direct call.
4626 (find_func_clobbers): Likewise.
4627 (ipa_pta_execute): Likewise.
4628 (create_variable_info_for): For functions that are ifunc_resolver
4629 resolve to a varinfo that contains the result of the resolver call.
4630 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4633 2018-05-31 Michael Collison <michael.collison@arm.com>
4635 * config/aarch64/aarch64.md:
4636 (*fix_to_zero_extenddfdi2): New pattern.
4637 * gcc.target/aarch64/fix_extend1.c: New testcase.
4639 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4643 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4644 and BUILT_IN_STRNCMP_EQ.
4645 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4646 BUILT_IN_STRNCMP_EQ.
4647 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4648 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4649 (gimple_fold_builtin): Likewise.
4650 * tree-ssa-strlen.c (compute_string_length): New function.
4651 (determine_min_obsize): New function.
4652 (handle_builtin_string_cmp): New function to handle calls to
4653 string compare functions.
4654 (strlen_optimize_stmt): Add handling to builtin string compare
4656 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4657 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4658 * tree.c (build_common_builtin_nodes): Add new defines of
4659 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4661 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4664 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4665 functions with naked attribute.
4667 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4669 * config/i386/sse.md (avx_vec_concat<mode>):
4670 Substitute concat_tg_mode mode attribute with xtg_mode.
4671 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4672 (concat_tg_mode): Remove mode attribute.
4674 2018-05-31 Martin Sebor <msebor@redhat.com>
4677 * calls.c (alloc_max_size): Correct a logic error/typo.
4678 Treat excessive arguments as infinite. Warn for invalid arguments.
4679 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4681 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4684 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4685 and movx for Haswell.
4687 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4688 Cesar Philippidis <cesar@codesourcery.com>
4691 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4692 when emitting error on private/firstprivate reductions.
4693 * omp-low.c (lower_omp_target): Avoid reference-type processing
4694 on pointers for firstprivate clause.
4696 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4698 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4701 * config/aarch64/aarch64-simd.md
4702 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4703 (aarch64_ld1_x3_<mode>): Likewise
4704 (aarch64_st1x2<VALLDIF:mode>): Likewise
4705 (aarch64_st1_x2_<mode>): Likewise
4706 (aarch64_st1x3<VALLDIF:mode>): Likewise
4707 (aarch64_st1_x3_<mode>): Likewise
4708 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4709 (vld1_s8_x3): Likewise.
4710 (vld1_u16_x3): Likewise.
4711 (vld1_s16_x3): Likewise.
4712 (vld1_u32_x3): Likewise.
4713 (vld1_s32_x3): Likewise.
4714 (vld1_u64_x3): Likewise.
4715 (vld1_s64_x3): Likewise.
4716 (vld1_f16_x3): Likewise.
4717 (vld1_f32_x3): Likewise.
4718 (vld1_f64_x3): Likewise.
4719 (vld1_p8_x3): Likewise.
4720 (vld1_p16_x3): Likewise.
4721 (vld1_p64_x3): Likewise.
4722 (vld1q_u8_x3): Likewise.
4723 (vld1q_s8_x3): Likewise.
4724 (vld1q_u16_x3): Likewise.
4725 (vld1q_s16_x3): Likewise.
4726 (vld1q_u32_x3): Likewise.
4727 (vld1q_s32_x3): Likewise.
4728 (vld1q_u64_x3): Likewise.
4729 (vld1q_s64_x3): Likewise.
4730 (vld1q_f16_x3): Likewise.
4731 (vld1q_f32_x3): Likewise.
4732 (vld1q_f64_x3): Likewise.
4733 (vld1q_p8_x3): Likewise.
4734 (vld1q_p16_x3): Likewise.
4735 (vld1q_p64_x3): Likewise.
4736 (vst1_s64_x2): Likewise.
4737 (vst1_u64_x2): Likewise.
4738 (vst1_f64_x2): Likewise.
4739 (vst1_s8_x2): Likewise.
4740 (vst1_p8_x2): Likewise.
4741 (vst1_s16_x2): Likewise.
4742 (vst1_p16_x2): Likewise.
4743 (vst1_s32_x2): Likewise.
4744 (vst1_u8_x2): Likewise.
4745 (vst1_u16_x2): Likewise.
4746 (vst1_u32_x2): Likewise.
4747 (vst1_f16_x2): Likewise.
4748 (vst1_f32_x2): Likewise.
4749 (vst1_p64_x2): Likewise.
4750 (vst1q_s8_x2): Likewise.
4751 (vst1q_p8_x2): Likewise.
4752 (vst1q_s16_x2): Likewise.
4753 (vst1q_p16_x2): Likewise.
4754 (vst1q_s32_x2): Likewise.
4755 (vst1q_s64_x2): Likewise.
4756 (vst1q_u8_x2): Likewise.
4757 (vst1q_u16_x2): Likewise.
4758 (vst1q_u32_x2): Likewise.
4759 (vst1q_u64_x2): Likewise.
4760 (vst1q_f16_x2): Likewise.
4761 (vst1q_f32_x2): Likewise.
4762 (vst1q_f64_x2): Likewise.
4763 (vst1q_p64_x2): Likewise.
4764 (vst1_s64_x3): Likewise.
4765 (vst1_u64_x3): Likewise.
4766 (vst1_f64_x3): Likewise.
4767 (vst1_s8_x3): Likewise.
4768 (vst1_p8_x3): Likewise.
4769 (vst1_s16_x3): Likewise.
4770 (vst1_p16_x3): Likewise.
4771 (vst1_s32_x3): Likewise.
4772 (vst1_u8_x3): Likewise.
4773 (vst1_u16_x3): Likewise.
4774 (vst1_u32_x3): Likewise.
4775 (vst1_f16_x3): Likewise.
4776 (vst1_f32_x3): Likewise.
4777 (vst1_p64_x3): Likewise.
4778 (vst1q_s8_x3): Likewise.
4779 (vst1q_p8_x3): Likewise.
4780 (vst1q_s16_x3): Likewise.
4781 (vst1q_p16_x3): Likewise.
4782 (vst1q_s32_x3): Likewise.
4783 (vst1q_s64_x3): Likewise.
4784 (vst1q_u8_x3): Likewise.
4785 (vst1q_u16_x3): Likewise.
4786 (vst1q_u32_x3): Likewise.
4787 (vst1q_u64_x3): Likewise.
4788 (vst1q_f16_x3): Likewise.
4789 (vst1q_f32_x3): Likewise.
4790 (vst1q_f64_x3): Likewise.
4791 (vst1q_p64_x3): Likewise.
4793 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4795 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4796 user_label_prefix to name.
4798 * tree-core.h: Update comment about the format of NAME string
4799 passed to handler in attribute_spec.
4801 * config/msp430/msp430.md: Remove erroneous subreg expression from
4802 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4803 zero_extend{q,h}isi2.
4805 2018-05-30 Borislav Petkov <bp@suse.de>
4807 * doc/extend.texi: Document some architecture specific
4808 constraints and sort entries.
4810 2018-05-30 Martin Sebor <msebor@redhat.com>
4813 * builtins.c (expand_builtin_stpcpy_1): New function.
4814 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4815 only if the former succeeds.
4817 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4819 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4822 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4824 * doc/invoke.texi (-flinker-output): Document
4826 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4828 * passes.c (ipa_write_summaries): Only modify statements if body
4830 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4831 incrementally linking.
4832 (ipa_passes): Likewise.
4833 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4834 pass down resolution info.
4835 * common.opt (flag_incremental_link): Update info.
4836 * gcc.c (plugin specs): Turn flinker-output=* to
4837 -plugin-opt=-linker-output-known
4838 * toplev.c (compile_file): Also cut compilation when doing incremental
4840 * flag-types. (enum lto_partition_model): Add
4841 LTO_LINKER_OUTPUT_NOLTOREL.
4842 (invoke.texi): Add -flinker-output docs.
4843 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4844 link same way as WPA; do not stream in dead initializers.
4846 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4847 darf2out_register_external_decl): Support incremental link.
4849 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4851 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4853 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4855 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4856 it down to simple_object_copy_lto_debug_sections.
4857 (run_gcc): Determine incremental LTO link time and configure
4858 lto1 into non-wpa mode, disable renaming of debug sections.
4860 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4862 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4863 descriptions of various incorrectly documented functions.
4865 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4868 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4869 address check not strict.
4871 2018-05-30 Richard Biener <rguenther@suse.de>
4873 PR tree-optimization/85964
4874 * tracer.c (better_p): Drop initialized count check, we only
4875 call the function with initialized counts now.
4876 (find_best_successor): Do find a best edge if one
4877 has uninitialized count.
4878 (find_best_predecessor): Likewise. Do BB frequency check only
4879 if count is initialized.
4881 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4883 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4884 (aarch64_ldrstr_offset_compare): New.
4885 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4886 load/store orderings.
4887 (aarch64_gen_adjusted_ldpstp): Likewise.
4889 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4891 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4892 Check for subset of GENERAL_REGS and FP_REGS.
4893 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4896 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4898 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4899 and wi::to_poly_offset. Add the current offset and then check
4900 whether the sum fits, rather than using an unchecked addition of
4901 a checked term. Check for a shwi rather than a uhwi.
4902 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4903 (store_constructor): Use poly_int_tree_p.
4904 (expand_expr_real_1): Likewise.
4905 * function.c (assign_temp): Likewise.
4906 * fold-const.c (const_binop): Use poly_int_tree_p and
4908 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4910 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4911 to_poly_offset for MEM offsets.
4912 * ipa-icf.c (sem_variable::equals): Likewise.
4913 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4914 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4915 wi::to_poly_offset for BIT_FIELD_REF offsets.
4916 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4918 * var-tracking.c (emit_note_insn_var_location): Use
4919 tree_to_poly_uint64.
4921 2018-05-29 Jim Wilson <jimw@sifive.com>
4923 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4925 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4928 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4929 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4931 (sse4_1_round<mode>2): Use nonimmediate_operand
4932 for operand 1 predicate.
4934 2018-05-29 Martin Sebor <msebor@redhat.com>
4935 Richard Biener <rguenther@suse.de>
4938 * calls.c (get_size_range): Call determine_value_range instead
4939 of get_value_range..
4940 * tree-vrp.h (determine_value_range): Declared new function.
4941 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4943 2018-05-29 Richard Biener <rguenther@suse.de>
4945 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4946 sure to use non-pattern stmts for get_earlier_stmt arguments.
4947 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4948 called on pattern stmts.
4949 (get_later_stmt): Likewise.
4951 2018-05-29 Martin Liska <mliska@suse.cz>
4953 PR gcov-profile/85759
4954 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4957 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4959 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4961 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4965 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4966 VEC_PACK_FLOAT_EXPR): New tree codes.
4967 * tree-pretty-print.c (op_code_prio): Handle
4968 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4969 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4970 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4971 * tree-inline.c (estimate_operator_cost): Likewise.
4972 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4973 * fold-const.c (const_binop): Likewise.
4974 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4975 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4976 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4977 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4978 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4979 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4980 * expr.c (expand_expr_real_2): Likewise.
4981 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4982 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4983 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4985 * optabs.c (expand_widen_pattern_expr): For
4986 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4987 sign from result type rather than operand's type.
4988 (expand_binop_directly): For vec_packu_float_optab and
4989 vec_packs_float_optab allow result type to be different from operand's
4991 * optabs-tree.c (optab_for_tree_code): Handle
4992 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4993 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4994 * tree-vect-generic.c (expand_vector_operations_1): Handle
4995 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4996 VEC_PACK_FLOAT_EXPR.
4997 * tree-vect-stmts.c (supportable_widening_operation): Handle
4999 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5000 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5001 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5002 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5003 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5005 (vec_pack<floatprefix>_float_<mode>): New expander.
5006 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5008 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5009 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5010 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5011 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5012 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5014 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5015 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5016 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5017 VEC_PACK_FLOAT_EXPR): Document.
5019 2018-05-29 Richard Biener <rguenther@suse.de>
5021 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5023 (stmt_vec_info_vec): Make pointer.
5024 (init_stmt_vec_info_vec): Remove.
5025 (free_stmt_vec_info_vec): Likewise.
5026 (set_stmt_vec_info_vec): New function.
5027 (free_stmt_vec_infos): Likewise.
5028 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5029 (set_vinfo_for_stmt): Likewise.
5030 (get_earlier_stmt): Likewise.
5031 (get_later_stmt): Likewise.
5032 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5033 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5034 (vec_info::~vec_info): Free stmt_vec_infos.
5035 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5036 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5037 (pass_slp_vectorize::execute): Likewise.
5038 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5039 (free_stmt_vec_info_vec): Likewise.
5040 (set_stmt_vec_info_vec): New function.
5041 (free_stmt_vec_infos): Likewise.
5042 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5043 the global stmt_vec_info_vec.
5044 * tree-parloops.c (gather_scalar_reductions): Use
5045 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5048 2018-05-29 Richard Biener <rguenther@suse.de>
5050 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5052 2018-05-29 Martin Liska <mliska@suse.cz>
5053 David Malcolm <dmalcolm@redhat.com>
5055 * vec.c (test_reverse): New.
5056 (vec_c_tests): Add new test.
5057 * vec.h (vl_ptr>::reverse): New function.
5059 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5061 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5063 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5066 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5068 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5070 2018-05-28 Richard Biener <rguenther@suse.de>
5072 PR tree-optimization/85933
5073 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5074 look at stmts marked as vectorizable.
5076 2018-05-28 Richard Biener <rguenther@suse.de>
5078 PR tree-optimization/85934
5079 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5080 vector boolean check before scalar optimization.
5082 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5084 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5087 2018-05-28 Mark Wielaard <mark@klomp.org>
5089 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5090 if it is an expression containing a minus sign.
5092 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5094 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5096 2018-05-27 Paul Koning <ni1d@arrl.net>
5098 * config/pdp11/pdp11.md (truncsihi2): Remove.
5100 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5101 Chung-Ju Wu <jasonwucj@gmail.com>
5103 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5105 (unaligned_store_dw): Ditto.
5106 * config/nds32/nds32-memory-manipulation.c
5107 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5108 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5109 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5110 (emit_setmem_word_loop): Rename to ...
5111 (emit_setmem_doubleword_loop): ... this.
5112 (nds32_gen_dup_4_byte_to_word_value): New function.
5113 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5114 (nds32_expand_setmem_loop): Refine implementation.
5115 (nds32_expand_setmem_loop_v3m): Ditto.
5116 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5119 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5121 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5123 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5125 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5126 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5127 machine->attr_no_prologue_p.
5128 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5129 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5130 (nds32_expand_epilogue): Consider attr_naked_p.
5131 (nds32_expand_epilogue_v3pop): Likewise.
5132 (nds32_can_use_return_insn): Likewise.
5133 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5134 attr_no_prologue_p fields.
5135 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5137 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5140 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5142 * config/i386/sse.md
5143 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5145 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5147 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5149 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5151 (*<floatsuffix>floatv2div2sf2): Rename to ...
5152 (*float<floatunssuffix>v2div2sf2): ... this.
5153 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5154 (float<floatunssuffix>v2div2sf2_mask): ... this.
5155 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5156 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5157 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5159 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5161 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5163 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5165 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5167 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5169 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5170 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5171 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5172 gen_ufix_truncv8dfv8si2.
5173 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5174 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5175 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5176 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5177 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5178 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5179 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5180 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5182 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5186 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5188 2018-05-25 Jim Wilson <jimw@sifive.com>
5190 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5191 * config/riscv/riscv.c (struct machine_function): Add
5192 interrupt_handler_p and attribute_checked_p fields.
5193 (riscv_attribute_table): Add interrupt.
5194 (riscv_interrupt_type_p): New.
5195 (riscv_save_reg_p): Save extra regs for interrupt handler.
5196 (riscv_use_save_libcall): Return false for interrupt handler.
5197 (riscv_first_stack_step): Add forward declaration.
5198 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5199 for interrupt handler with large frame. Use it for saved reg list.
5200 (riscv_expand_prologue): Move flag_stack_usage_info support to
5201 eliminate duplication.
5202 (riscv_expand_epilogue): Generate mret for interrupt handler.
5203 (riscv_epilogue_uses): New.
5204 (riscv_can_use_return_insn): Return false for interrupt handler.
5205 (riscv_function_ok_for_sibcall): Likewise.
5206 (riscv_set_current_function): Add interrupt handler support.
5207 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5208 * config/riscv/riscv.md (UNSPECV_MRET): New.
5210 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5212 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5214 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5216 PR tree-optimization/85712
5217 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5218 this candidate has already been replaced in-situ by a copy.
5220 2018-05-25 Jason Merrill <jason@redhat.com>
5222 PR c++/80485 - inline function non-zero address.
5223 * symtab.c (nonzero_address): Check DECL_COMDAT.
5225 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5228 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5229 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5230 extension of SImode operation. Use const123_operand predicate.
5231 (*saddsi_1): Remove.
5232 (*saddl_se_1): Ditto.
5234 (*ssubl_se_1): Ditto.
5235 * config/alpha/predicates.md (const123_operand): New predicate.
5236 * config/alpha/constraints.md (P): Use IN_RANGE.
5238 2018-05-25 Richard Biener <rguenther@suse.de>
5240 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5242 (ref_maybe_used_by_stmt_p): Likewise.
5243 (stmt_may_clobber_ref_p): Likewise.
5244 (stmt_may_clobber_ref_p_1): Likewise.
5245 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5247 (ref_maybe_used_by_stmt_p): Likewise.
5248 (stmt_may_clobber_ref_p): Likewise.
5249 (stmt_may_clobber_ref_p_1): Likewise.
5250 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5251 the alias oracle to disambiguate DRs with stmts DR analysis
5253 (vect_analyze_data_refs): Do not give up on not analyzable
5254 DRs for BB vectorization. Remove code truncating the dataref
5257 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5260 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5261 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5262 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5264 2018-05-25 Richard Biener <rguenther@suse.de>
5266 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5267 function, combining stmt data ref gathering and fatal analysis
5269 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5270 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5271 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5272 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5273 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5275 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5277 PR tree-optimization/85720
5278 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5279 SCC if all partitions are builtins.
5280 (version_loop_by_alias_check): New parameter. Generate cancelable
5281 runtime alias check if all partitions are builtins.
5282 (distribute_loop): Update call to above function.
5284 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5286 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5287 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5288 (parm_default_def_partition_arg): Ditto.
5289 (set_parm_default_def_partition): Ditto.
5290 (get_parm_default_def_partitions): Ditto and make it static.
5291 (get_undefined_value_partitions): Ditto and make it static.
5292 (remove_ssa_form): Refactor call to init_var_map here.
5293 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5294 computation for loop region.
5295 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5296 (register_default_def): Delete.
5297 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5298 (parm_default_def_partition_arg): Ditto.
5299 (set_parm_default_def_partition): Ditto.
5300 (get_parm_default_def_partitions): Ditto and make it static.
5301 (get_undefined_value_partitions): Ditto and make it static.
5302 (coalesce_with_default, coalesce_with_default): Update comment.
5303 (create_coalesce_list_for_region): New func factored out from
5304 create_outofssa_var_map.
5305 (populate_coalesce_list_for_outofssa): New func factored out from
5306 create_outofssa_var_map and coalesce_ssa_name.
5307 (create_outofssa_var_map): Delete.
5308 (coalesce_ssa_name): Refactor to support live range computation.
5309 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5310 (get_parm_default_def_partitions): Delete.
5311 (get_undefined_value_partitions): Ditto.
5312 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5313 computation for loop region.
5314 (new_tree_live_info, loe_visit_block): Ditto.
5315 (live_worklist, set_var_live_on_entry): Ditto.
5316 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5317 * tree-ssa-live.h (struct _var_map): New fields.
5318 (init_var_map): Change decl.
5319 (region_contains_p): New.
5321 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5323 * tree-ssa-live.h (live_merge_and_clear): Delete.
5325 2018-05-25 Richard Biener <rguenther@suse.de>
5328 * tree-dump.c (dequeue_and_dump): Remove access to removed
5329 operand 2 of a SWITCH_EXPR.
5331 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5333 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5334 multiplication and division.
5335 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5336 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5337 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5338 (cond_udiv_optab, cond_umod_optab): New optabs.
5339 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5340 (IFN_COND_RDIV): New internal functions.
5341 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5342 TRUNC_MOD_EXPR and RDIV_EXPR.
5343 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5344 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5346 (SVE_INT_BINARY): Include mult.
5347 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5348 (optab, sve_int_op): Handle mult.
5349 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5351 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5352 for SVE_INT_BINARY_SD.
5354 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5356 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5357 (optab, sve_int_op): Handle div and udiv.
5358 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5359 for SVE_INT_BINARY_SD.
5360 (*<optab><mode>3): New insn for the same.
5362 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5364 * tree-vect-patterns.c: Include predict.h.
5365 (vect_recog_divmod_pattern): Restrict check for division support
5366 to when optimizing for size.
5368 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5370 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5371 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5372 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5373 (gimple_match_op::set_op): Likewise.
5374 (gimple_resimplify4): Declare.
5375 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5376 (expr::gen_transform): Likewise.
5377 (decision_tree::gen): Generate a simplification routine for 4 operands.
5378 * gimple-match-head.c (gimple_simplify): Add an overload for
5379 4 operands. In the top-level function, handle up to 4 call
5380 arguments and call gimple_resimplify4.
5381 (gimple_resimplify4): New function.
5382 (build_call_internal): Pass a fourth operand.
5383 (maybe_push_to_seq): Likewise.
5384 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5385 Fold VEC_COND_EXPRs of an operation and a default value into
5386 an IFN_COND_* function if possible.
5387 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5389 (SVE_COND_FP_BINARY): Include them.
5390 (optab, sve_fp_op): Handle them.
5391 (SVE_INT_BINARY_REV): New code iterator.
5392 (SVE_COND_FP_BINARY_REV): New int iterator.
5393 (commutative): New int attribute.
5394 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5396 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5398 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5399 (*cond_<optab><mode>): New patterns for reversed operands.
5401 2018-05-25 Richard Biener <rguenther@suse.de>
5403 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5404 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5405 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5406 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5407 * tree-vect-data-refs.c (everywhere): Adjust users.
5408 * tree-vect-loop.c (everywhere): Likewise.
5409 * tree-vect-slp.c (everywhere): Likewise.
5410 * tree-vect-stmts.c (everywhere): Likewise.
5411 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5413 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5415 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5417 (gcc_cv_as_section_exclude): ... this.
5418 Try Solaris as #exclude syntax.
5419 * configure: Regenerate.
5420 * config.in: Regenerate.
5421 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5423 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5424 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5426 * varasm.c (default_elf_asm_named_section): Don't check if
5427 HAVE_GAS_SECTION_EXCLUDE is defined.
5429 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5431 * doc/md.texi: Update the documentation of the cond_* optabs
5432 to mention the new final operand. Fix GET_MODE_NUNITS call.
5433 Describe the scalar case too.
5434 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5435 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5437 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5438 (get_conditional_internal_fn): Update comment.
5439 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5440 accumulator value as a final argument to conditional functions.
5441 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5442 a define_expand and add an "else" operand. Assert for now that
5443 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5444 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5445 (*cond_<optab><mode>): New patterns.
5446 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5447 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5448 (UNSPEC_COND_EOR): Delete.
5449 (optab): Remove associated mappings.
5450 (SVE_INT_BINARY): New code iterator.
5451 (sve_int_op): Remove int attribute and add "minus" to the code
5453 (SVE_COND_INT_OP): Delete.
5454 (SVE_COND_FP_OP): Rename to...
5455 (SVE_COND_FP_BINARY): ...this.
5457 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5459 * optabs.c (can_reuse_operands_p): New function.
5460 (maybe_legitimize_operands): Try to reuse the results for
5463 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5465 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5466 Add {q} suffix to insn mnemonic.
5468 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5470 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5471 (msp430_warn_func_return): New.
5473 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5475 * fold-const.c (tree_nonzero_bits): New function.
5476 * fold-const.h (tree_nonzero_bits): Likewise.
5477 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5478 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5480 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5484 * varasm.c (assemble_alias): Check ifunc_resolver only on
5487 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5490 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5491 when memory input operand is handled.
5493 2018-05-24 Luis Machado <luis.machado@linaro.org>
5495 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5497 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5499 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5501 * match.pd: Delay FMA folds until after vectorization.
5503 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5506 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5507 address check not strict.
5509 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5511 * gimple-match.h (gimple_match_op): New class.
5512 (mprts_hook): Replace parameters with a gimple_match_op *.
5513 (maybe_build_generic_op): Likewise.
5514 (gimple_simplified_result_is_gimple_val): Replace parameters with
5515 a const gimple_match_op *.
5516 (gimple_simplify): Replace code_helper * and tree * parameters with
5517 a gimple_match_op * parameter.
5518 (gimple_resimplify1): Replace code_helper *, tree and tree *
5519 parameters with a gimple_match_op * parameter.
5520 (gimple_resimplify2): Likewise.
5521 (gimple_resimplify3): Likewise.
5522 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5523 parameters with a gimple_match_op * parameter.
5524 * gimple-match-head.c (gimple_simplify): Change prototypes of
5525 auto-generated functions to take a gimple_match_op * instead of
5526 separate code_helper * and tree * parameters. Make the same
5527 change in the top-level overload and update calls to the
5528 gimple_resimplify routines. Update calls to the auto-generated
5529 functions and to maybe_push_res_to_seq in the publicly-facing
5530 operation-specific gimple_simplify overloads.
5531 (gimple_match_op::MAX_NUM_OPS): Define.
5532 (gimple_resimplify1): Replace rcode and ops with a single res_op
5533 parameter. Update call to gimple_simplify.
5534 (gimple_resimplify2): Likewise.
5535 (gimple_resimplify3): Likewise.
5536 (mprts_hook): Replace parameters with a gimple_match_op *.
5537 (maybe_build_generic_op): Likewise.
5538 (build_call_internal): Replace type, nargs and ops with
5539 a gimple_match_op *.
5540 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5541 with a single gimple_match_op *. Update calls to mprts_hook,
5542 build_call_internal and gimple_simplified_result_is_gimple_val.
5543 Factor out code that is common to the tree_code and combined_fn cases.
5544 * genmatch.c (expr::gen_transform): Replace tem_code and
5545 tem_ops with a gimple_match_op called tem_op. Update calls
5546 to the gimple_resimplify functions and maybe_push_res_to_seq.
5547 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5548 res_ops. Update call to the gimple_resimplify functions.
5549 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5550 (decision_tree::gen): Make the functions take a gimple_match_op *
5551 called res_op instead of separate res_code and res_ops parameters.
5552 Update call accordingly.
5553 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5554 and ops with a single res_op parameter. Update calls to
5555 maybe_build_generic_op and maybe_push_res_to_seq.
5556 (fold_stmt_1): Update calls to gimple_simplify and
5557 replace_stmt_with_simplification.
5558 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5559 and gimple_simplified_result_is_gimple_val.
5560 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5562 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5563 with a gimple_match_op *.
5564 (vn_nary_build_or_lookup): Likewise. Update call to
5565 vn_nary_build_or_lookup_1.
5566 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5567 gimple_match_op *. Update calls to the gimple_resimplify routines
5568 and to gimple_simplified_result_is_gimple_val.
5569 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5570 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5571 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5572 (visit_nary_op): Likewise.
5573 (visit_reference_op_load): Likewise.
5575 2018-05-23 Luis Machado <luis.machado@linaro.org>
5577 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5578 modifier for printing the step amount.
5580 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5583 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5586 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5588 * doc/sourcebuild.texi (Endianness): New subsubsection.
5590 2018-05-23 Luis Machado <luis.machado@linaro.org>
5592 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5593 <prefetch_dynamic_strides>: New const bool field.
5594 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5595 prefetch_dynamic_strides.
5596 (exynosm1_prefetch_tune): Likewise.
5597 (thunderxt88_prefetch_tune): Likewise.
5598 (thunderx_prefetch_tune): Likewise.
5599 (thunderx2t99_prefetch_tune): Likewise.
5600 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5602 (aarch64_override_options_internal): Update to set
5603 PARAM_PREFETCH_DYNAMIC_STRIDES.
5604 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5605 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5606 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5607 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5608 prefetch-dynamic-strides setting.
5610 2018-05-23 Luis Machado <luis.machado@linaro.org>
5612 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5613 <minimum_stride>: New const int field.
5614 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5615 minimum_stride field defaulting to -1.
5616 (exynosm1_prefetch_tune): Likewise.
5617 (thunderxt88_prefetch_tune): Likewise.
5618 (thunderx_prefetch_tune): Likewise.
5619 (thunderx2t99_prefetch_tune): Likewise.
5620 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5621 <default_opt_level>: Set to 3.
5622 (aarch64_override_options_internal): Update to set
5623 PARAM_PREFETCH_MINIMUM_STRIDE.
5624 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5625 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5626 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5627 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5628 stride is constant and is below the minimum stride threshold.
5630 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5632 * config/arm/arm-cpus.in (mode26): Delete.
5633 (armv4): Delete mode26 reference.
5634 * config/arm/arm.c (arm_configure_build_target): Delete use of
5637 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5639 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5641 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5642 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5644 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5645 (floatunsdidf2): Ditto.
5647 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5649 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5650 (fixuns_trunc<mode>si2_avx512f): Ditto.
5651 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5652 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5653 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5655 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5657 PR rtl-optimization/79985
5658 * df-scan.c (df_insn_refs_collect): Remove special case for
5659 global registers and asm statements.
5661 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5663 * extend.texi (Global Register Variables): Rewrite the bullet list.
5664 Note that the register is available for allocation. Note that access
5665 via inline asm must use constraints. Add note about async-signal
5666 handlers. Remove paragraph about automagic register selection.
5668 2018-05-23 Richard Biener <rguenther@suse.de>
5670 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5671 of fixed offset from memset VN.
5673 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5675 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5677 (alloc_cand_and_find_basis): Initialize first_interp field.
5678 (slsr_process_mul): Modify first_interp field.
5679 (slsr_process_add): Likewise.
5680 (slsr_process_cast): Modify first_interp field for each new
5682 (slsr_process_copy): Likewise.
5683 (dump_candidate): Dump first_interp field.
5684 (replace_mult_candidate): Process all interpretations, not just
5686 (replace_rhs_if_not_dup): Likewise.
5687 (replace_one_candidate): Likewise.
5689 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5691 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5693 (aarch64_needs_frame_chain): New function.
5694 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5696 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5699 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5700 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5701 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5702 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5703 as true for strict-align.
5704 (aarch64_can_inline_p): Perform checks even when callee has no
5705 attributes to check for strict alignment.
5706 * doc/extend.texi (AArch64 Function Attributes): Document
5708 * doc/invoke.texi: (AArch64 Options): Likewise.
5710 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5712 PR tree-optimization/85853
5713 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5714 the handling of the root of the node to...
5715 (vect_slp_analyze_node_operations_1): ...this new function,
5716 and run the whole thing with the child nodes' def types
5717 set according to their SLP node's def type.
5719 2018-05-23 Richard Biener <rguenther@suse.de>
5722 * tree-data-ref.c (create_runtime_alias_checks): Defer
5723 and ignore overflow warnings.
5725 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5727 PR tree-optimization/85822
5728 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5731 2018-05-23 Richard Biener <rguenther@suse.de>
5733 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5734 memset constants via native_interpret_expr.
5736 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5739 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5741 (cgraph_node::create_alias): Likewise.
5742 (cgraph_node::get_availability): Check ifunc_resolver instead
5743 of looking up ifunc attribute.
5744 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5745 * varasm.c (do_assemble_alias): Likewise.
5746 (assemble_alias): Likewise.
5747 (default_binds_local_p_3): Likewise.
5748 * cgraph.h (cgraph_node): Add ifunc_resolver.
5749 (cgraph_node::only_called_directly_or_aliased_p): Return false
5751 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5753 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5754 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5755 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5756 instead of looking up ifunc attribute.
5758 2018-05-22 Luis Machado <luis.machado@linaro.org>
5760 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5762 2018-05-22 Martin Sebor <msebor@redhat.com>
5765 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5766 only when expasion succeeds.
5767 (expand_builtin_strcmp): Same.
5768 (expand_builtin_strncmp): Same.
5770 2018-05-22 Martin Sebor <msebor@redhat.com>
5772 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5774 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5775 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5777 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5778 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5780 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5781 Add check for legitimate_address.
5782 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5783 (aarch64_swap_ldrstr_operands): New.
5784 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5787 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5788 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5790 * config/aarch64/aarch64.md: New patterns to generate stp
5792 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5793 single words and double words.
5794 (load_pair_sw, load_pair_dw): Likewise.
5795 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5797 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5799 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5800 for different mode ldpstp and add peephole for merged zero stores.
5802 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5804 (aarch64_gen_store_pair): Rename calls to match new patterns.
5805 (aarch64_gen_load_pair): Rename calls to match new patterns.
5806 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5807 (load_pair<DREG:mode><DREG2:mode>): ... This.
5808 (store_pair<mode>): Rename to...
5809 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5810 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5812 (V_INT_EQUIV): Handle SImode.
5813 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5816 2018-05-22 Martin Sebor <msebor@redhat.com>
5819 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5820 or ajust the presumed bound on an operation to avoid unnecessary
5823 2018-05-22 Martin Sebor <msebor@redhat.com>
5825 PR tree-optimization/85826
5826 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5827 assuming that a DECL necesarily has a constant size.
5829 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5832 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5834 2018-05-22 Richard Biener <rguenther@suse.de>
5836 PR tree-optimization/85834
5837 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5838 non-constant and non-zero memset arguments.
5840 2018-05-22 Martin Liska <mliska@suse.cz>
5843 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5845 2018-05-22 Richard Biener <rguenther@suse.de>
5847 PR tree-optimization/85863
5848 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5849 comparisons when vectype is specified.
5850 (vectorizable_condition): Do not specify vectype for
5851 vect_is_simple_cond when SLP vectorizing.
5853 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5856 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5857 define __ibm128 as long double.
5858 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5859 as a distinct type when IEEE 128-bit support is enabled.
5860 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5861 128-bit types to use the correct functions.
5862 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5863 convert between 128-bit floating point types that have different
5864 modes but the same representation, instead of using gen_lowpart to
5866 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5868 (IFKF_reg): New attributes to give the register constraints for
5870 (extend<mode>tf2_internal): New insns to mark an explicit
5871 conversion between 128-bit floating point types that have a
5872 different mode but share the same representation.
5874 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5876 PR tree-optimization/85814
5877 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5878 a null return from get_strinfo when unsharing the next
5879 strinfo in the chain.
5881 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5884 * varasm.c (weak_finish): Clean up weak_decls.
5886 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5888 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5889 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5890 UNSPEC_UADALP values.
5891 * config/aarch64/iterators.md (ABAL): New int iterator.
5894 (sur): Add mappings for the above.
5895 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5897 (aarch64_<sur>abal<mode>_4): Likewise.
5898 (aarch64_<sur>adalp<mode>_3): Likewise.
5899 (<sur>sadv16qi): New define_expand.
5901 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5903 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5904 (*movdf_internal): Ditto.
5905 (*rcpsf2_sse): Ditto.
5906 (*rsqrtsf2_sse): Ditto.
5907 (*sqrt<mode>2_sse): Ditto.
5909 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5911 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5913 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5914 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5915 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5916 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5918 * config/aarch64/arm_neon.h: Likewise.
5919 * config/aarch64/iterators.md (VQ_I): New.
5921 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5923 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5925 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5928 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5930 * config/nds32/constraints.md (S): New constraint.
5931 * config/nds32/nds32.md (call_internal): Use constraint S.
5932 (call_value_internal): Likewise.
5933 (sibcall_internal): Likewise.
5934 (sibcall_value_internal): Likewise.
5936 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5937 Chung-Ju Wu <jasonwucj@gmail.com>
5939 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5942 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5943 Chung-Ju Wu <jasonwucj@gmail.com>
5945 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5946 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5947 (nds32_rtx_costs_impl): Simplify.
5948 (nds32_address_cost_impl): Simplify.
5949 (nds32_init_rtx_costs): New function.
5950 (nds32_rtx_costs_speed_prefer): Likewise.
5951 (nds32_rtx_costs_size_prefer): Likewise.
5952 (nds32_address_cost_speed_prefer): Likewise.
5953 (nds32_address_cost_speed_fwprop): Likewise.
5954 (nds32_address_cost_size_prefer): Likewise.
5955 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5956 * config/nds32/nds32.c (nds32_option_override): Use
5957 nds32_init_rtx_costs function.
5959 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5961 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5962 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5963 (TARGET_PIPELINE_N8): Likewise.
5964 (TARGET_PIPELINE_N10): Likewise.
5965 (TARGET_PIPELINE_N13): Likewise.
5966 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5968 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5970 * config/nds32/nds32-fpu.md: Update copyright year.
5972 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5974 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5976 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5978 * config/nds32/nds32.c
5979 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5980 * config/nds32/nds32.opt (minline-asm-r15): New option.
5982 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5984 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5986 * config/nds32/nds32.md (abssi2): New pattern.
5988 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5990 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5991 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5992 Merge insn pattern from sse_cvtsi2ss<round_name> and
5993 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5994 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5995 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5996 using SWI48 mode iterator.
5997 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5998 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5999 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6000 pattern from sse_cvttss2si<round_saeonly_name>
6001 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6002 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6003 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6004 using SWI48 mode iterator.
6005 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6006 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6007 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6008 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6009 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6010 using SWI48 mode iterator.
6011 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6012 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6013 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6014 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6015 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6016 SWI48 mode iterator.
6017 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6018 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6019 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6020 pattern from sse_cvttsd2si<round_saeonly_name>
6021 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6023 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6025 * config/nds32/nds32-md-auxiliary.c
6026 (nds32_valid_smw_lwm_base_p): Refine.
6027 (nds32_output_smw_single_word): Refine.
6028 (nds32_output_smw_double_word): New.
6029 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6031 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6033 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6034 (nds32_output_stack_pop): Refine.
6035 (nds32_expand_unaligned_load): Refine.
6036 (nds32_expand_unaligned_store): Refine.
6038 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6039 Chung-Ju Wu <jasonwucj@gmail.com>
6041 * config/nds32/constants.md: Add TP_REGNUM constant.
6042 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6043 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6045 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6046 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6047 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6048 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6049 and PIC code generation.
6050 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6052 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6054 * config/nds32/nds32.md: Support TLS and PIC.
6055 * config/nds32/nds32.c: Support TLS and PIC.
6056 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6057 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6060 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6062 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6063 mode with E_ prefix.
6065 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6066 Chung-Ju Wu <jasonwucj@gmail.com>
6068 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6069 * config/nds32/nds32-md-auxiliary.c
6070 (symbolic_reference_mentioned_p): New.
6071 (nds32_legitimize_ict_address): New.
6072 (nds32_expand_ict_move): New.
6073 (nds32_indirect_call_referenced_p): New.
6074 (nds32_symbol_binds_local_p): Delete.
6075 (nds32_long_call_p): Modify.
6076 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6077 * config/nds32/nds32-protos.h
6078 (symbolic_reference_mentioned_p): Declare.
6079 (nds32_legitimize_ict_address): Declare.
6080 (nds32_expand_ict_move): Declare.
6081 (nds32_indirect_call_referenced_p): Declare.
6082 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6083 (nds32_relax_group): Use nds32_ict_const_p as condition.
6084 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6085 (nds32_asm_file_start): Output ict_model directive in asm code.
6086 (nds32_legitimate_address_p): Consider indirect call.
6087 (nds32_print_operand): Consider indirect call.
6088 (nds32_print_operand_address): Consider indirect call.
6089 (nds32_insert_attributes): Handle "indirect_call" attribute.
6090 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6091 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6092 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6093 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6094 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6095 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6096 (TARGET_ICT_MODEL_SMALL): Define.
6097 (TARGET_ICT_MODEL_LARGE): Define.
6098 * config/nds32/nds32.md (movsi): Consider ict model.
6099 (call, call_value): Consider ict model.
6100 (sibcall, sibcall_value): Consider ict model.
6101 * config/nds32/nds32.opt (mict-model): New option.
6102 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6105 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6106 Monk Chiang <sh.chiang04@gmail.com>
6107 Jim Wilson <jimw@sifive.com>
6109 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6110 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6111 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6112 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6113 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6114 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6115 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6116 compute save_libcall_adjustment properly.
6117 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6118 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6119 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6120 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6121 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6122 (ABI_SPEC): Handle mabi=ilp32e.
6123 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6124 (RVE): Add RVE mask.
6125 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6126 <-march>: Add rv32e as an example.
6128 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6131 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6132 (intra_create_variable_infos): Handle C++ constructors.
6134 2018-05-18 Martin Liska <mliska@suse.cz>
6136 * passes.def: Remove a redundant pass.
6138 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6141 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6143 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6145 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6147 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6149 (armv2, armv2a, armv3, armv3m): Delete architectures.
6150 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6151 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6152 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6154 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6155 (*mulsidi3adddi): Likewise.
6156 (mulsidi3): Likewise.
6157 (*mulsidi3_nov6): Likewise.
6158 (umulsidi3): Likewise.
6159 (umulsidi3_nov6): Likewise.
6160 (umaddsidi4): Likewise.
6161 (*umulsidi3adddi): Likewise.
6162 (smulsi3_highpart): Likewise.
6163 (*smulsi3_highpart_nov6): Likewise.
6164 (umulsi3_highpart): Likewise.
6165 (*umulsi3_highpart_nov6): Likewise.
6166 * config/arm/arm.h (arm_arch3m): Delete.
6167 * config/arm/arm.c (arm_arch3m): Delete.
6168 (arm_option_override_internal): Update armv3-related comment.
6169 (arm_configure_build_target): Delete use of isa_bit_mode32.
6170 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6171 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6172 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6174 (mulusa3): Likewise.
6175 * config/arm/arm-protos.h (arm_arch3m): Delete.
6176 * config/arm/arm-tables.opt: Regenerate.
6177 * config/arm/arm-tune.md: Likewise.
6178 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6179 deleted architectures.
6181 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6183 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6184 (armv5t, armv5te): New features.
6185 (ARMv5, ARMv5e): Delete fgroups.
6186 (ARMv5t, ARMv5te): Adjust for above changes.
6188 (armv5, armv5e): Delete arches.
6189 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6191 (*call_reg_arm): Likewise.
6192 (*call_value_reg_armv5): Likewise.
6193 (*call_value_reg_arm): Likewise.
6194 (*call_symbol): Likewise.
6195 (*call_value_symbol): Likewise.
6196 (*sibcall_insn): Likewise.
6197 (*sibcall_value_insn): Likewise.
6199 (prefetch): Likewise.
6200 (define_split and define_peephole2 dependent on arm_arch5):
6202 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6204 (TARGET_ARM_QBIT): Likewise.
6205 (TARGET_DSP_MULTIPLY): Likewise.
6206 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6207 (arm_arch5, arm_arch5e): Delete.
6208 (arm_arch5t, arm_arch5te): Declare.
6209 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6210 (arm_arch5t): Declare.
6211 (arm_option_reconfigure_globals): Update for the above.
6212 (arm_options_perform_arch_sanity_checks): Update comment, replace
6213 use of arm_arch5 with arm_arch5t.
6214 (use_return_insn): Likewise.
6215 (arm_emit_call_insn): Likewise.
6216 (output_return_instruction): Likewise.
6217 (arm_final_prescan_insn): Likewise.
6218 (arm_coproc_builtin_available): Likewise.
6219 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6220 arm_arch5e with arm_arch5t and arm_arch5te.
6221 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6222 (arm_arch5t, arm_arch5te): Declare.
6223 * config/arm/arm-tables.opt: Regenerate.
6224 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6225 * config/arm/t-multilib: Likewise.
6226 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6227 instead of arm_arch5.
6228 (*call_reg_thumb1): Likewise.
6229 (*call_value_reg_thumb1_v5): Likewise.
6230 (*call_value_reg_thumb1): Likewise.
6231 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6233 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6235 2018-05-18 Martin Liska <mliska@suse.cz>
6237 PR gcov-profile/84846
6238 * doc/gcov.texi: Document -t option of gcov tool.
6240 2018-05-18 Martin Liska <mliska@suse.cz>
6242 PR gcov-profile/84846
6243 * gcov.c (print_usage): Add new -t option.
6244 (process_args): Handle the option.
6245 (generate_results): Use stdout as output when requested by
6248 2018-05-18 Martin Liska <mliska@suse.cz>
6250 PR gcov-profile/84846
6251 * coverage.c (coverage_init): Write PWD to .gcno file.
6252 * doc/gcov.texi: Document how working directory is printed.
6253 * gcov-dump.c (dump_gcov_file): Print PWD.
6254 * gcov.c (output_intermediate_file): Likewise.
6255 (read_graph_file): Read PWD string.
6256 (output_lines): Print PWD.
6258 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6261 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6262 for retval and return false if all args to phi are zero.
6264 2018-05-18 Richard Biener <rguenther@suse.de>
6266 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6268 (evrp_dom_walker::before_dom_children): Call it.
6270 2018-05-18 Richard Biener <rguenther@suse.de>
6272 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6273 results when processing array refs with variable index.
6275 2018-05-18 Toon Moene <toon@moene.org>
6277 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6278 directly after that of -floop-interchange. Indicate that both
6279 options are enabled by default when specifying -O3.
6281 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6283 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6284 iterator. Delete separate integer-mode vec_set<mode> expander.
6285 (aarch64_simd_vec_setv2di): Delete.
6286 (vec_setv2di): Delete.
6287 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6288 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6289 the "w, r" alternative.
6291 2018-05-18 Martin Liska <mliska@suse.cz>
6293 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6294 * tree-pass.h (make_pass_lower_switch_O0): New function.
6295 * tree-switch-conversion.c (node_has_low_bound): Remove.
6296 (node_has_high_bound): Likewise.
6297 (node_is_bounded): Likewise.
6298 (class pass_lower_switch): Make it a template type and create
6300 (pass_lower_switch::execute): Add template argument.
6301 (make_pass_lower_switch): New function.
6302 (make_pass_lower_switch_O0): New function.
6303 (do_jump_if_equal): Remove.
6304 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6305 all the hard work to tree optimization passes.
6307 2018-05-18 Martin Liska <mliska@suse.cz>
6309 * dbgcnt.c (limit_low): Renamed from limit.
6310 (limit_high): New variable.
6311 (dbg_cnt_is_enabled): Check for upper limit.
6312 (dbg_cnt): Adjust dumping.
6313 (dbg_cnt_set_limit_by_index): Add new argument for high
6315 (dbg_cnt_set_limit_by_name): Likewise.
6316 (dbg_cnt_process_single_pair): Parse new format.
6317 (dbg_cnt_process_opt): Use strtok.
6318 (dbg_cnt_list_all_counters): Remove 'value' and add
6320 * doc/invoke.texi: Document changes.
6322 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6324 * doc/sourcebuild.texi (scalar_all_fma): Document.
6325 * tree.def (FMA_EXPR): Delete.
6326 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6327 * internal-fn.c (ternary_direct): New macro.
6328 (expand_ternary_optab_fn): Likewise.
6329 (direct_ternary_optab_supported_p): Likewise.
6330 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6331 * builtins.c (fold_builtin_fma): Delete.
6332 (fold_builtin_3): Don't call it.
6333 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6334 * expr.c (expand_expr_real_2): Likewise.
6335 * fold-const.c (operand_equal_p): Likewise.
6336 (fold_ternary_loc): Likewise.
6337 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6338 * gimple.c (DEFTREECODE): Likewise.
6339 * gimplify.c (gimplify_expr): Likewise.
6340 * optabs-tree.c (optab_for_tree_code): Likewise.
6341 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6342 * tree-eh.c (operation_could_trap_p): Likewise.
6343 (stmt_could_throw_1_p): Likewise.
6344 * tree-inline.c (estimate_operator_cost): Likewise.
6345 * tree-pretty-print.c (dump_generic_node): Likewise.
6346 (op_code_prio): Likewise.
6347 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6348 * tree-ssa-operands.c (get_expr_operands): Likewise.
6349 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6350 * fold-const-call.h (fold_fma): Delete.
6351 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6352 CFN_FNMA and CFN_FNMS.
6354 * genmatch.c (combined_fn): New enum.
6355 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6356 (commutative_op): New function.
6357 (commutate): Use it. Handle more than 2 operands.
6358 (dt_operand::gen_gimple_expr): Use commutative_op.
6359 (parser::parse_expr): Allow :c to be used with non-binary
6360 operators if the commutative operand is known.
6361 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6362 CFN_FMS, CFN_FNMA and CFN_FNMS.
6363 (backprop::process_assign_use): Remove FMA_EXPR handling.
6364 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6365 (gen_hsa_fma): New function.
6366 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6367 IFN_FNMA and IFN_FNMS.
6368 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6369 * gimple-fold.h (follow_all_ssa_edges): Declare.
6370 * gimple-fold.c (follow_all_ssa_edges): New function.
6371 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6372 gimple_build interface and use follow_all_ssa_edges to fold the result.
6373 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6374 instead of checking for optabs directly.
6375 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6376 rather than FMA_EXPRs.
6377 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6378 call to IFN_FMA instead of an FMA_EXPR.
6380 2018-05-17 Jim Wilson <jimw@sifive.com>
6382 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6383 sign extended promoted subreg, and the range does not have the sign bit
6384 set, then do a sign extend.
6386 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6387 test, check for sign extended subreg and/or constant operands, and
6388 do a sign extend in that case.
6390 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6392 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6393 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6395 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6396 Change logics_shift_reg to logics_shift_imm.
6397 (thunderx2t99_fp_loadpair_basic): Delete.
6398 (thunderx2t99_fp_storepair_basic): Delete.
6399 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6400 (thunderx2t99_asimd_polynomial): Delete.
6401 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6402 and neon_fp_mul_d_scalar_q.
6403 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6404 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6405 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6406 (thunderx2t99_asimd_lut): Add missing tbl types.
6407 (thunderx2t99_asimd_ext): Delete.
6408 (thunderx2t99_asimd_load1_1_mult): Delete.
6409 (thunderx2t99_asimd_load1_2_mult): Delete.
6410 (thunderx2t99_asimd_load1_ldp): New.
6411 (thunderx2t99_asimd_load1): New.
6412 (thunderx2t99_asimd_load2): Add missing *load2* types.
6413 (thunderx2t99_asimd_load3): New.
6414 (thunderx2t99_asimd_load4): New.
6415 (thunderx2t99_asimd_store1_1_mult): Delete.
6416 (thunderx2t99_asimd_store1_2_mult): Delete.
6417 (thunderx2t99_asimd_store2_mult): Delete.
6418 (thunderx2t99_asimd_store2_onelane): Delete.
6419 (thunderx2t99_asimd_store_stp): New.
6420 (thunderx2t99_asimd_store1): New.
6421 (thunderx2t99_asimd_store2): New.
6422 (thunderx2t99_asimd_store3): New.
6423 (thunderx2t99_asimd_store4): New.
6425 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6427 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6428 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6430 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6431 Segher Boessenkool <segher@kernel.crashing.org>
6434 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6437 2018-05-17 Richard Biener <rguenther@suse.de>
6439 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6440 for pruning loop and prune defs feeding only already visited PHIs.
6442 2018-05-17 Richard Biener <rguenther@suse.de>
6444 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6446 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6447 Richard Biener <rguenther@suse.de>
6449 PR tree-optimization/85793
6450 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6451 for VMAT_ELEMENTWISE.
6453 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6455 * internal-fn.h (lookup_internal_fn): Declare
6456 * internal-fn.c (lookup_internal_fn): New function.
6457 * gimple.c (gimple_build_call_from_tree): Handle calls to
6459 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6460 internal function names.
6461 * tree-pretty-print.c (dump_generic_node): Likewise.
6462 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6464 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6466 * gimple-fold.h (gimple_build): Make the function forms take
6467 combined_fn rather than built_in_function.
6468 (gimple_simplify): Likewise.
6469 * gimple-match-head.c (gimple_simplify): Likewise.
6470 * gimple-fold.c (gimple_build): Likewise.
6471 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6472 rather than gimple_build_call_internal.
6473 (get_initial_defs_for_reduction): Likewise.
6474 (vect_create_epilog_for_reduction): Likewise.
6475 (vectorizable_live_operation): Likewise.
6477 2018-05-17 Martin Liska <mliska@suse.cz>
6479 * gimple-ssa-sprintf.c (format_directive): Do not use
6480 space in between 'G_' and '('.
6482 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6485 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6486 even if the mask is not all ones.
6489 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6491 (ix86_gimple_fold_builtin): Likewise.
6494 * config/i386/i386.c: Include tree-vector-builder.h.
6495 (ix86_vector_shift_count): New function.
6496 (ix86_fold_builtin): Fold shift builtins by scalar count.
6497 (ix86_gimple_fold_builtin): Likewise.
6499 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6500 _mm512_setzero): New intrinsics.
6502 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6503 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6505 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6506 code generation for cases where splatting a value is not useful.
6507 * simplify-rtx.c (simplify_ternary_operation): Simplify
6508 vec_merge across a vec_duplicate and a paradoxical subreg forming
6509 a vector mode to a vec_concat.
6511 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6513 * config.gcc: Support "goldmont-plus".
6514 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6516 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6517 PROCESSOR_GOLDMONT_PLUS.
6518 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6519 (processor_target_table): Add "goldmont-plus".
6520 (PTA_GOLDMONT_PLUS): Define.
6521 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6522 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6523 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6524 (fold_builtin_cpu): Add "goldmont-plus".
6525 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6526 (ix86_option_override_internal): Add "goldmont-plus".
6527 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6528 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6529 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6530 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6532 2018-05-17 Richard Biener <rguenther@suse.de>
6534 PR tree-optimization/85757
6535 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6536 remove defs that only feed that PHI from further processing.
6538 2018-05-16 Jim Wilson <jimw@sifive.com>
6540 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6542 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6544 2018-05-16 Mark Wielaard <mark@klomp.org>
6546 * dwarf2out.c (count_index_strings): New function.
6547 (output_indirect_strings): Call count_index_strings and generate
6548 header for dwarf_version >= 5.
6550 2018-05-16 Mark Wielaard <mark@klomp.org>
6552 * dwarf2out.c (dwarf_FORM): New function.
6553 (set_indirect_string): Use dwarf_FORM.
6554 (reset_indirect_string): Likewise.
6555 (size_of_die): Likewise.
6556 (value_format): Likewise.
6557 (output_die): Likewise.
6558 (add_skeleton_AT_string): Likewise.
6559 (output_macinfo_op): Likewise.
6560 (index_string): Likewise.
6561 (output_index_string_offset): Likewise.
6562 (output_index_string): Likewise.
6563 (count_index_strings): Likewise.
6565 2018-05-16 Carl Love <cel@us.ibm.com>
6567 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6568 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6570 2018-05-16 Martin Jambor <mjambor@suse.cz>
6572 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6573 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6575 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6577 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6578 (fnma<mode>4): Likewise.
6579 (fms<mode>4): Likewise.
6580 (fnms<mode>4): Likewise.
6581 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6582 (aarch64_fnma<mode>4): Likewise.
6583 (aarch64_fms<mode>4): Likewise.
6584 (aarch64_fnms<mode>4): Likewise.
6585 (aarch64_fnmadd<mode>4): Likewise.
6587 2018-05-16 Jason Merrill <jason@redhat.com>
6589 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6591 2018-05-16 Richard Biener <rguenther@suse.de>
6593 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6594 (dump_stmt_cost): Declare.
6595 (add_stmt_cost): Dump cost we add.
6596 (add_stmt_costs): New function.
6597 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6599 (vect_analyze_stmt): Adjust prototype.
6600 (vectorizable_condition): Likewise.
6601 (vectorizable_live_operation): Likewise.
6602 (vectorizable_reduction): Likewise.
6603 (vectorizable_induction): Likewise.
6604 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6605 cost vector to pass to vectorizable_ and record afterwards.
6606 (vect_model_reduction_cost): Take cost vector argument and adjust.
6607 (vect_model_induction_cost): Likewise.
6608 (vectorizable_reduction): Likewise.
6609 (vectorizable_induction): Likewise.
6610 (vectorizable_live_operation): Likewise.
6611 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6612 SLP_TREE_NUMBER_OF_VEC_STMTS.
6613 (vect_analyze_slp_cost_1): Remove.
6614 (vect_analyze_slp_cost): Likewise.
6615 (vect_slp_analyze_node_operations): Take visited args and
6616 a target cost vector. Avoid processing already visited stmt sets.
6617 (vect_slp_analyze_operations): Use a local cost vector to gather
6618 costs and register those of non-discarded instances.
6619 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6620 (vect_schedule_slp_instance): Remove copying of
6621 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6623 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6624 adding cost. Record cost entry location.
6625 (vect_prologue_cost_for_slp_op): Function to compute cost of
6626 a constant or invariant generated for SLP vect in the prologue,
6627 split out from vect_analyze_slp_cost_1.
6628 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6629 (vect_model_promotion_demotion_cost): Likewise.
6630 (vect_model_store_cost): Likewise, make static.
6631 (vect_model_load_cost): Likewise.
6632 (vectorizable_bswap): Add cost vector arg and adjust.
6633 (vectorizable_call): Likewise.
6634 (vectorizable_simd_clone_call): Likewise.
6635 (vectorizable_conversion): Likewise.
6636 (vectorizable_assignment): Likewise.
6637 (vectorizable_shift): Likewise.
6638 (vectorizable_operation): Likewise.
6639 (vectorizable_store): Likewise.
6640 (vectorizable_load): Likewise.
6641 (vectorizable_condition): Likewise.
6642 (vectorizable_comparison): Likewise.
6643 (can_vectorize_live_stmts): Likewise.
6644 (vect_analyze_stmt): Likewise.
6645 (vect_transform_stmt): Adjust calls to vectorizable_*.
6646 * tree-vectorizer.c: Include gimple-pretty-print.h.
6647 (dump_stmt_cost): New function.
6649 2018-05-16 Richard Biener <rguenther@suse.de>
6651 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6652 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6653 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6654 (check_name): New callback.
6655 (dse_classify_store): Track cycles via a visited bitmap of PHI
6656 defs and simplify handling of in-loop and across loop dead stores
6657 and properly fail for loop-variant refs. Handle byte-tracking with
6658 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6661 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6663 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6664 (vect_get_mask_type_for_stmt): Likewise.
6665 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6667 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6668 to determine the statement's vector type and the vector type that
6669 should be used for calculating nunits. Deal with cases in which
6670 the type has to be deferred.
6671 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6672 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6673 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6674 (vect_determine_vf_for_stmt): New functions, split out from...
6675 (vect_determine_vectorization_factor): ...here.
6676 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6677 (vect_get_mask_type_for_stmt): New functions, split out from
6678 vect_determine_vectorization_factor.
6680 2018-05-16 Richard Biener <rguenther@suse.de>
6682 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6683 verify the [VEC_]COND_EXPR embedded comparison.
6685 2018-05-15 Martin Sebor <msebor@redhat.com>
6687 PR tree-optimization/85753
6688 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6689 RECORD_TYPE in addition to ARRAY_TYPE.
6691 2018-05-15 Martin Sebor <msebor@redhat.com>
6694 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6696 2018-05-15 Richard Biener <rguenther@suse.de>
6698 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6699 add by_clobber_p one. Change algorithm to collect all defs
6700 representing uses we need to walk and try reducing them to
6701 a single one before failing.
6702 (dse_dom_walker::dse_optimize_stmt): Adjust.
6704 2018-05-13 Mark Wielaard <mark@klomp.org>
6706 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6707 (size_of_loc_descr): Likewise.
6708 (output_loc_operands): Likewise.
6709 (output_loc_operands_raw): Likewise.
6710 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6711 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6712 (hash_loc_operands): Likewise.
6713 (compare_loc_operands): Likewise.
6715 2018-05-14 Mark Wielaard <mark@klomp.org>
6717 * dwarf2out.c (count_index_addrs): New function.
6718 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6720 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6722 PR tree-optimization/83648
6723 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6724 return value as malloc candidate.
6726 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6729 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6730 param as true in call to suggest_attribute.
6732 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6734 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6735 -mreadonly-in-sdata.
6737 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6739 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6741 (aarch64_crypto_aesd_fused): Likewise.
6743 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6745 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6746 (movsi_aarch64): Likewise.
6747 (load_pairsi): Likewise.
6748 (load_pairdi): Likewise.
6749 (store_pairsi): Likewise.
6750 (store_pairdi): Likewise.
6751 (load_pairsf): Likewise.
6752 (load_pairdf): Likewise.
6753 (store_pairsf): Likewise.
6754 (store_pairdf): Likewise.
6755 (zero_extend): Likewise.
6756 (trunc): Swap alternatives.
6757 (fcvt_target): Add '?' to prefer w over r.
6759 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6762 * config/i386/i386.md: Disallow non-commutative arithmetics in
6763 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6764 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6765 in the peephole2 before it.
6767 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6769 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6770 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6771 (ix86_handle_option): Handle -mcldemote.
6772 * config.gcc: New header.
6773 * config/i386/cldemoteintrin.h: New file.
6774 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6775 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6777 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6778 OPTION_MASK_ISA_CLDEMOTE.
6779 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6780 (ix86_valid_target_attribute_inner_p): Ditto.
6781 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6782 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6783 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6784 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6785 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6787 * config/i386/i386.opt: Add -mcldemote.
6788 * config/i386/x86intrin.h: New header.
6789 * doc/invoke.texi: Add -mcldemote.
6791 2018-05-14 Richard Biener <rguenther@suse.de>
6793 * doc/match-and-simplify.texi: Adjust :s documentation.
6795 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6797 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6798 intended memcpy size.
6799 (REORDER_45): Likewise.
6801 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6803 * sort.cc: New file.
6804 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6805 * vec.c (qsort_chk): Use gcc_qsort.
6806 * Makefile.in (OBJS-libcommon): Add sort.o.
6807 (build/sort.o): New target. Use it...
6808 (BUILD_RTL): ... here, and...
6809 (build/gencfn-macros): ... here, and...
6810 (build/genmatch): ... here.
6812 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6813 Chung-Ju Wu <jasonwucj@gmail.com>
6815 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6816 * config/nds32/nds32-graywolf.md: New file.
6817 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6818 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6820 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6821 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6822 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6823 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6824 * config/nds32/pipelines.md: Include n15 settings.
6826 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6827 Chung-Ju Wu <jasonwucj@gmail.com>
6829 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6830 * config/nds32/nds32-n13.md: New file.
6831 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6832 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6834 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6835 * config/nds32/nds32.md (pipeline_model): Add n13.
6836 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6837 * config/nds32/pipelines.md: Include n13 settings.
6839 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6840 Chung-Ju Wu <jasonwucj@gmail.com>
6842 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6843 * config/nds32/nds32-n10.md: New file.
6844 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6845 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6847 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6848 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6849 * config/nds32/nds32.md (pipeline_model): Add n10.
6850 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6851 * config/nds32/pipelines.md: Include n10 settings.
6853 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6854 Kito Cheng <kito.cheng@gmail.com>
6855 Chung-Ju Wu <jasonwucj@gmail.com>
6857 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6858 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6859 Add enum values for DSP extension instructions.
6860 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6862 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6863 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6865 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6866 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6867 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6868 * config/nds32/nds32-intrinsic.md: Likewise.
6869 * config/nds32/nds32_intrinsic.h: Likewise.
6870 * config/nds32/nds32-md-auxiliary.c: Likewise.
6871 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6872 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6873 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6874 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6875 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6876 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6877 TYPE_DMAC in switch statement.
6878 * config/nds32/nds32.c: New checking and implementation for DSP
6879 extension instructions.
6880 * config/nds32/nds32.h: Likewise.
6881 * config/nds32/nds32.md: Likewise.
6882 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6883 * config/nds32/predicates.md: Implement new predicates for DSP
6886 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6888 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6889 Reformat alternatives and attributes so it is easier to identify
6890 which constraints/attributes go with which instruction.
6891 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6892 (mov<mode>_softfloat32, FMOVE64): Likewise.
6893 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6894 (mov<mode>_softfloat64, FMOVE64): Likewise.
6896 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6898 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6900 (Basic PowerPC Built-in Functions): The new name of the
6901 subsection previously known as "PowerPC Built-in Functions".
6902 (Basic PowerPC Built-in Functions Available on all Configurations):
6904 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6905 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6906 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6907 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6909 2018-05-11 Martin Jambor <mjambor@suse.cz>
6912 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6915 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6918 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6920 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6922 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6923 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6924 (ix86_handle_option): Handle -mwaitpkg.
6925 * config.gcc: New header.
6926 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6927 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6928 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6930 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6931 OPTION_MASK_ISA_WAITPKG.
6932 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6933 (ix86_option_override_internal): Add PTA_WAITPKG.
6934 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6935 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6936 IX86_BUILTIN_TPAUSE.
6937 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6938 __builtin_ia32_umwait and __builtin_ia32_tpause.
6939 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6940 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6941 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6942 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6943 UNSPECV_TPAUSE): New.
6944 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6945 * config/i386/i386.opt: Add -mwaitpkg.
6946 * config/i386/waitpkgintrin.h: New file.
6947 * config/i386/x86intrin.h: New header.
6948 * doc/invoke.texi: Add -mwaitpkg.
6950 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6953 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6955 (cortex-m0): Use armv6s-m isa.
6956 (cortex-m0plus): Likewise.
6957 (cortex-m1): Likewise.
6958 (cortex-m0.small-multiply): Likewise.
6959 (cortex-m0plus.small-multiply): Likewise.
6960 (cortex-m1.small-multiply): Likewise.
6962 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6963 Jakub Jelinek <jakub@redhat.com>
6965 PR tree-optimization/85692
6966 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6967 source permute as well.
6969 2018-05-11 Martin Liska <mliska@suse.cz>
6972 * doc/extend.texi: Document LLVM style format for no_sanitize
6975 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6977 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6978 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6979 (mode_supports_vsx_dform_quad): Likewise.
6980 (mode_supports_vmx_dform): Move these functions to be next to the
6981 other mode_supports functions.
6982 (mode_supports_dq_form): Likewise.
6983 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6984 mode_supports_dq_form.
6985 (reg_offset_addressing_ok_p): Likewise.
6986 (offsettable_ok_by_alignment): Likewise.
6987 (rs6000_legitimate_offset_address_p): Likewise.
6988 (legitimate_lo_sum_address_p): Likewise.
6989 (rs6000_legitimize_address): Likewise.
6990 (rs6000_legitimize_reload_address): Likewise.
6991 (rs6000_secondary_reload_inner): Likewise.
6992 (rs6000_preferred_reload_class): Likewise.
6993 (rs6000_output_move_128bit): Likewise.
6995 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6997 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6998 Generate SImode target register for null target.
6999 <case IX86_BUILTIN_XGETBV>: Ditto.
7000 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7001 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7003 2018-05-10 Carl Love <cel@us.ibm.com>
7005 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7006 dcbtt and dcbtstt if operands[2] is 0.
7008 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7011 * config/i386/sse.md (usadv64qi): New expander.
7013 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7015 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7016 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7017 -maltivec=be support.
7018 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7019 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7020 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7021 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7022 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7023 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7024 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7025 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7026 altivec_vsumsws): Adjust.
7027 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7028 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7029 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7031 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7032 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7033 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7034 (altivec_lve<VI_char>x): Delete expand.
7035 (*altivec_lve<VI_char>x_internal): Rename to...
7036 (altivec_lve<VI_char>x): ... this.
7037 (altivec_lvxl_<mode>): Delete expand.
7038 (*altivec_lvxl_<mode>_internal): Rename to ...
7039 (altivec_lvxl_<mode>): ... this.
7040 (altivec_stvxl_<mode>): Delete expand.
7041 (*altivec_stvxl_<mode>_internal): Rename to ...
7042 (altivec_stvxl_<mode>): ... this.
7043 (altivec_stve<VI_char>x): Delete expand.
7044 (*altivec_stve<VI_char>x_internal): Rename to ...
7045 (altivec_stve<VI_char>x): ... this.
7046 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7047 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7048 reduc_plus_scal_<mode>): Adjust.
7049 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7051 (rs6000_cpu_cpp_builtins): Adjust.
7052 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7053 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7054 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7055 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7056 -maltivec=be support.
7057 (rs6000_split_vec_extract_var): Adjust.
7058 (rs6000_split_v4si_init): Adjust.
7059 (swap_selector_for_mode): Delete.
7060 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7061 altivec_expand_stvex_be): Delete.
7062 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7063 -maltivec=be support.
7064 (rs6000_gimple_fold_builtin): Ditto.
7065 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7067 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7068 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7069 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7070 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7071 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7072 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7073 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7074 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7075 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7076 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7077 anonymous split): Adjust.
7078 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7079 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7081 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7083 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7084 when --with-gxx-include-dir is also specified.
7085 * configure: Regenerate.
7087 2018-05-09 Jim Wilson <jimw@sifive.com>
7090 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7091 * config/riscv/t-withmultilib: New.
7092 * config/riscv/withmultilib.h: New.
7093 * doc/install.texi: Document RISC-V --with-multilib-list support.
7095 2018-05-09 Richard Biener <rguenther@suse.de>
7097 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7099 (vect_bb_vectorization_profitable_p): Adjust. Compute
7100 actual scalar cost using the cost vector and the add_stmt_cost
7103 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7105 PR rtl-optimization/85645
7106 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7107 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7109 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7111 PR rtl-optimization/85645
7112 * shrink-wrap.c (spread_components): Return a boolean saying if
7113 anything was changed.
7114 (try_shrink_wrapping_separate): Iterate spread_components until
7115 nothing changes anymore.
7117 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7119 PR rtl-optimization/85645
7120 * regrename.c (build_def_use): Also kill the chains that include the
7121 destination of a REG_CFA_REGISTER note.
7123 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7125 PR rtl-optimization/85645
7126 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7127 insn that has a REG_CFA_REGISTER note.
7129 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7131 * cfgexpand.c (expand_clobber): New function.
7132 (expand_gimple_stmt_1): Use it.
7133 * tree-vect-stmts.c (vect_clobber_variable): New function,
7135 (vectorizable_simd_clone_call): ...here.
7136 (vectorizable_store): Emit a clobber either side of an
7137 IFN_STORE_LANES sequence.
7138 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7140 2018-05-09 Tom de Vries <tom@codesourcery.com>
7143 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7144 (define_insn "trap_if_false"): Add exit after trap.
7146 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7148 PR rtl-optimization/85638
7149 * bb-reorder.c: Include common/common-target.h.
7150 (create_forwarder_block): New function extracted from...
7151 (fix_up_crossing_landing_pad): ...here. Rename into...
7152 (dw2_fix_up_crossing_landing_pad): ...this.
7153 (sjlj_fix_up_crossing_landing_pad): New function.
7154 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7155 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7156 from both partitions and exit the loop after one iteration.
7158 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7161 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7163 (Basic PowerPC Built-in Functions): The new name of the
7164 subsection previously known as "PowerPC Built-in Functions".
7165 (Basic PowerPC Built-in Functions Available on all Configurations):
7167 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7169 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7171 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7173 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7176 2018-05-08 Jim Wilson <jimw@sifive.com>
7178 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7179 (LD_EMUL_SUFFIX): New.
7180 (LINK_SPEC): Use it.
7182 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7184 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7186 (Basic PowerPC Built-in Functions): The new name of the
7187 subsection previously known as "PowerPC Built-in Functions".
7188 (Basic PowerPC Built-in Functions Available on all Configurations):
7190 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7192 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7194 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7196 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7199 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7202 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7203 after cmpelim optimization.
7205 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7207 * config.gcc: Support "goldmont".
7208 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7209 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7211 * config/i386/i386.c (m_GOLDMONT): Define.
7212 (processor_target_table): Add "goldmont".
7213 (PTA_GOLDMONT): Define.
7214 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7215 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7216 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7217 (fold_builtin_cpu): Add "goldmont".
7218 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7219 (ix86_option_override_internal): Add "goldmont".
7220 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7221 (processor_type): Add PROCESSOR_GOLDMONT.
7222 * config/i386/i386.md: Add CPU "glm".
7223 * config/i386/glm.md: New file.
7224 * config/i386/x86-tune.def: Add m_GOLDMONT.
7225 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7227 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7230 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7232 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7233 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7234 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7237 * config/i386/i386.c (ix86_fold_builtin): Handle
7238 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7241 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7242 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7244 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7247 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7248 (check_arch): Likewise.
7249 (check_fpu): Return the result rather than printing it.
7250 (end arch): Fix operator precedence.
7251 (end cpu): Likewise.
7252 (END): Print the result from check_fpu.
7254 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7255 Alan Hayward <alan.hayward@arm.com>
7256 David Sherwood <david.sherwood@arm.com>
7258 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7259 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7260 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7261 (*fcmuo<mode>_and): New patterns.
7263 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7265 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7266 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7267 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7268 (cmp_op, sve_imm_con): New code attributes.
7269 (SVE_COND_INT_CMP, imm_con): Delete.
7270 (cmp_op): Remove above unspecs from int attribute.
7271 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7273 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7274 comparison-specific unspecs.
7275 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7276 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7277 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7278 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7279 (*vec_fcm<cmp_op><mode>): Rename to...
7280 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7281 (*vec_fcmuo<mode>): Rename to...
7282 (*fcmuo<mode>): ...this and adjust likewise.
7283 (*pred_fcm<cmp_op><mode>): New pattern.
7284 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7285 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7287 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7289 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7290 (aarch64_emit_sve_predicated_cond): New function.
7291 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7292 (aarch64_emit_unspec_cond_or): Replace with...
7293 (aarch64_emit_sve_or_conds): ...this new function. Use
7294 aarch64_emit_sve_ptrue_op for the individual comparisons and
7295 aarch64_emit_binop to OR them together.
7296 (aarch64_emit_inverted_unspec_cond): Replace with...
7297 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7298 aarch64_emit_sve_ptrue_op for the comparison and
7299 aarch64_emit_unop to invert the result.
7300 (aarch64_expand_sve_vec_cmp_float): Update after the above
7301 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7303 2018-05-07 Nathan Sidwell <nathan@acm.org>
7305 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7306 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7307 (Backwards Compatibility): Likewise.
7309 2018-05-07 Luis Machado <luis.machado@linaro.org>
7313 2018-05-07 Luis Machado <luis.machado@linaro.org>
7315 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7316 <prefetch_dynamic_strides>: New const bool field.
7317 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7318 prefetch_dynamic_strides.
7319 (exynosm1_prefetch_tune): Likewise.
7320 (thunderxt88_prefetch_tune): Likewise.
7321 (thunderx_prefetch_tune): Likewise.
7322 (thunderx2t99_prefetch_tune): Likewise.
7323 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7325 (aarch64_override_options_internal): Update to set
7326 PARAM_PREFETCH_DYNAMIC_STRIDES.
7327 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7328 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7329 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7330 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7331 prefetch-dynamic-strides setting.
7333 2018-05-07 Luis Machado <luis.machado@linaro.org>
7335 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7336 <minimum_stride>: New const int field.
7337 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7338 minimum_stride field.
7339 (exynosm1_prefetch_tune): Likewise.
7340 (thunderxt88_prefetch_tune): Likewise.
7341 (thunderx_prefetch_tune): Likewise.
7342 (thunderx2t99_prefetch_tune): Likewise.
7343 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7344 (aarch64_override_options_internal): Update to set
7345 PARAM_PREFETCH_MINIMUM_STRIDE.
7346 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7347 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7348 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7349 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7350 stride is constant and is below the minimum stride threshold.
7352 2018-05-07 Luis Machado <luis.machado@linaro.org>
7354 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7357 2018-05-07 Luis Machado <luis.machado@linaro.org>
7359 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7360 <prefetch_dynamic_strides>: New const bool field.
7361 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7362 prefetch_dynamic_strides.
7363 (exynosm1_prefetch_tune): Likewise.
7364 (thunderxt88_prefetch_tune): Likewise.
7365 (thunderx_prefetch_tune): Likewise.
7366 (thunderx2t99_prefetch_tune): Likewise.
7367 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7369 (aarch64_override_options_internal): Update to set
7370 PARAM_PREFETCH_DYNAMIC_STRIDES.
7371 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7372 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7373 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7374 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7375 prefetch-dynamic-strides setting.
7377 2018-05-07 Luis Machado <luis.machado@linaro.org>
7379 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7380 <minimum_stride>: New const int field.
7381 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7382 minimum_stride field.
7383 (exynosm1_prefetch_tune): Likewise.
7384 (thunderxt88_prefetch_tune): Likewise.
7385 (thunderx_prefetch_tune): Likewise.
7386 (thunderx2t99_prefetch_tune): Likewise.
7387 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7388 (aarch64_override_options_internal): Update to set
7389 PARAM_PREFETCH_MINIMUM_STRIDE.
7390 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7391 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7392 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7393 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7394 stride is constant and is below the minimum stride threshold.
7396 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7399 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7400 the type is addressable. Don't force op into register if it has
7403 2018-05-05 Roland McGrath <mcgrathr@google.com>
7406 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7407 any section for which we don't know a specific type it should have,
7408 regardless of name. Previously this was done only for the exact
7409 names ".init_array", ".fini_array", and ".preinit_array".
7410 (default_elf_asm_named_section): Add comment about
7411 relationship with default_section_type_flags and SECTION_NOTYPE.
7412 (get_section): Don't consider it a type conflict if one side has
7413 SECTION_NOTYPE and the other doesn't, as long as neither has the
7414 SECTION_BSS et al used in the default_section_type_flags logic.
7416 2018-05-05 Tom de Vries <tom@codesourcery.com>
7419 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7420 (workaround_barsyncs): New function.
7421 (nvptx_reorg): Use workaround_barsyncs.
7422 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7423 (define_expand "nvptx_membar_cta"): New define_expand.
7424 (define_insn "*nvptx_membar_cta"): New insn.
7426 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7428 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7429 To improve optimization opportunities.
7430 * builtin-types.def: The new needed builtin types for the above.
7432 2018-05-04 Richard Biener <rguenther@suse.de>
7434 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7435 * gimple-ssa-store-merging.c
7436 (imm_store_chain_info::output_merged_store): Remove redundant create,
7437 release split_store vector contents on failure.
7438 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7439 scalar stmt vector on cache hit.
7441 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7443 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7445 * config.gcc (powerpc-xilinx-eabi*): Remove.
7446 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7448 (fusion_addis_mem_combo_load): Ditto.
7449 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7451 (rs6000_cpu_cpp_builtins): Ditto.
7452 * config/rs6000/rs6000-linux.c
7453 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7454 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7455 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7457 (rs6000_setup_reg_addr_masks): Ditto.
7458 (rs6000_init_hard_regno_mode_ok): Ditto.
7459 (rs6000_option_override_internal): Ditto.
7460 (legitimate_lo_sum_address_p): Ditto.
7461 (rs6000_legitimize_address): Ditto.
7462 (rs6000_legitimize_reload_address): Ditto.
7463 (rs6000_legitimate_address_p): Ditto.
7464 (abi_v4_pass_in_fpr): Ditto.
7465 (setup_incoming_varargs): Ditto.
7466 (rs6000_gimplify_va_arg): Ditto.
7467 (rs6000_split_multireg_move): Ditto.
7468 (rs6000_savres_strategy): Ditto.
7469 (rs6000_emit_prologue_components): Ditto.
7470 (rs6000_emit_epilogue_components): Ditto.
7471 (rs6000_emit_prologue): Ditto.
7472 (rs6000_emit_epilogue): Ditto.
7473 (rs6000_elf_file_end): Ditto.
7474 (rs6000_function_value): Ditto.
7475 (rs6000_libcall_value): Ditto.
7476 * config/rs6000/rs6000.h: Ditto.
7477 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7478 (TARGET_MINMAX): ... this. New.
7479 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7480 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7481 (*movsi_internal1_single): Delete.
7482 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7483 mfpu=, mxilinx-fpu): Delete.
7484 * config/rs6000/singlefp.h: Delete.
7485 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7486 * config/rs6000/t-rs6000: Ditto.
7487 * config/rs6000/t-xilinx: Delete.
7488 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7489 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7490 (VStype_simple): Delete.
7491 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7492 * config/rs6000/xfpu.h: Delete.
7493 * config/rs6000/xfpu.md: Delete.
7494 * config/rs6000/xilinx.h: Delete.
7495 * config/rs6000/xilinx.opt: Delete.
7496 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7497 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7499 2018-05-04 Tom de Vries <tom@codesourcery.com>
7502 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7505 2018-05-04 Richard Biener <rguenther@suse.de>
7508 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7509 (expand_complex_div_wide): Likewise.
7510 (expand_complex_operations_1): Likewise.
7511 (expand_complex_libcall): Preserve EH info of the original stmt.
7512 (tree_lower_complex): Handle removed blocks.
7513 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7514 on complex multiplication and division libcall builtins.
7516 2018-05-04 Richard Biener <rguenther@suse.de>
7519 * fold-const.c (negate_expr_p): Restrict negation of operand
7520 zero of a division to when we know that can happen without
7522 (fold_negate_expr_1): Likewise.
7524 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7527 * real.h (real_nextafter): Declare.
7528 * real.c (real_nextafter): New function.
7529 * fold-const-call.c (fold_const_nextafter): New function.
7530 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7531 CASE_CFN_NEXTTOWARD.
7532 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7533 even when arg1_mode is different from arg0_mode.
7535 2018-05-03 Nathan Sidwell <nathan@acm.org>
7537 * doc/extend.texi (Deprecated Features): Remove
7539 (Backwards Compatibility): Likewise.
7540 * doc/invoke.texi (C++ Language Options): Likewise.
7541 (C++ Dialect Options): Likewise.
7543 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7546 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7547 _mm512_mask_mullox_epi64): New intrinsics.
7549 2018-05-03 Tom de Vries <tom@codesourcery.com>
7552 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7553 dump files): Add offload-tree.
7555 2018-05-03 Richard Biener <rguenther@suse.de>
7557 PR tree-optimization/85615
7558 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7559 to loops not nested in BBs loop father to avoid creating multi-entry
7562 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7564 PR tree-optimization/70291
7565 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7566 arguments. Change return type to tree. Emit libcall as a new
7567 statement rather than replacing existing one when inplace_p is true.
7568 (expand_complex_multiplication_components): New function.
7569 (expand_complex_multiplication): Expand floating-point complex
7570 multiplication using the above.
7571 (expand_complex_division): Rename inner_type parameter to type.
7572 Update expand_complex_libcall call-site.
7573 (expand_complex_operations_1): Update expand_complex_multiplication
7574 and expand_complex_division call-sites.
7576 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7579 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7580 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7581 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7582 the highest significant bit of the shift count mask is clear. In
7583 check whether and[sq]i3 is needed verify that all significant bits
7584 of the shift count other than the highest are set.
7586 2018-05-02 Tom de Vries <tom@codesourcery.com>
7589 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7590 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7591 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7592 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7593 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7594 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7595 __builtin_goacc_parlevel_size.
7597 2018-05-02 Richard Biener <rguenther@suse.de>
7599 PR tree-optimization/85597
7600 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7601 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7604 2018-05-02 Tom de Vries <tom@codesourcery.com>
7607 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7608 dump files): Add ltrans-tree.
7610 2018-05-02 Tom de Vries <tom@codesourcery.com>
7613 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7614 dump files): Add wpa-ipa.
7616 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7618 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7619 powerpc*-*-linux*paired* target.
7620 * config/rs6000/750cl.h: Delete.
7621 * config/rs6000/paired.h: Delete.
7622 * config/rs6000/paired.md: Delete.
7623 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7625 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7626 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7627 comment. Remove paired float support.
7628 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7629 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7631 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7632 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7634 * config/rs6000/rs6000.c: Remove paired float support.
7635 (paired_expand_vector_init, paired_expand_vector_move,
7636 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7637 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7638 paired_expand_builtin, paired_expand_predicate_builtin,
7639 paired_init_builtins): Delete.
7640 * config/rs6000/rs6000.h: Remove paired float support.
7641 * config/rs6000/rs6000.md: Remove paired float support.
7642 (move_from_CR_ov_bit): Delete.
7643 * config/rs6000/rs6000.opt (mpaired): Delete.
7644 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7645 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7647 2018-05-02 Richard Biener <rguenther@suse.de>
7650 * gimplify.c (gimplify_save_expr): When in SSA form allow
7651 SAVE_EXPRs to compute to SSA vars.
7653 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7656 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7657 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7658 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7659 clobber operands[2], instead use a new pseudo. Formatting fixes.
7661 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7663 PR tree-optimization/85586
7664 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7665 exit early for statements in the same group if the accesses are
7668 2018-05-02 Tom de Vries <tom@codesourcery.com>
7671 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7674 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7676 PR tree-optimization/85143
7677 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7679 2018-05-01 Tom de Vries <tom@codesourcery.com>
7682 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7683 not found" error message.
7685 2018-05-01 Tom de Vries <tom@codesourcery.com>
7688 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7689 * vec.c (test_ordered_remove_if): New function.
7690 (vec_c_tests): Call test_ordered_remove_if.
7691 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7692 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7693 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7694 VEC_ORDERED_REMOVE_IF.
7696 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7698 PR tree-optimization/82665
7699 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7700 pointer subtraction where arguments come from a memchr call.
7702 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7704 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7705 --push-state --as-needed and --pop-state instead of --as-needed and
7706 --no-as-needed if ld supports it.
7707 * configure: Regenerated.
7710 * doc/install.texi2html: Replace _002d with - and _002a with * in
7711 generated html files using sed.
7713 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7716 * gcc-rich-location.c (blank_line_before_p): New function.
7717 (use_new_line): New function.
7718 (gcc_rich_location::add_fixit_insert_formatted): New function.
7719 * gcc-rich-location.h
7720 (gcc_rich_location::add_fixit_insert_formatted): New function.
7722 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7724 * selftest.c (assert_streq): Rename "expected" and "actual" to
7725 "val1" and "val2". Extend NULL-handling to cover both inputs
7726 symmetrically, while still requiring both to be non-NULL for a pass.
7727 * selftest.h (assert_streq): Rename "expected" and "actual" to
7729 (ASSERT_EQ): Likewise.
7730 (ASSERT_EQ_AT): Likewise.
7731 (ASSERT_KNOWN_EQ): Likewise.
7732 (ASSERT_KNOWN_EQ_AT): Likewise.
7733 (ASSERT_NE): Likewise.
7734 (ASSERT_MAYBE_NE): Likewise.
7735 (ASSERT_MAYBE_NE_AT): Likewise.
7736 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7737 the assertion to pass.
7738 (ASSERT_STREQ_AT): Likewise.
7740 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7742 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7743 interaction with -pie.
7745 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7747 * selftest.h: Fix alphabetization of per-source-file selftest
7750 2018-04-30 Jason Merrill <jason@redhat.com>
7752 PR c++/61982 - dead stores to destroyed objects.
7753 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7756 2018-04-30 Jason Merrill <jason@redhat.com>
7758 * tree.c (build_clobber): New.
7759 * tree.h: Declare it.
7760 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7762 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7764 * diagnostic-show-locus.c (layout::layout): Update for
7765 location_get_source_line returning a char_span.
7766 (struct char_span): Move to input.h.
7767 (struct correction): Update for fields in char_span becoming
7769 (struct source_line): Update for location_get_source_line
7770 returning a char_span.
7771 (layout::print_line): Likewise.
7772 * edit-context.c (edited_file::print_content): Likewise.
7773 (edited_file::print_diff_hunk): Likewise.
7774 (edited_file::print_run_of_changed_lines): Likewise.
7775 (edited_file::get_num_lines): Likewise.
7776 (edited_line::edited_line): Likewise.
7777 * final.c (asm_show_source): Likewise.
7778 * input.c (location_get_source_line): Convert return type
7779 from const char * to char_span, losing the final "line_len"
7781 (dump_location_info): Update for the above.
7782 (get_substring_ranges_for_loc): Likewise. Use a char_span
7783 when handling the literal within the line.
7784 (test_reading_source_line): Update for location_get_source_line
7785 returning a char_span.
7786 * input.h (class char_span): Move here from
7787 diagnostic-show-locus.c, converting from a struct to a class.
7788 Make data members private.
7789 (char_span::operator bool): New.
7790 (char_span::length): New.
7791 (char_span::get_buffer): New.
7792 (char_span::operator[]): New.
7793 (char_span::subspan): Make const.
7794 (char_span::xstrdup): New.
7795 (location_get_source_line): Convert return type from const char *
7796 to char_span, losing the final "line_size" param.
7798 2018-04-30 Jan Hubicka <jh@suse.cz>
7800 * lto-wrapper.c (ltrans_priorities): New static var.
7801 (cmp_priority): New.
7802 (run_gcc): Read priorities and if doing parallel build order
7803 the Makefile by them.
7805 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7807 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7809 2018-04-30 Richard Biener <rguenther@suse.de>
7811 * tree-cfg.c (verify_address): Remove base argument, add
7812 flag whether to check TREE_ADDRESSABLE and do that.
7813 (verify_expr): Remove.
7814 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7815 (verify_gimple_assign_single): Likewise.
7816 (verify_gimple_switch): Likewise.
7817 (verify_expr_location_1): Dereference tp once. Add (disabled)
7818 piece from verify_expr.
7819 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7821 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7823 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7825 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7827 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7828 (small_data_pattern): Likewise.
7829 (arc_rewrite_small_data): Likewise.
7830 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7831 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7832 (get_symbol_alignment): New function.
7833 (legitimate_small_data_address_p): Likewise.
7834 (legitimate_scaled_address): Update, call
7835 legitimate_small_data_address_p.
7836 (output_sdata): New static variable.
7837 (arc_print_operand): Update how we handle small data operands.
7838 (arc_print_operand_address): Likewise.
7839 (arc_legitimate_address_p): Update, use
7840 legitimate_small_data_address_p.
7841 (arc_rewrite_small_data_p): Remove.
7842 (arc_rewrite_small_data_1): Likewise.
7843 (arc_rewrite_small_data): Likewise.
7844 (small_data_pattern): Likewise.
7845 (compact_sda_memory_operand): Update to use
7846 legitimate_small_data_address_p and get_symbol_alignment.
7847 (prepare_move_operands): Don't rewite sdata pattern.
7848 (prepare_extend_operands): Remove.
7849 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7851 (zero_extendqisi2): Likewise.
7852 (zero_extendhisi2): Likewise.
7853 (extendqihi2): Likewise.
7854 (extendqisi2): Likewise.
7855 (extendhisi2): Likewise.
7859 * config/arc/constraints.md (Usd): Change it to memory constraint.
7861 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7863 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7864 as source of std instructions.
7865 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7866 allow 6-bit constants as source for store instructions.
7867 (movdi_insn): Update instruction pattern to allow 6-bit constants
7868 as source for store instructions.
7870 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7872 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7874 2018-04-30 Nathan Sidwell <nathan@acm.org>
7875 Sandra Loosemore <sandra@codesourcery.com>
7877 * dumpfile.c (dump_open): Allow '-' for stdout.
7878 * doc/invoke.texi (Developer Options): Document dump filename
7879 determination early. Document stdin/stdout selection.
7881 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7883 Microblaze Target: PIC data text relative
7885 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7886 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7888 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7889 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7890 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7891 New addressing mode for data-text relative position indepenedent code.
7892 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7893 'ADDRESS_SYMBOLIC_TXT_REL'.
7894 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7895 (microblaze_legitimate_pic_operand): Exclude function calls from
7896 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7897 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7899 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7900 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7901 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7902 for 'address + offset'.
7903 (microblaze_expand_prologue): Add new function prologue call for
7905 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7906 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7907 table in case of TARGET_PIC_DATA_TEXT_REL.
7908 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7909 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7910 Add new macros 'UNSPEC_TEXT',
7911 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7912 + exclude function calls from 'UNSPEC_PLT' in case of data text
7914 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7915 new target hook for generating address diff vector tables in case of
7917 * doc/tm.texi : Regenerate.
7918 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7919 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7920 of addr diff vector generation.
7921 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7922 target hook definition.
7923 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7924 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7925 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7926 option for data text relative.
7928 2018-04-30 Richard Biener <rguenther@suse.de>
7930 * tree-chrec.h (evolution_function_is_constant_p): Remove
7932 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7934 2018-04-30 Richard Biener <rguenther@suse.de>
7937 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7939 2018-04-30 Richard Biener <rguenther@suse.de>
7941 PR tree-optimization/28364
7942 PR tree-optimization/85275
7943 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7944 copying first exit test.
7946 2018-04-28 Mark Wielaard <mark@klomp.org>
7948 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7950 (dwarf_AT): Handle DW_AT_addr_base.
7951 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7953 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7956 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7957 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7958 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7959 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7961 2018-04-28 Richard Biener <rguenther@suse.de>
7963 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7964 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7965 to reflect use. Only add interesting stmts.
7967 2018-04-27 Martin Jambor <mjambor@suse.cz>
7970 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7971 the jump function allows for passing through aggregate values.
7973 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7975 * input.h (in_system_header_at): Convert from macro to inline
7977 (from_macro_expansion_at): Likewise.
7978 (from_macro_definition_at): Likewise.
7980 2018-04-27 Jeff Law <law@redhat.com>
7982 * config.gcc: Mark tile* targets as deprecated/obsolete.
7984 2018-04-27 Richard Biener <rguenther@suse.de>
7986 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7989 2018-04-27 Richard Biener <rguenther@suse.de>
7991 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7993 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7995 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7996 with Yd constraint. Set "preferred_for_speed" attribute from
7997 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7999 (*movdi_internal): Ditto.
8000 (movti_interunit splitters): Remove
8001 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8002 (movdi_interunit splitters): Ditto.
8003 * config/i386/constraints.md (Ye): Remove.
8004 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8006 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8009 * config/aarch64/constraints.md (Usg): Limit to 31.
8012 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8014 PR tree-optimization/85529
8015 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8016 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8017 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8018 zero extension or masking of the MSB bit.
8019 (optimize_range_tests): Add FIRST_BB argument, pass it through
8020 to optimize_range_tests_var_bound.
8021 (maybe_optimize_range_tests, reassociate_bb): Adjust
8022 optimize_range_tests callers.
8024 2018-04-26 Richard Biener <rguenther@suse.de>
8025 Jakub Jelinek <jakub@redhat.com>
8027 * cgraph.h (symbol_table): Just declare debug method here.
8028 * symtab.c (symbol_table::debug): Define.
8030 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8032 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8034 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8036 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8037 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8038 (*movdi_internal): Substitute Yi and Yj constraint with x
8039 and Ym and Yn constraint with y constraint. Update "isa"
8040 attribute and set "preferred_for_speed" attribute from
8041 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8042 (*movsi_internal): Ditto.
8043 (*movdf_internal): Ditto.
8044 (*movsf_internal): Ditto.
8045 (*zero_extendsidi2): Ditto.
8046 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8047 (sse2_loadld): Ditto.
8048 (*vec_extract<ssevecmodelower>_0): Ditto.
8049 (*vec_extractv4si_0_zext_sse4): Ditto.
8050 (vec_concatv2di): Ditto.
8051 (*vec_dup<mode>): Ditto.
8052 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8053 * config/i386/constraints.md (Yi): Remove.
8058 2018-04-26 Nathan Sidwell <nathan@acm.org>
8060 * dumpfile.c (dump_open): New.
8061 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8062 (dump_finish): Detect stdio/stderr by value not name.
8064 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8066 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8068 2018-04-26 Tom de Vries <tom@codesourcery.com>
8071 * config/nvptx/nvptx.c (verify_neutering_jumps)
8072 (verify_neutering_labels): New function
8073 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8075 2018-04-26 Tom de Vries <tom@codesourcery.com>
8078 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8079 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8082 2018-04-26 Richard Biener <rguenther@suse.de>
8083 Tom de Vries <tom@codesourcery.com>
8086 * lto-streamer-out.c (output_function): Fixup loops if required to match
8087 discovery done in the reader.
8089 2018-04-26 Richard Biener <rguenther@suse.de>
8091 PR tree-optimization/85116
8092 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8093 have a loop exit from the single latch predecessor. Remove
8094 case of header with just condition.
8095 (ch_base::copy_headers): Exclude infinite loops from any
8097 (pass_ch::execute): Record exits.
8099 2018-04-26 Richard Biener <rguenther@suse.de>
8101 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8102 prologue cost vector and pass it to vect_get_load_cost.
8103 (vect_get_peeling_costs_all_drs): Likewise.
8104 (vect_peeling_hash_get_lowest_cost): Likewise.
8105 (vect_enhance_data_refs_alignment): Likewise.
8107 2018-04-26 Richard Biener <rguenther@suse.de>
8110 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8111 checking of integer<->pointer conversions.
8112 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8113 sign-/zero-extending pointer types.
8114 (expand_omp_for_static_chunk): Likewise.
8116 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8117 Jean Lee <xiaoyur347@gmail.com>
8119 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8120 (TARGET_ASAN_SHADOW_OFFSET): Define.
8121 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8122 true for -fsanitize=address.
8124 2018-04-25 Mark Wielaard <mark@klomp.org>
8126 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8129 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8131 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8132 than "alu", remove explicit "memory" and "imm_disp" attributes.
8133 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8136 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8137 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8138 gen_lowpart_no_emit.
8140 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8143 * config/i386/i386.c (ix86_expand_builtin): Change memory
8144 operand to XI, extend p0 to Pmode.
8145 * config/i386/i386.md: Change unspec volatile and operand
8146 1 mode to XI, change operand 0 mode to P.
8148 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8150 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8151 GET_MODE_MASK before any checking.
8152 (nds32_can_use_bset_p): Likewise.
8153 (nds32_can_use_btgl_p): Likewise.
8155 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8157 * config/nds32/nds32-doubleword.md: New define_split pattern for
8158 illegal register number.
8160 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8162 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8164 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8166 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8168 2018-04-25 Richard Biener <rguenther@suse.de>
8170 * lto-streamer.h (LTO_major_version): Bump to 8.
8172 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8174 * BASE-VER: Set to 9.0.0.
8176 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8178 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8179 in __abskf2 and __powikf2.
8181 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8184 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8185 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8186 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8187 Use the above on operand 2. Reindent.
8188 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8190 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8193 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8195 * config/i386/i386.opt (mcet): Removed.
8196 * doc/install.texi: Remove -mcet documentation.
8197 * doc/invoke.texi: Likewise.
8199 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8202 * doc/install.texi: Remove -mcet from bootstrap-cet.
8204 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8207 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8208 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8212 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8213 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8214 containing a CONST_VECTOR.
8216 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8218 * doc/install.texi: Update newlib dependency for nvptx.
8220 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8223 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8224 instead of INTVAL when shifting x left.
8226 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8228 PR tree-optimization/85478
8229 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8230 vect_grouped_store_supported for single element vectors.
8232 2018-04-24 Richard Biener <rguenther@suse.de>
8235 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8236 load cost increase to the case of non-constant step.
8238 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8241 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8242 destination if any_malformed_asm.
8244 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8247 * expr.c (store_field): In the bitfield case, if the value comes from
8248 a function call and is returned in registers by means of a PARALLEL,
8249 do not change the mode of the temporary unless BLKmode and VOIDmode.
8251 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8253 PR rtl-optimization/85423
8254 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8255 dependencies to debug insns when the previous insn is non-debug.
8257 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8259 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8260 enums into a single definition.
8261 (fls): Fix predicates and printing.
8264 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8266 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8267 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8268 and short u6 immediate.
8269 (check_if_valid_sleep_operand): Remove.
8270 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8272 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8274 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8275 flag_always_save_lp condition.
8276 * config/nds32/nds32.opt (malways-save-lp): New option.
8278 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8280 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8281 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8282 * config/nds32/nds32.h
8283 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8284 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8286 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8288 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8289 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8291 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8292 Chung-Ju Wu <jasonwucj@gmail.com>
8294 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8296 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8297 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8299 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8301 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8303 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8305 * config/nds32/nds32-protos.h (nds32_data_alignment,
8306 nds32_local_alignment): Declare.
8307 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8308 nds32_local_alignment): New functions.
8309 (TARGET_CONSTANT_ALIGNMENT): Define.
8310 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8312 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8314 * config/nds32/nds32.c
8315 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8316 (TARGET_MODES_TIEABLE_P): Likewise.
8318 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8320 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8323 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8324 Chung-Ju Wu <jasonwucj@gmail.com>
8326 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8327 for unaligned access.
8328 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8330 * config/nds32/nds32-intrinsic.md: Likewise.
8331 * config/nds32/nds32_intrinsic.h: Likewise.
8332 * config/nds32/nds32.h (nds32_builtins): Likewise.
8333 * config/nds32/nds32.opt (munaligned-access): New option.
8334 * config/nds32/nds32.c (nds32_asm_file_start): Display
8335 flag_unaligned_access status.
8337 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8339 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8340 -mno-relax is present.
8341 * config/riscv/linux.h (LINK_SPEC): Ditto.
8343 2018-04-20 Martin Sebor <msebor@redhat.com>
8346 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8348 (gimple_fold_builtin_stxcpy_chk): Same.
8349 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8351 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8354 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8355 __powikf2 when long double is IEEE 128-bit.
8357 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8359 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8360 step to make sure stack always aligned.
8362 2018-04-20 Carl Love <cel@us.ibm.com>
8365 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8366 size check for arg0.
8368 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8369 Tom de Vries <tom@codesourcery.com>
8372 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8373 Emit insns for calls too.
8374 (nvptx_find_par): Always look for worker-level predecessor insn.
8375 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8377 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8378 (nvptx_process_pars): Propagate frames for calls.
8380 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8383 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8385 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8386 (ix86_handle_option): Don't handle OPT_mibt.
8387 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8389 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8391 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8392 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8393 (ix86_target_macros): Define __CET__ with flag_cf_protection
8394 for -fcf-protection.
8395 * config/i386/i386.c (isa2_opts): Remove -mibt.
8396 * config/i386/i386.h (TARGET_IBT): Removed.
8397 (TARGET_IBT_P): Likewise.
8398 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8399 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8400 * config/i386/i386.opt (mcet): Update help message.
8403 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8404 -mcet as an alias for -mshstk.
8406 2018-04-20 Richard Biener <rguenther@suse.de>
8409 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8410 complexity by forcing a single use of the multiply operand.
8412 2018-04-20 Martin Jambor <mjambor@suse.cz>
8415 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8416 recursion dependency to only apply to non-clones.
8418 2018-04-20 Martin Jambor <mjambor@suse.cz>
8421 * ipa-cp.c (create_specialized_node): Check that clones of
8422 self-recursive edges exist during IPA-CP.
8424 2018-04-19 Toon Moene <toon@moene.org>
8426 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8429 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8431 PR tree-optimization/85467
8432 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8433 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8434 VECTOR_CST element to type.
8436 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8439 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8440 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8441 (builtin_longjmp): Likewise.
8442 (save_stack_nonlocal): New pattern.
8443 (restore_stack_nonlocal): Likewise.
8445 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8448 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8449 Replace ASM_OUTPUT_LABEL with fprintf.
8451 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8454 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8455 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8456 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8457 define __IBT__ and __SHSTK__ for -fcf-protection.
8458 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8460 (ix86_trampoline_init): Likewise.
8461 (x86_output_mi_thunk): Likewise.
8462 (ix86_notrack_prefixed_insn_p): Likewise.
8463 (ix86_option_override_internal): Don't disallow -fcf-protection.
8464 * config/i386/i386.md (rdssp<mode>): Also enable for
8466 (incssp<mode>): Likewise.
8467 (nop_endbr): Likewise.
8468 * config/i386/i386.opt (mcet): Change help message to built-in
8472 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8473 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8474 enable CET built-in functions.
8476 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8478 * common/config/i386/i386-common.c
8479 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8480 OPTION_MASK_ISA_MOVDIRI_UNSET,
8481 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8482 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8483 * config.gcc (movdirintrin.h): New header.
8484 * config/i386/cpuid.h (bit_MOVDIRI,
8485 bit_MOVDIR64B): New bits.
8486 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8488 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8489 (VOID, PVOID, PCVOID)): New function types.
8490 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8491 __builtin_ia32_directstoreu_u64,
8492 __builtin_ia32_movdir64b): New builtins.
8493 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8494 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8496 (ix86_valid_target_attribute_inner_p): Ditto.
8497 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8498 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8499 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8500 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8501 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8502 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8503 (movdiri<mode>, movdir64b_<mode>): New.
8504 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8505 * config/i386/immintrin.h: Include movdirintrin.h.
8506 * config/i386/movdirintrin.h: New file.
8507 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8509 2018-04-19 Richard Biener <rguenther@suse.de>
8512 * cfg.c (clear_bb_flags): When loop state says we have
8513 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8515 2018-04-19 Richard Biener <rguenther@suse.de>
8517 PR tree-optimization/84737
8518 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8519 copying restrict info.
8520 (vect_setup_realignment): Use it.
8521 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8522 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8523 the first DR to all generated stores.
8524 (vectorizable_load): Likewise for loads.
8526 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8528 PR tree-optimization/85446
8529 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8530 the integral and pointer types to have the same precision.
8532 * doc/install.texi: Document --disable-cet being the default and
8535 2018-04-18 Martin Liska <mliska@suse.cz>
8537 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8540 2018-04-18 Martin Liska <mliska@suse.cz>
8543 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8546 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8547 arguments if they are comparable.
8549 2018-04-18 Martin Liska <mliska@suse.cz>
8552 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8555 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8558 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8561 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8562 ENDBR after calling __morestack.
8564 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8567 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8568 by using gcc_base_ver to generate a gcc_driver_version, and use
8569 it when generating GCC_DRIVER_NAME.
8570 * configure: Regenerate.
8572 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8575 * config.gcc: Obsolete powerpc*-*-*spe*.
8577 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8580 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8581 (stabstr_D): Change type of unum from unsigned int to
8582 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8585 2018-04-17 Jim Wilson <jimw@sifive.com>
8588 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8589 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8590 Set arg_pointer_offset after using pretend_args_size.
8592 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8594 PR rtl-optimization/85431
8595 * dse.c (record_store): Ignore zero width stores.
8598 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8599 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8600 __builtin_stack_restore rather than after it.
8601 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8602 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8603 argument instead of virtual_dynamic_stack_rtx.
8605 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8607 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8609 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8610 Add note to error message to explain internal mapping of overloaded
8611 built-in function name to non-overloaded built-in function name.
8612 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8615 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8618 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8619 where the inputs overlap with the output.
8621 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8623 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8624 (=v, v) alternative and explicit "memory" attribute.
8625 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8626 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8628 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8629 "sselog1" type instead of "sselog".
8630 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8631 "sselog". Remove explicit "memory" attribute.
8632 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8633 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8635 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8636 "sselog1" type instead of "sselog". Remove explicit "memory"
8638 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8639 use "sselog1" type instead of "sselog". Remove explicit "memory"
8641 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8642 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8644 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8645 "sselog1" type instead of "sselog". Remove explicit "memory"
8647 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8648 use "sselog1" type instead of "sselog". Remove explicit "memory"
8652 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8655 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8658 2018-04-17 Martin Jambor <mjambor@suse.cz>
8661 * ipa-cp.c (create_specialized_node): Call
8662 expand_all_artificial_thunks if necessary.
8664 2018-04-17 Martin Liska <mliska@suse.cz>
8667 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8668 in message, remote space in between '_G' and '('.
8670 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8673 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8674 avx512f_vmcmp<mode>3<round_saeonly_name>,
8675 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8676 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8677 avx512f_rndscale<mode><round_saeonly_name>,
8678 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8679 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8680 Use %<iptr>2 instead of %2 for -masm=intel.
8681 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8682 avx512f_vcvttss2usi<round_saeonly_name>,
8683 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8685 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8686 avx512f_vcvttsd2usi<round_saeonly_name>,
8687 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8688 Use %q1 instead of %1 for -masm=intel.
8689 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8690 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8691 of %3 for -masm=intel.
8692 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8694 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8696 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8698 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8699 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8701 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8703 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8704 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8705 %g1 and one with %0 and %1.
8706 (avx512er_vmrcp28<mode><round_saeonly_name>,
8707 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8709 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8710 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8711 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8712 of %0 and %{%4%} for -masm=intel.
8713 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8714 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8715 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8716 order of %0 and %{%5%}%{z%} for -masm=intel.
8718 2018-04-17 Jan Hubicka <jh@suse.cz>
8721 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8723 2018-04-17 Martin Liska <mliska@suse.cz>
8726 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8727 for target_clone error message. Make default implementation
8728 clone to be a local declaration.
8729 (separate_attrs): Add new argument and check for an empty
8731 (expand_target_clones): Handle it.
8732 (ipa_target_clone): Make redirection just for target_clones
8735 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8736 Tom de Vries <tom@codesourcery.com>
8739 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8740 tiled basic blocks without omp continue statements.
8742 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8745 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8746 vec_extract expression as having side effects to make sure it gets
8749 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8752 * config/i386/i386.c (get_builtin_code_for_version): Check
8755 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8758 * gcc/config.gcc: Support "skylake".
8759 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8761 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8762 (processor_target_table): Add "skylake".
8763 (ix86_option_override_internal): Add "skylake".
8764 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8765 PROCESSOR_CANNONLAKE.
8766 (get_builtin_code_for_version): Fix priority for
8767 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8768 PROCESSOR_SKYLAKE-AVX512.
8769 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8770 (processor_type): Add PROCESSOR_SKYLAKE.
8772 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8773 Jason Merrill <jason@redhat.com>
8776 * convert.c (convert_to_integer_1): Use direct recursion for
8777 enumeral types and types with a precision less than the number
8778 of bits in their mode.
8780 2018-04-16 Julia Koval <julia.koval@intel.com>
8783 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8784 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8786 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8789 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8790 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8791 and -mno-direct-move.
8793 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8796 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8797 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8798 Ensure negative shifts result in {0}.
8800 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8802 PR rtl-optimization/79916
8803 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8804 regs (if any) to define how to gnerate SD moves when LRA is in
8807 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8809 PR rtl-optimization/85393
8810 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8811 * except.c (expand_dw2_landing_pad_for_region): Make static.
8812 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8813 a label and unconditional jump to old_bb, rather than
8814 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8817 PR rtl-optimization/85376
8818 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8819 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8820 instead of a specific value.
8822 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8823 Bin Cheng <bin.cheng@arm.com>
8825 PR tree-optimization/82965
8826 PR tree-optimization/83991
8827 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8828 by_profile_only parameter.
8829 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8830 information if the loop was predicted to iterate too many times.
8831 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8833 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8836 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8839 2018-04-13 Martin Liska <mliska@suse.cz>
8840 Jakub Jelinek <jakub@redhat.com>
8843 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8844 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8845 * builtins.c (expand_builtin_memory_copy_args): Use
8846 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8847 handle dest_addr == pc_rtx.
8849 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8852 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8853 asked to not generate direct moves.
8854 (fix_trunc<mode>si2_stfiwx): Similar.
8855 (fix_trunc<mode>si2_internal): Similar.
8857 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8860 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8861 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8862 lookup if dest in some wider mode is known to be const0_rtx and
8863 if so, record permanent equivalence for it to be ZERO_EXTEND of
8864 the narrower mode destination.
8866 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8868 * lto-streamer-out.c (output_function): Revert 259346.
8869 * omp-expand.c (expand_oacc_for): Likewise.
8871 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8873 PR rtl-optimization/85354
8874 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8875 * sel-sched.c (sel_global_init): ... here.
8877 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8880 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8881 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8882 mode for PE-COFF targets.
8883 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8884 (i386_pe_asm_lto_end): Likewise.
8885 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8886 (TARGET_ASM_LTO_END): Likewise.
8887 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8888 (i386_pe_asm_lto_start): New function.
8889 (i386_pe_asm_lto_end): Likewise.
8891 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8892 Richard Biener <rguenther@suse.de>
8895 * lto-streamer-out.c (output_function): Fix CFG loop state before
8897 * omp-expand.c (expand_oacc_for): Handle calls to internal
8898 functions like regular functions.
8900 2018-04-12 Richard Biener <rguenther@suse.de>
8903 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8904 for the early LTO debug to properly generate references to it
8905 during DIE emission. Do not re-use that for the skeleton for
8907 (dwarf2out_early_finish): Likewise.
8909 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8912 * config/i386/sse.md
8913 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8914 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8915 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8916 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8917 and output is a reg, avoid creating invalid lowpart subreg, but
8918 instead split into a 512-bit move. Don't split if not AVX512VL,
8919 input is xmm16+ reg and output is a mem.
8920 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8921 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8922 xmm16+ reg and output is a mem.
8924 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8926 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8927 also for flag_dwarf2_cfi_asm.
8929 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8931 PR rtl-optimization/85342
8932 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8933 a bool scalar var inside of the loop instead. Don't try to update
8934 recog_data.operand after failed apply_change_group.
8936 2018-04-12 Tom de Vries <tom@codesourcery.com>
8939 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8940 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8941 array with flexible array member as array without given dimension.
8942 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8943 argument for undefined param to true.
8945 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8948 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8949 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8950 from PowerPC section.
8951 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8952 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8953 help text that is too long.
8954 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8955 help text that is too long.
8956 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8957 help text that is too long.
8959 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8961 * config/alpha/alpha.md (stack_probe_internal): Rename
8962 from "probe_stack". Update all callers.
8964 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8966 PR rtl-optimization/84566
8967 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8968 sched_macro_fuse_insns.
8970 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8973 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8974 (compute_block_dependences): ... from here.
8976 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8978 PR tree-optimization/85331
8979 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8980 from int to HOST_WIDE_INT.
8982 2018-04-11 Martin Jambor <mjambor@suse.cz>
8985 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8986 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8987 not the same as the source val.
8988 (cgraph_edge_brings_value_p): New parameter.
8989 (gather_edges_for_value): Pass destination value to
8990 cgraph_edge_brings_value_p.
8991 (perhaps_add_new_callers): Likewise.
8992 (get_info_about_necessary_edges): Likewise and exclude values brought
8993 only by self-recursive edges.
8994 (create_specialized_node): Redirect only clones of self-calling edges.
8995 (+self_recursive_pass_through_p): New function.
8996 (find_more_scalar_values_for_callers_subset): Use it.
8997 (find_aggregate_values_for_callers_subset): Likewise.
8998 (known_aggs_to_agg_replacement_list): Removed.
8999 (decide_whether_version_node): Re-calculate known constants for all
9000 remaining context clones.
9002 2018-04-11 Richard Biener <rguenther@suse.de>
9005 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9006 from early DWARF output.
9007 (dwarf2out_early_finish): Output line info unconditionally into
9008 early DWARF and add reference to it.
9010 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9013 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9014 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9015 other than V2DFmode using iptr mode attribute.
9016 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9018 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9020 PR rtl-optimization/84659
9021 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9023 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9026 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9028 (output_loc_list): Pass address of a dummy size variable even in the
9029 locview handling loop.
9030 (index_location_lists): Add comment on why skip_loc_list_entry can't
9033 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9036 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9039 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9042 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9043 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9044 and -mstring-compare-inline-limit.
9046 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9049 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9050 for stack clash protection in a register whenever we need it to be in
9053 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9055 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9056 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9058 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9061 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9063 (mlong-double-): Ditto.
9064 * config/rs6000/sysv4.opt (msdata=): Ditto.
9065 (mtls-size=): Ditto.
9067 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9069 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9070 erroneous entries for
9071 "vector int vec_ldl (int, long int *)", and
9072 "vector unsigned int vec_ldl (int, unsigned long int *)".
9073 Add comments and entries for
9074 "vector bool char vec_ldl (int, bool char *)",
9075 "vector bool short vec_ldl (int, bool short *)",
9076 "vector bool int vec_ldl (int, bool int *)",
9077 "vector bool long long vec_ldl (int, bool long long *)",
9078 "vector pixel vec_ldl (int, pixel *)",
9079 "vector long long vec_ldl (int, long long *)",
9080 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9081 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9082 type tree bool_long_long_type_node and correct definition of
9083 bool_V2DI_type_node to make reference to this new type tree.
9084 (rs6000_mangle_type): Replace erroneous reference to
9085 bool_long_type_node with bool_long_long_type_node.
9086 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9087 comments to emphasize sign distinctions for char and int types and
9088 replace RS6000_BTI_bool_long constant with
9089 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9090 use of RS6000_BTI_pixel.
9091 (bool_long_type_node): Remove this macro definition.
9092 (bool_long_long_type_node): New macro definition
9094 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9096 PR rtl-optimization/85300
9097 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9098 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9099 simplify_unary_operation fails.
9101 2018-04-10 Martin Liska <mliska@suse.cz>
9103 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9104 cgraph_edge and ipa_ref.
9106 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9110 * config/i386/sse.md
9111 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9112 computation of the VEC_MERGE selector from mask.
9113 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9114 Fix decoding of the VEC_MERGE selector into mask.
9116 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9118 PR tree-optimization/85286
9119 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9121 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9123 * final.c (final_1): Set insn_last_address as well as
9124 insn_current_address.
9126 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9129 * explow.c (emit_stack_probe): Call validize_mem on memory location
9130 before passing it to gen_probe_stack. Create address operand and
9131 legitimize it for the probe_stack_address case.
9133 2018-04-09 Jan Hubicka <jh@suse.cz>
9136 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9137 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9138 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9140 (free_lang_data): Rebuild type inheritance graph.
9142 2018-04-09 Martin Sebor <msebor@redhat.com>
9144 * invoke.texi (-finline-small-functions): Mention other optimization
9146 (-findirect-inlining, -fpartial-inlining): Same.
9147 (-finline-functions-called-once): Same.
9148 (-freorder-blocks-and-partition): Same.
9150 2018-04-09 Jan Hubicka <jh@suse.cz>
9153 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9154 jumps; choose last target that matches the criteria (i.e.
9155 no partition changes for non-crossing jumps).
9156 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9157 support for redirecting crossing jumps to non-crossing.
9159 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9161 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9162 also for naked functions.
9164 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9166 * config/arc/arc.md (add_shift): New pattern.
9167 (add_shift2): Likewise.
9168 (sub_shift): Likewise.
9169 (sub_shift_cmp0_noout): Likewise.
9170 (compare_si_ashiftsi): Likewise.
9171 (xbfu_cmp0_noout): New combine pattern.
9172 (xbfu_cmp0"): Likewise.
9173 (movsi_set_cc_insn): Place the predicable variant first.
9174 (commutative_binary_cmp0_noout): Remove clobber.
9175 (commutative_binary_cmp0): New pattern.
9176 (noncommutative_binary_cmp0): Likewise.
9177 (noncommutative_binary_cmp0_noout): Likewise.
9178 (noncommutative_binary_comparison_result_used): Removed.
9179 (rsub_cmp0): New pattern.
9180 (rsub_cmp0_noout): Likewise.
9181 (extzvsi): Changed, keep only meaningful variants.
9182 (SQH, SEZ): New iterators.
9183 (SQH_postfix): New mode attribute.
9184 (SEZ_prefix): New code attribute.
9185 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9186 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9187 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9189 (noncommutative_operator): Check the availability of barrel
9192 2018-04-09 Richard Biener <rguenther@suse.de>
9194 PR tree-optimization/85284
9195 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9196 Only use the niter constraining form of simple_iv when the exit
9199 2018-04-09 Tom de Vries <tom@codesourcery.com>
9202 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9203 (define_expand "*memory_barrier"): New define_expand.
9204 (define_insn "memory_barrier"): New insn.
9206 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9208 PR rtl-optimization/80463
9209 PR rtl-optimization/83972
9210 PR rtl-optimization/83480
9212 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9213 correct producer for the insn.
9214 (tidy_control_flow): Fixup seqnos in case of debug insns.
9216 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9218 PR rtl-optimization/83913
9220 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9221 different sched-times when merging exprs.
9223 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9225 PR rtl-optimization/83962
9227 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9228 tidy_fallthru_edge and tidy_control_flow.
9230 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9232 PR rtl-optimization/83530
9234 * sel-sched.c (force_next_insn): New global variable.
9235 (remove_insn_for_debug): When force_next_insn is true, also leave only
9236 next insn in the ready list.
9237 (sel_sched_region): When the region wasn't scheduled, make another pass
9238 over it with force_next_insn set to 1.
9240 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9242 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9244 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9245 for interrupt control.
9246 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9247 functions for interrupt control.
9248 * config/nds32/nds32-intrinsic.md: Likewise.
9249 * config/nds32/nds32_intrinsic.h: Likewise.
9250 * config/nds32/nds32.h (nds32_builtins): Likewise.
9252 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9254 * config/nds32/nds32.c (nds32_init_machine_status,
9255 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9256 strict_aligned_p field.
9257 (nds32_expand_to_rtl_hook): New function.
9258 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9259 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9261 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9262 Chung-Ju Wu <jasonwucj@gmail.com>
9264 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9265 * config/nds32/nds32-n7.md: New file.
9266 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9267 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9269 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9270 * config/nds32/nds32.md (pipeline_model): Add n7.
9271 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9272 * config/nds32/pipelines.md: Include n7 settings.
9274 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9275 Chung-Ju Wu <jasonwucj@gmail.com>
9277 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9278 * config/nds32/nds32-e8.md: New file.
9279 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9280 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9282 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9283 * config/nds32/nds32.md (pipeline_model): Add e8.
9284 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9285 * config/nds32/pipelines.md: Include e8 settings.
9287 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9288 Chung-Ju Wu <jasonwucj@gmail.com>
9290 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9291 * config/nds32/nds32-n8.md: New file.
9292 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9293 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9295 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9296 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9297 * config/nds32/nds32.md (pipeline_model): Add n8.
9298 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9299 * config/nds32/pipelines.md: Include n8 settings.
9301 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9302 Chung-Ju Wu <jasonwucj@gmail.com>
9304 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9305 * config/nds32/nds32-n9-2r1w.md: New file.
9306 * config/nds32/nds32-n9-3r2w.md: New file.
9307 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9308 nds32_register_ports): New or modify for cpu n9.
9309 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9311 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9312 * config/nds32/nds32-utils.c: New file.
9313 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9314 TARGET_MUL_SLOW): Define.
9315 * config/nds32/nds32.md (pipeline_model): New attribute.
9316 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9317 New options that support cpu n9.
9318 * config/nds32/pipelines.md: Include n9 settings.
9319 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9321 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9323 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9324 information if necessary.
9325 (output_cond_branch_compare_zero): Likewise.
9326 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9327 (nds32_target_alignment): Refine for alignment.
9328 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9329 (FUNCTION_BOUNDARY): Modify.
9330 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9332 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9334 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9336 * config/nds32/constants.md (unspec_volatile_element): Add values for
9337 TLB operation and data prefetch.
9338 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9339 functions for TLB operation and data prefetch.
9340 * config/nds32/nds32-intrinsic.md: Likewise.
9341 * config/nds32/nds32_intrinsic.h: Likewise.
9342 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9343 (nds32_print_operand): Likewise.
9344 * config/nds32/nds32.h (nds32_builtins): Likewise.
9346 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9347 Andrew Pinski <pinsika@gcc.gnu.org>
9350 * match.pd: Use constant_boolean_node of correct type instead of
9351 boolean_true_node or boolean_false_node for simplifying
9352 pointer comparisons to zero.
9354 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9356 PR tree-optimization/80021
9357 * tree.c (verify_type_variant): Make error call in verify_variant_match
9358 translatable and remove final full stop.
9360 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9362 * config/nds32/constants.md (unspec_volatile_element): Add
9363 UNSPEC_VOLATILE_EH_RETURN.
9364 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9365 nds32_output_stack_pop): Support dwarf exception handling process.
9366 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9367 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9368 exception handling process.
9369 (nds32_compute_stack_frame): Likewise.
9370 (nds32_return_addr_rtx): Likewise.
9371 (nds32_initial_elimination_offset): Likewise.
9372 (nds32_expand_prologue): Likewise.
9373 (nds32_expand_epilogue): Likewise.
9374 (nds32_dynamic_chain_address): New function.
9375 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9377 (DYNAMIC_CHAIN_ADDRESS): Define.
9378 (EH_RETURN_DATA_REGNO): Define.
9379 (EH_RETURN_STACKADJ_RTX): Define.
9380 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9381 patterns for dwarf exception handling.
9383 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9385 * config/nds32/nds32.h: Clean up obsolete macros.
9387 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9389 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9390 Add enum values for particular instructions.
9391 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9392 particular intrinsic functions.
9393 * config/nds32/nds32-intrinsic.md: Likewise.
9394 * config/nds32/nds32_intrinsic.h: Likewise.
9395 * config/nds32/nds32.h (nds32_builtins): Likewise.
9396 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9397 (btst, ave): New patterns for particular instructions.
9399 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9401 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9402 Add enum values for atomic load/store and memory sync.
9403 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9405 * config/nds32/nds32-intrinsic.md: Likewise.
9406 * config/nds32/nds32_intrinsic.h: Likewise.
9407 * config/nds32/nds32.h (nds32_builtins): Likewise.
9409 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9411 PR tree-optimization/85257
9412 * fold-const.c (native_encode_vector): If not all elts could fit
9413 and off is -1, return 0 rather than offset.
9414 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9415 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9416 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9417 adjust buffer in native_interpret_expr call.
9419 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9421 * config/nds32/constants.md (unspec_volatile_element): Add cache
9422 control enum values.
9423 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9424 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9425 * config/nds32/nds32.c (nds32_cctl_names): New.
9426 (nds32_print_operand): Handle cache control register names.
9427 * config/nds32/nds32.h (nds32_builtins): New enum values.
9428 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9430 * config/nds32/nds32.md (type): Add mmu.
9431 * config/nds32/pipelines.md (simple_insn): Add mmu.
9433 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9435 * config/nds32/nds32.md (type): Remove call.
9436 * config/nds32/pipelines.md (simple_insn): Likewise.
9438 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9440 * config/nds32/constants.md (unspec_volatile_element): Add
9441 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9442 UNSPEC_VOLATILE_FMFCFG.
9443 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9444 description for fmfcfg and fmfcsr.
9445 (bdesc_1arg): Add fmtcsr.
9446 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9447 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9448 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9449 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9450 unspec_fmfcfg): New patterns.
9451 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9452 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9453 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9454 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9455 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9456 __nds32__fmfcfg): Define.
9458 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9460 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9461 intrinsic register names.
9462 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9463 intrinsic register enum values and macros.
9465 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9467 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9468 for load/store addressing form.
9469 (nds32_print_operand_address): Likewise.
9471 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9474 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9475 based on LABEL_REF. Remove useless assertion.
9476 (pic_address_needs_scratch): Fix formatting.
9477 (sparc_legitimize_pic_address): Minor tweaks.
9478 (sparc_delegitimize_address): Adjust assertion accordingly.
9479 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9480 into symbolic_operand.
9481 (movsi_high_pic_label_ref): Likewise.
9482 (movsi_lo_sum_pic_label_ref): Likewise.
9483 (movdi_pic_label_ref): Likewise.
9484 (movdi_high_pic_label_ref): Likewise.
9485 (movdi_lo_sum_pic_label_ref): Likewise.
9487 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9489 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9490 custom LIB_SPEC setup.
9492 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9493 Kito Cheng <kito.cheng@gmail.com>
9495 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9496 * config/riscv/freebsd.h: New.
9498 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9500 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9501 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9504 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9505 Kito Cheng <kito.cheng@gmail.com>
9507 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9508 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9509 * config/nds32/nds32-protos.h (nds32_output_call,
9510 nds32_output_return): Declare.
9511 * config/nds32/nds32.md: Refine all the call and return patterns.
9513 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9516 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9517 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9519 PR rtl-optimization/84872
9520 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9521 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9524 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9526 * expr.c (copy_blkmode_to_reg): Revert 254862.
9527 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9529 2018-04-06 Richard Biener <rguenther@suse.de>
9532 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9533 after seeing a component reference with an adjacent field. Treat
9534 refs to arrays at struct end of external decls similar to
9535 refs to unconstrained commons.
9537 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9540 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9541 look through SAVE_EXPRs with non-side-effects argument. Adjust
9543 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9546 2018-04-06 Richard Biener <rguenther@suse.de>
9549 * alias.c (find_base_term): New wrapper around find_base_term
9550 unwinding CSELIB_VAL_PTR changes.
9551 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9554 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9556 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9558 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9559 constant definitions.
9560 ("nop"): lr 0,0 -> nopr r0
9561 ("nop_lr0", "nop_lr1"): New insn definitions.
9563 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9565 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9566 NDS32_V3PUSH_AVAILABLE_P macro.
9568 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9569 Chung-Ju Wu <jasonwucj@gmail.com>
9571 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9572 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9573 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9574 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9575 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9576 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9577 * config/nds32/constraints.md: New constraints and checking for hard
9578 float configuration.
9579 * config/nds32/iterators.md: New mode iterator and attribute for hard
9580 float configuration.
9581 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9583 * config/nds32/nds32-fpu.md: New file.
9584 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9585 deal with hard float code generation.
9586 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9588 (abi_type, float_reg_number): New enum type.
9589 * config/nds32/nds32-predicates.c: New predicates for hard float.
9590 * config/nds32/nds32-protos.h: Declare functions for hard float.
9591 * config/nds32/nds32.c: Implementation for hard float configuration.
9592 * config/nds32/nds32.h: Definitions for hard float configuration.
9593 * config/nds32/nds32.md: Include hard float machine description and
9594 modify patterns for hard float configuration.
9595 * config/nds32/nds32.opt: New options for hard float configuration.
9596 * config/nds32/predicates.md: New predicates for hard float
9599 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9601 * common/config/nds32/nds32-common.c
9602 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9604 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9607 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9608 CONSTRUCTOR_ELT (ctor, ...)->value.
9610 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9613 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9615 2018-04-05 Tom de Vries <tom@codesourcery.com>
9618 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9621 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9622 Kito Cheng <kito.cheng@gmail.com>
9624 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9625 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9626 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9627 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9629 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9630 Kito Cheng <kito.cheng@gmail.com>
9632 * config/nds32/constraints.md (Ufe): New memory constraint.
9633 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9634 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9635 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9637 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9638 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9640 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9642 * config/nds32/nds32.md: Use optimize_size in the condition for
9643 alu-shift instructions.
9645 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9647 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9649 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9651 * config/nds32/nds32.md (negsi2): Refine pattern.
9653 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9654 Chung-Ju Wu <jasonwucj@gmail.com>
9656 * config/nds32/iterators.md (shift_rotate): New code iterator.
9657 (shift): New code attribute.
9658 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9659 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9660 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9661 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9662 bit-wise operations.
9663 (andsi3, *andsi3): Ditto.
9664 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9665 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9666 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9667 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9668 nds32_ior_operand, nds32_xor_operand): New predicates.
9670 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9672 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9673 (addsi3, subsi3): ... this.
9675 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9677 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9679 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9681 * config/nds32/nds32.md: Adjust indention.
9683 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9685 * config/nds32/nds32.md (feature): New attribute.
9687 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9689 * config/nds32/nds32.md (subtype): New attribute.
9691 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9694 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9695 expansion to perform a bitwise AND of the argument followed by a
9696 boolean negation of the result.
9698 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9700 PR rtl-optimization/84878
9701 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9702 the basic block. Assert the use reference is not artificial and that
9703 it has an associated insn.
9705 2018-04-04 Michael Matz <matz@suse.de>
9707 * builtins.c (compute_objsize): Pass correct operand
9708 to array_at_struct_end_p.
9710 2018-04-04 Richard Biener <rguenther@suse.de>
9713 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9714 from contexts for DINFO_LEVEL_TERSE and below.
9716 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9718 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9719 resiter_operand condition.
9720 * config/nds32/nds32.md (*move<mode>): Ditto.
9722 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9723 Monk Chiang <sh.chiang04@gmail.com>
9725 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9727 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9729 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9731 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9732 Kito Cheng <kito.cheng@gmail.com>
9734 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9735 nds32_cond_code_str, output_cond_branch,
9736 output_cond_branch_compare_zero, nds32_expand_cbranch,
9737 nds32_expand_cstore, nds32_expand_movcc,
9738 nds32_output_cbranchsi4_equality_zero,
9739 nds32_output_cbranchsi4_equality_reg,
9740 nds32_output_cbranchsi4_equality_reg_or_const_int,
9741 nds32_output_cbranchsi4_greater_less_zero: New functions.
9742 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9743 nds32_expand_cstore, nds32_expand_movcc,
9744 nds32_output_cbranchsi4_equality_zero,
9745 nds32_output_cbranchsi4_equality_reg,
9746 nds32_output_cbranchsi4_equality_reg_or_const_int,
9747 nds32_output_cbranchsi4_greater_less_zero): Declare.
9748 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9749 nds32_rimm11s_operand): New predicates.
9750 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9751 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9754 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9756 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9757 * config/nds32/nds32.md: Ditto.
9758 * config/nds32/pipelines.md: Ditto.
9760 2018-04-04 Richard Biener <rguenther@suse.de>
9762 PR tree-optimization/85168
9763 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9764 propagating abnormals.
9766 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9768 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9770 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9771 Kito Cheng <kito.cheng@gmail.com>
9773 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9774 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9775 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9776 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9777 * config/nds32/nds32.md (sibcall_internal): New.
9778 (sibcall_register): Remove.
9779 (sibcall_immediate): Remove.
9780 (sibcall_value_internal): New.
9781 (sibcall_value_register): Remove.
9782 (sibcall_value_immediate): Remove.
9783 * config/nds32/predicates.md (nds32_general_register_operand): New.
9784 (nds32_call_address_operand): New.
9786 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9788 PR rtl-optimization/85167
9789 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9790 bb_defs if *split_p, instead preinitialize it to NULL.
9792 PR tree-optimization/85156
9793 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9794 evaluating the argument multiple times.
9796 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9798 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9800 (_mm_cvtpd_ps): Likewise.
9801 (_mm_cvttpd_epi32): Likewise.
9802 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9803 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9804 vector, pixel, and bool following altivec.h include.
9806 2018-04-03 Martin Sebor <msebor@redhat.com>
9808 * doc/extend.texi (Common Function Attributes): Clarify.
9809 (const attribute): Likewise.
9810 (pure attribute): Likewise.
9812 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9815 * config/i386/i386.c (ix86_expand_vector_set): Use
9816 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9818 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9820 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9821 instructions when changing rounding bits to preserve precision bits
9822 in the x87 control word.
9824 2018-04-03 Martin Liska <mliska@suse.cz>
9826 PR tree-optimization/82491
9827 * rtl.h (strip_offset_and_add): Replace += suboffset with
9828 poly_uint64 () + suboffset.
9830 2018-03-29 Martin Liska <mliska@suse.cz>
9831 Martin Jambor <mjambor@suse.cz>
9834 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9835 param_type is not an integral or pointer type.
9837 2018-04-03 Richard Biener <rguenther@suse.de>
9839 * sese.h (recompute_all_dominators): Remove.
9841 2018-04-02 Martin Sebor <msebor@redhat.com>
9843 * doc/invoke.texi (-Wrestrict): Fix typos.
9845 2018-04-02 Jim Wilson <jimw@sifive.com>
9847 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9848 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9849 (<optab>di3, <optab>si3_extend): Likewise.
9850 (<optab>si3_mask, <optab>si3_mask_1): New.
9851 (<optab>di3_mask, <optab>di3_mask_1): New.
9852 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9853 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9854 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9856 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9858 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9861 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9863 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9864 (nds32_canonicalize_comparison): New function.
9866 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9867 Kito Cheng <kito.cheng@gmail.com>
9868 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9870 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9871 * config/nds32/constants.md (unspec_volatile_element): Add
9872 UNSPEC_VOLATILE_RELAX_GROUP.
9873 * config/nds32/nds32-relax-opt.c: New file.
9874 * config/nds32/nds32-predicates.c
9875 (nds32_symbol_load_store_p): New function.
9876 * config/nds32/nds32-protos.h
9877 (nds32_symbol_load_store_p): Declare function.
9878 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9879 * config/nds32/nds32.c
9880 (nds32_register_pass): New function to register pass.
9881 (nds32_register_passes): New function to register passes.
9882 * config/nds32/nds32.md (relax_group): New pattern.
9883 * config/nds32/nds32.opt (mrelax-hint): New option.
9884 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9886 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9888 * config/nds32/t-nds32: Modify files dependency.
9890 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9892 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9893 (PROFILE_HOOK): Define its implementation.
9895 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9897 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9898 type and 32-bit size.
9900 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9903 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9904 (V_128_256): New mode iterator.
9905 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9906 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9907 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9909 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9910 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9912 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9915 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9916 NaN inputs correctly.
9918 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9921 * config/rs6000/vsx.md (??r): New mode attribute.
9922 (*vsx_mov<mode>_64bit): Use it.
9923 (*vsx_mov<mode>_32bit): Likewise.
9925 2018-03-30 Martin Sebor <msebor@redhat.com>
9927 PR tree-optimization/84818
9928 * builtins.c (check_access): Use warning_n.
9930 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9933 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9935 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9938 2018-03-30 Julia Koval <julia.koval@intel.com>
9941 * x86-tune.def (movx, partial_reg_dependency): Enable for
9944 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9947 * lra-constraints.c (process_alt_operands): Move setting
9948 this_alternative_matches below.
9950 2018-03-29 Martin Liska <mliska@suse.cz>
9953 * doc/invoke.texi: Document how LTO works with debug info.
9954 Describe auto-load support of binutils. Mention 'x86-64'
9955 as valid option value of -march option.
9957 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9959 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9962 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9963 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9964 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9967 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9970 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9971 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9972 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9973 for RS6000_BTM_POWERPC64.
9974 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9975 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9976 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9981 2018-03-28 Carl Love <cel@us.ibm.com>
9984 2017-09-27 Carl Love <cel@us.ibm.com>
9986 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9987 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9988 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9991 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9993 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9994 instead of __vector bool.
9995 (_mm_max_pu8): Likewise.
9996 (_mm_min_pi16): Likewise.
9998 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10001 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10002 (DIVWEUO): Likewise.
10003 (DIVDEO): Likewise.
10004 (DIVDEUO): Likewise.
10005 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10006 DIVWEUO and DIVDEUO.
10007 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10008 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10009 (div_extend): Likewise.
10010 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10012 (__builtin_divweuo): Likewise.
10013 (__builtin_divdeo): Likewise.
10014 (__builtin_divdeuo): Likewise.
10016 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10019 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10020 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10022 PR tree-optimization/82004
10023 * gimple-match-head.c (optimize_pow_to_exp): New function.
10024 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10025 Don't fold to exp if optimize_pow_to_exp is false.
10027 2018-03-28 Martin Liska <mliska@suse.cz>
10030 * calls.c (initialize_argument_information): Fix trailing space.
10031 * common.opt: Fix typo and provide better explanation for
10032 -fsanitize-coverage option.
10033 * config/i386/i386.opt: Fix typo.
10035 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10036 Martin Liska <mliska@suse.cz>
10039 * gimplify.c (asan_poison_variable): Don't do the check for
10040 gimplify_omp_ctxp here.
10041 (gimplify_decl_expr): Do it here.
10042 (gimplify_target_expr): Likewise.
10044 2018-03-28 Martin Liska <mliska@suse.cz>
10047 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10048 chkp_type_bounds_count if MPX is not enabled.
10050 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10052 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10054 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10057 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10058 function to create the function decl for complex long double
10059 multiply and divide for -mabi=ieeelongdouble.
10060 (init_float128_ieee): Call it.
10062 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10065 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10066 -fcf-protection=branch -mibt.
10067 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10069 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10072 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10073 arm_word_relocations.
10075 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10078 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10079 extern array declarations.
10081 2018-03-27 Richard Biener <rguenther@suse.de>
10083 PR middle-end/84067
10084 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10085 explicit single_use checks.
10087 2018-03-27 Richard Biener <rguenther@suse.de>
10089 PR tree-optimization/85082
10090 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10093 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10095 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10096 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10097 Turn on fasynchronous-unwind-tables and funwind-tables.
10099 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10102 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10103 (*bmi_blsr_<mode>_ccz): Ditto.
10105 2018-03-26 Tom de Vries <tom@codesourcery.com>
10107 PR tree-optimization/85063
10108 * omp-general.c (offloading_function_p): New function. Factor out
10110 * omp-offload.c (pass_omp_target_link::gate): ... here.
10111 * omp-general.h (offloading_function_p): Declare.
10112 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10113 with attribute omp declare target for offloading functions.
10115 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10117 PR tree-optimization/84005
10118 * tree-data-ref.h (get_base_for_alignment): Declare.
10119 * tree-data-ref.c (get_base_for_alignment_1): New function.
10120 (get_base_for_alignment): Likewise.
10121 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10122 get_base_for_alignment to find a suitable base object, instead
10123 of always using drb->base_address.
10125 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10127 PR inline-asm/85022
10128 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10129 known size by default.
10131 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10133 PR inline-asm/85030
10134 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10135 and non BLKmode operands.
10137 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10140 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10141 Clean up attributes.
10143 2018-03-23 Richard Biener <rguenther@suse.de>
10146 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10147 we are going to emit early debug for LTO.
10149 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10151 PR inline-asm/85034
10152 * function.c (match_asm_constraints_1): Don't optimize if input
10153 doesn't satisfy general_operand predicate for output's mode.
10155 PR inline-asm/85022
10156 * alias.c (write_dependence_p): Don't require for x_canonicalized
10157 non-VOIDmode if x has VOIDmode.
10160 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10161 just don't try to optimize it rather than assert it never happens.
10163 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10165 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10166 macro expansions for definition of ST_INTERNAL_<mode> and
10167 LD_INTERNAL_<mode> builtins.
10168 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10170 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10172 (altivec_expand_st_builtin): Likewise.
10173 (altivec_expand_builtin): Remove calls to deleted functions.
10174 (rs6000_address_for_altivec): Delete this function.
10175 * config/rs6000/vector.md: Remove expands for
10176 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10178 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10181 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10182 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10183 re-computing once computed.
10184 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10185 (arm_init_machine_status): Initialize
10186 machine->static_chain_stack_bytes.
10188 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10191 * doc/extend.texi: Add four new prototypes for vec_ld.
10192 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10193 definitions for more logical presentation.
10194 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10195 entries for V1TI variants of __builtin_altivec_ld builtin.
10196 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10197 handling of V1TI variant of LVX icode pattern.
10198 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10199 (rs6000_gimple_fold_builtin): Likewise.
10200 (altivec_init_builtins): Add code to define
10201 __builtin_altivec_lvx_v1ti function.
10203 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10205 PR inline-asm/84941
10206 * function.c (match_asm_constraints_1): Don't do the optimization
10207 if input isn't a REG, SUBREG, MEM or constant.
10209 2018-03-22 Tom de Vries <tom@codesourcery.com>
10211 PR tree-optimization/84956
10212 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10213 bb_has_abnormal_pred.
10215 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10218 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10219 DECL_INITIAL (decl) to decl at the end.
10220 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10221 adjust the comment.
10223 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10225 * doc/extend.texi (__builtin_tgmath): Document when complex
10226 integer types are treated as _Complex _Float64.
10228 2018-03-21 Tom de Vries <tom@codesourcery.com>
10230 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10232 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10234 PR tree-optimization/84960
10235 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10236 if it is ENTRY block, move them into single succ of ENTRY in that case.
10238 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10240 PR tree-optimization/84811
10241 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10242 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10243 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10244 (known_subrange_p): Update accordingly. Cast each value involved
10245 in the size comparison, rather than casting the result of the
10248 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10250 PR tree-optimization/84982
10251 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10252 by flipping the least significant bit rather than all bits from
10253 bitpos to bitpos + bitsize - 1.
10255 2018-03-21 Nathan Sidwell <nathan@acm.org>
10257 * doc/extend.texi (Deprecated Features): Remove mention of
10258 long-deleted deprecations.
10260 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10263 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10264 * configure: Regenerate.
10266 2018-03-21 Tom de Vries <tom@codesourcery.com>
10268 PR tree-optimization/83126
10269 * tree-parloops.c (num_phis): New function.
10270 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10272 2018-03-21 Nathan Sidwell <nathan@acm.org>
10274 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10275 mention anon-struct/union members and trailing attributes.
10277 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10279 PR tree-optimization/84969
10280 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10281 builtin memset partitions if they set different rhs values.
10283 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10285 PR rtl-optimization/84989
10286 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10287 VEC_DUPLICATE with scalar result mode.
10289 2018-03-21 Martin Liska <mliska@suse.cz>
10292 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10293 not intended return statement.
10295 2018-03-21 Martin Liska <mliska@suse.cz>
10298 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10299 (chkp_find_bound_slots_1): Limit number of iterations.
10301 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10304 * Minor grammar fixes for x86 options.
10306 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10309 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10310 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10312 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10315 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10316 (altivec_lvx_<mode>_1op): Likewise.
10317 (altivec_stvx_<mode>_2op): Likewise.
10318 (altivec_stvx_<mode>_1op): Likewise.
10319 (altivec_lvx_<VM2:mode>): New define_expand.
10320 (altivec_stvx_<VM2:mode>): Likewise.
10321 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10322 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10323 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10324 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10325 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10326 (rs6000_gen_lvx): Likewise.
10327 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10328 (altivec_expand_stv_builtin): Likewise.
10329 (altivec_expand_builtin): Likewise.
10330 * config/rs6000/vector.md: Likewise.
10332 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10335 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10338 2018-03-20 Richard Biener <rguenther@suse.de>
10341 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10342 sign-conversions as zero, fall back to standard scalar_stmt
10345 2018-03-20 Martin Liska <mliska@suse.cz>
10348 * predict.c (rebuild_frequencies): Handle case when we have
10349 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10351 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10354 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10355 flag_section_anchors.
10356 * varasm.c (use_blocks_for_decl_p): Remove hack for
10357 dw2_force_const_mem.
10360 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10362 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10363 be created, use lowpart_subreg of operands[0] rather than operands[0]
10365 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10366 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10367 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10368 and n constraint instead of aarch64_shift_imm_di and Usd.
10369 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10370 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10372 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10375 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10376 to favor GPR over NEON registers.
10377 (<shift>di3_neon): Likewise.
10379 2018-03-20 Tom de Vries <tom@codesourcery.com>
10382 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10383 (nvptx_process_pars): Emit bar.sync asap and alap.
10385 2018-03-20 Tom de Vries <tom@codesourcery.com>
10388 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10389 seen_label if seen_label is already set.
10391 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10394 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10395 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10396 Use 1U instead of 1. Formatting fixes.
10399 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10400 instead of TREE_TYPE (s1) for the return value.
10402 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10404 PR tree-optimization/84946
10405 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10406 bitsize + bitsize in poly_uint64 rather than poly_int64.
10409 * dwarf2asm.c: Include fold-const.c.
10410 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10411 of decl rather than decl itself.
10413 PR rtl-optimization/84643
10414 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10416 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10419 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10420 calling assemble_variable.
10422 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10425 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10426 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10428 2018-03-19 Jim Wilson <jimw@sifive.com>
10431 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10432 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10433 (riscv_first_stack_step): Likewise.
10434 (riscv_option_override): Use STACK_BOUNDARY instead of
10435 MIN_STACK_BOUNDARY.
10436 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10437 MIN_STACK_BOUNDARY.
10438 (BIGGEST_ALIGNMENT): Set to 128.
10439 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10440 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10443 2018-03-19 Richard Biener <rguenther@suse.de>
10445 PR tree-optimization/84933
10446 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10447 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10449 2018-03-19 Richard Biener <rguenther@suse.de>
10451 PR tree-optimization/84859
10452 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10453 (cond_if_else_store_replacement): Perform sinking operation on
10454 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10455 Generalize what a BB with a single eligible store is.
10457 2018-03-19 Richard Biener <rguenther@suse.de>
10459 PR tree-optimization/84929
10460 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10461 chrec_is_positive against non-chrec arg.
10463 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10466 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10468 2018-03-18 Martin Liska <mliska@suse.cz>
10470 PR rtl-optimization/84635
10471 * regrename.c (build_def_use): Use matches_mode only when
10474 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10476 PR tree-optimization/84913
10477 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10478 vectorize chains of COND_EXPRs.
10480 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10482 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10484 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10486 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10488 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10490 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10492 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10493 Kito Cheng <kito.cheng@gmail.com>
10495 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10496 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10497 (nds32_adjust_reg_alloc_order): New function.
10498 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10500 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10502 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10503 nds32_print_operand, nds32_print_operand_address): Use
10504 HOST_WIDE_INT_PRINT_DEC instead.
10506 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10508 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10510 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10513 * config/i386/i386.c (initial_ix86_tune_features,
10514 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10515 unsigned long long.
10516 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10517 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10518 rather than 1u << ix86_tune. Formatting fix.
10519 (ix86_option_override_internal): Change ix86_arch_mask from
10520 unsigned int to unsigned HOST_WIDE_INT, initialize to
10521 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10522 (ix86_function_specific_restore): Likewise.
10524 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10527 * postreload.c (reload_combine_recognize_pattern): Perform
10528 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10529 truncate_int_for_mode the result for the destination's mode.
10532 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10533 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10535 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10538 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10541 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10542 regno_allocno_class_array and sorted_pseudos.
10543 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10544 insns where regno is used.
10546 2018-03-16 Martin Liska <mliska@suse.cz>
10549 * multiple_target.c (create_dispatcher_calls): Redirect
10550 reference in the symbol table.
10552 2018-03-16 Martin Liska <mliska@suse.cz>
10555 * multiple_target.c (create_dispatcher_calls): Redirect also
10558 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10562 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10563 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10564 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10566 2018-03-16 Julia Koval <julia.koval@intel.com>
10568 * doc/invoke.texi (Skylake Server): Add CLWB.
10569 Cannonlake): Remove CLWB.
10571 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10573 PR tree-optimization/84841
10574 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10576 (FLOAT_ONE_CONST_TYPE): Define.
10577 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10578 (sort_by_operand_rank): Put entries with higher constant_type last
10579 rather than first to match comments.
10581 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10583 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10586 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10589 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10590 check and use address of target always.
10592 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10595 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10596 (indirect_thunk_bnd_needed): Likewise.
10597 (indirect_thunks_used): Likewise.
10598 (indirect_thunks_bnd_used): Likewise.
10599 (indirect_return_needed): New.
10600 (indirect_return_bnd_needed): Likewise.
10601 (output_indirect_thunk_function): Add a bool argument for
10603 (output_indirect_thunk_function): Don't generate alias for
10604 function return thunk.
10605 (ix86_code_end): Call output_indirect_thunk_function to generate
10606 function return thunks.
10607 (ix86_output_function_return): Set indirect_return_bnd_needed
10608 and indirect_return_needed instead of indirect_thunk_bnd_needed
10609 and indirect_thunk_needed.
10611 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10613 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10614 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10615 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10617 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10618 Paul Hua <paul.hua.gm@gmail.com>
10621 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10623 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10625 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10626 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10627 resp. SFmode cases.
10629 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10632 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10633 instead of GET_MODE_SIZE when comparing Units.
10635 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10638 * varasm.c (hash_section): Return an unchangeble hash value
10639 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10640 Return !aarch64_can_use_per_function_literal_pools_p ().
10642 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10645 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10646 rather than address of cmode as last argument to prepare_cmp_insn.
10648 2018-03-15 Julia Koval <julia.koval@intel.com>
10650 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10651 F_AVX512VNNI, F_AVX512BITALG): New.
10653 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10656 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10657 insn for floating-point loads and stores.
10659 2018-03-14 Carl Love <cel@us.ibm.com>
10661 * config/rs6000/rs6000-c.c: Add macro definitions for
10662 ALTIVEC_BUILTIN_VEC_PERMXOR.
10663 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10664 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10665 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10666 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10668 * config/doc/extend.texi: Add prototypes for vec_permxor.
10670 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10673 * diagnostic-show-locus.c (class layout_point): Convert m_line
10674 from int to linenum_type.
10675 (line_span::comparator): Use linenum "compare" function when
10676 comparing line numbers.
10677 (test_line_span): New function.
10678 (layout_range::contains_point): Convert param "row" from int to
10680 (layout_range::intersects_line_p): Likewise.
10681 (layout::will_show_line_p): Likewise.
10682 (layout::print_source_line): Likewise.
10683 (layout::should_print_annotation_line_p): Likewise.
10684 (layout::print_annotation_line): Likewise.
10685 (layout::print_leading_fixits): Likewise.
10686 (layout::annotation_line_showed_range_p): Likewise.
10687 (struct line_corrections): Likewise for field m_row.
10688 (line_corrections::line_corrections): Likewise for param "row".
10689 (layout::print_trailing_fixits): Likewise.
10690 (layout::get_state_at_point): Likewise.
10691 (layout::get_x_bound_for_row): Likewise.
10692 (layout::print_line): Likewise.
10693 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10694 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10695 * input.c (selftest::test_linenum_comparisons): New function.
10696 (selftest::input_c_tests): Call it.
10697 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10698 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10699 * selftest.h (ASSERT_GT): New macro.
10700 (ASSERT_GT_AT): New macro.
10701 (ASSERT_LT): New macro.
10702 (ASSERT_LT_AT): New macro.
10704 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10706 PR rtl-optimization/84780
10707 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10709 2018-03-14 Martin Liska <mliska@suse.cz>
10711 * tree.c (record_node_allocation_statistics): Use
10712 get_stats_node_kind.
10713 (get_stats_node_kind): New function extracted from
10714 record_node_allocation_statistics.
10715 (free_node): Use get_stats_node_kind.
10717 2018-03-14 Richard Biener <rguenther@suse.de>
10719 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10720 that the value-set of ANTIC_IN doesn't grow.
10723 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10725 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10726 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10728 2018-03-14 Julia Koval <julia.koval@intel.com>
10730 * config.gcc (icelake-client, icelake-server): New.
10732 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10733 (initial_ix86_arch_features): Ditto.
10734 (PTA_SKYLAKE): Add SGX.
10735 (PTA_ICELAKE): Remove.
10736 (PTA_ICELAKE_CLIENT): New.
10737 (PTA_ICELAKE_SERVER): New.
10738 (ix86_option_override_internal): Split up icelake on icelake client and
10740 (get_builtin_code_for_version): Ditto.
10741 (fold_builtin_cpu): Ditto.
10742 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10743 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10744 * config/i386/i386.h (processor_type): Ditto.
10745 * doc/invoke.texi: Ditto.
10747 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10750 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10751 INTEGER_CST offset, add it together with bitpos / 8 and
10752 sign extend based on POINTER_SIZE.
10756 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10759 * config/i386/constraints.md (Yc): New register constraint.
10760 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10761 Use Yc constraint for alternative 2 of operand 0. Remove
10762 preferred_for_speed attribute.
10764 2018-03-14 Richard Biener <rguenther@suse.de>
10766 PR tree-optimization/84830
10767 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10768 with the old one to avoid oscillations.
10770 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10773 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10775 (assign_by_spills): Return a flag of reload assignment failure.
10776 Do not process the reload assignment failures. Do not spill other
10777 reload pseudos if they has the same reg class. Update n if
10779 (lra_assign): Add a return arg. Set up from the result of
10780 assign_by_spills call.
10781 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10782 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10783 usage_insns if it is not NULL.
10784 (spill_hard_reg_in_range): New function.
10785 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10786 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10787 function prototypes.
10788 (lra_assign): Change prototype.
10789 * lra.c (lra): Add code to deal with fails by splitting hard reg
10792 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10794 * config/riscv/riscv.opt (mrelax): New option.
10795 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10796 "norelax" when riscv_mrelax is disabled.
10797 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10799 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10802 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10803 reassociation for int modes.
10805 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10807 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10808 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10810 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10811 * config/aarch64/aarch64-sve.md
10812 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10813 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10814 (*extend<mode><Vwide>2): Rename to...
10815 (aarch64_sve_extend<mode><Vwide>2): ...this.
10816 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10817 renaming the old pattern to...
10818 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10820 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10821 define_expand, renaming the old pattern to...
10822 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10823 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10824 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10825 account when deciding which SVE instruction the optab should use.
10826 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10828 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10830 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10831 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10832 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10833 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10834 (tlsdesc_small_<mode>): Turn a define_expand and use
10835 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10836 (tlsdesc_small_advsimd_<mode>): ...this.
10837 (tlsdesc_small_sve_<mode>): New pattern.
10839 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10841 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10842 (UNSPEC_UMUL_HIGHPART): New constants.
10843 (MUL_HIGHPART): New int iteraor.
10844 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10845 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10847 (*<su>mul<mode>3_highpart): New define_insn.
10849 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10852 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10855 2018-03-13 Martin Liska <mliska@suse.cz>
10858 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10860 (sem_item_optimizer::~sem_item_optimizer): Release it.
10861 (sem_item_optimizer::merge_classes): Register variable aliases.
10862 (sem_item_optimizer::fixup_pt_set): New function.
10863 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10864 * ipa-icf.h: Declare new variables and functions.
10866 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10868 PR middle-end/84834
10869 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10870 integer_pow2p@2 and test integer_pow2p in condition.
10871 (A < 0 ? C : 0): Similarly for @1.
10873 PR middle-end/84831
10874 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10875 characters starting at p contain '\0' character, don't look beyond
10879 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10880 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10883 * reg-stack.c (change_stack): Change update_end var from int to
10884 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10885 also call set_block_for_insn on the newly added insns and rescan.
10888 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10889 on the last operand.
10892 * tree.c (stabilize_reference_1): Return save_expr (e) for
10893 STATEMENT_LIST even if it doesn't have side-effects.
10895 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10897 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10899 2018-03-12 Renlin Li <renlin.li@arm.com>
10901 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10902 aarch64_output_scalar_simd_mov_immediate.
10904 2018-03-12 Martin Sebor <msebor@redhat.com>
10906 PR tree-optimization/83456
10907 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10908 for perfectly overlapping calls to memcpy.
10909 (gimple_fold_builtin_memory_chk): Same.
10910 (gimple_fold_builtin_strcpy): Handle no-warning.
10911 (gimple_fold_builtin_stxcpy_chk): Same.
10912 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10914 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10916 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10917 parameter. Use it for SFmode.
10918 (rs6000_function_arg_advance_1): Adjust.
10919 (rs6000_function_arg): Adjust.
10920 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10922 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10924 PR rtl-optimization/84169
10925 PR rtl-optimization/84780
10926 * combine.c (can_combine_p): Check for a 2-insn combination whether
10927 the destination register is used between the two insns, too.
10929 2018-03-12 Richard Biener <rguenther@suse.de>
10931 PR tree-optimization/84803
10932 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10933 for refs DR analysis didn't process.
10935 2018-03-12 Richard Biener <rguenther@suse.de>
10937 PR tree-optimization/84777
10938 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10939 force-vectorize loops ignore whether we are optimizing for size.
10941 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10943 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10944 (TARGET_MD_ASM_ADJUST): Define.
10946 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10947 Kito Cheng <kito.cheng@gmail.com>
10948 Chung-Ju Wu <jasonwucj@gmail.com>
10950 * config/nds32/nds32.c (nds32_compute_stack_frame,
10951 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10952 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10953 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10954 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10955 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10956 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10957 * config/nds32/nds32.md (prologue, epilogue): Use macro
10958 NDS32_V3PUSH_AVAILABLE_P to do checking.
10960 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10963 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10964 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10965 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10966 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10967 addition of most attributes on !orig_type_die or the attribute not
10968 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10970 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10971 Chung-Ju Wu <jasonwucj@gmail.com>
10973 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10974 __NDS32_VH__ macro.
10975 * config/nds32/nds32.opt (mvh): New option.
10977 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10978 Chung-Ju Wu <jasonwucj@gmail.com>
10980 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10982 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10983 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10986 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10987 Chung-Ju Wu <jasonwucj@gmail.com>
10989 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10991 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10992 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10994 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10995 Kito Cheng <kito.cheng@gmail.com>
10996 Chung-Ju Wu <jasonwucj@gmail.com>
10998 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10999 UNSPEC_FFMISM and UNSPEC_FLMISM.
11000 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11001 for ffb, ffmism and flmism.
11002 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11003 (unspec_ffmism): Ditto.
11004 (unspec_flmism): Ditto.
11005 (nds32_expand_builtin_impl): Check if string extension is available.
11006 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11007 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11009 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11012 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11015 * lra-assigns.c (assign_by_spills): Return a flag of reload
11016 assignment failure. Do not process the reload assignment
11017 failures. Do not spill other reload pseudos if they has the same
11019 (lra_assign): Add a return arg. Set up from the result of
11020 assign_by_spills call.
11021 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11022 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11023 usage_insns if it is not NULL.
11024 (spill_hard_reg_in_range): New function.
11025 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11026 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11027 function prototypes.
11028 (lra_assign): Change prototype.
11029 * lra.c (lra): Add code to deal with fails by splitting hard reg
11032 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11035 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11037 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11040 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11041 (dw_val_node): Add val_symbolic_view.
11042 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11043 (symview_upper_bound): New.
11044 (new_line_info_table): Initialize symviews_since_reset.
11045 (dwarf2out_source_line): Count symviews_since_reset and set
11046 symview_upper_bound.
11047 (dw_val_equal_p): Handle symview.
11048 (add_AT_symview): New.
11049 (print_dw_val): Handle symview.
11050 (attr_checksum, attr_checksum_ordered): Likewise.
11051 (same_dw_val_p, size_of_die): Likewise.
11052 (value_format, output_die): Likewise.
11053 (add_high_low_attributes): Use add_AT_symview for entry_view.
11054 (dwarf2out_finish): Reset symview_upper_bound, clear
11057 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11060 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11061 Add strict argument and use it.
11062 (rs6000_split_multireg_move): Update for new strict argument.
11063 (mem_operand_gpr): Disallow all non-offsettable addresses.
11064 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11066 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11069 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11070 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11071 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11074 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11075 decl, use remap_type if we want to use the type.
11077 2018-03-09 Martin Sebor <msebor@redhat.com>
11079 PR tree-optimization/84526
11080 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11082 (builtin_access::generic_overlap): Be prepared to handle non-array
11085 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11087 PR rtl-optimization/84682
11088 * lra-constraints.c (process_address_1): Check is_address flag
11089 for address constraints.
11090 (process_alt_operands): Likewise.
11091 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11092 preprocess_constraints.
11093 * recog.h (preprocess_constraints): Add oploc parameter.
11095 * recog.c (preprocess_constraints): Test address_operand for
11096 CT_ADDRESS constraints.
11098 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11101 * lra-assigns.c (assign_by_spills): Return a flag of reload
11102 assignment failure. Do not process the reload assignment
11103 failures. Do not spill other reload pseudos if they has the same
11105 (lra_assign): Add a return arg. Set up from the result of
11106 assign_by_spills call.
11107 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11108 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11109 usage_insns if it is not NULL.
11110 (spill_hard_reg_in_range): New function.
11111 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11112 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11113 function prototypes.
11114 (lra_assign): Change prototype.
11115 * lra.c (lra): Add code to deal with fails by splitting hard reg
11118 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11121 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11122 Accept complain bool parameter. Only emit errors if it is true.
11123 (arm_parse_cpu_option_name): Likewise.
11124 (arm_target_thumb_only): Adjust callers of the above.
11125 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11126 prototype to take a default true bool parameter.
11127 (arm_parse_arch_option_name): Likewise.
11129 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11130 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11134 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11135 * configure: Regenerate.
11136 * configure.ac ("linker --version-script option"): New.
11137 ("linker soname option"): New.
11139 2018-03-09 Richard Biener <rguenther@suse.de>
11141 PR tree-optimization/84775
11142 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11143 immediate uses of predicate stmts and mark them modified.
11146 PR tree-optimization/84178
11147 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11149 (version_loop_for_if_conversion): Delay update_ssa call.
11150 (tree_if_conversion): Delay update_ssa until after predicate
11153 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11156 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11157 when the function accesses prior frames.
11159 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11162 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11163 gen_llsym, otherwise call maybe_gen_llsym.
11165 PR inline-asm/84742
11166 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11167 has ',' character inside of it.
11169 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11172 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11173 as clobbering CC_REGNUM.
11175 2018-03-08 Richard Biener <rguenther@suse.de>
11177 PR middle-end/84552
11178 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11179 (follow_copies_to_constant): Do not follow SSA names registered
11182 2018-03-08 Richard Biener <rguenther@suse.de>
11184 PR tree-optimization/84178
11185 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11187 (version_loop_for_if_conversion): Delay update_ssa call.
11188 (tree_if_conversion): Delay update_ssa until after predicate
11191 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11193 PR tree-optimization/84178
11194 * tree-if-conv.c (release_bb_predicate): Remove the
11195 the assertion that the stmts have NULL use_ops.
11196 Discard the statements, asserting that they haven't
11197 yet been added to a BB.
11199 2018-03-08 Richard Biener <rguenther@suse.de>
11201 PR tree-optimization/84746
11202 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11203 (phi_translate): Pass in destination ANTIC_OUT set.
11204 (phi_translate_1): Likewise. For a simplified result lookup
11205 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11206 (phi_translate_set): Adjust.
11207 (do_pre_regular_insertion): Likewise.
11208 (do_pre_partial_partial_insertion): Likewise.
11210 2018-03-08 Martin Liska <mliska@suse.cz>
11212 PR gcov-profile/84735
11213 * doc/gcov.texi: Document usage of profile files.
11214 * gcov-io.h: Document changes in the format.
11216 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11220 * dwarf2out.c (struct dw_line_info_table): Update comments for
11222 (FORCE_RESET_NEXT_VIEW): New.
11223 (FORCE_RESETTING_VIEW_P): New.
11224 (RESETTING_VIEW_P): Check for -1 too.
11225 (ZERO_VIEW_P): Likewise.
11226 (new_line_info_table): Force-reset next view.
11227 (dwarf2out_begin_function): Likewise.
11228 (dwarf2out_source_line): Simplify zero_view_p initialization.
11229 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11230 view directly. Omit view when omitting .loc at line 0.
11232 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11234 PR tree-optimization/84740
11235 * tree-switch-conversion.c (process_switch): Call build_constructors
11236 only if info.phi_count is non-zero.
11238 PR tree-optimization/84739
11239 * tree-tailcall.c (find_tail_calls): Check call arguments against
11240 DECL_ARGUMENTS (current_function_decl) rather than
11241 DECL_ARGUMENTS (func) when checking for tail recursion.
11243 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11245 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11246 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11247 Volker Reichelt's entry and add entries for people that perform
11248 GCC fuzzy testing and report numerous bugs.
11250 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11253 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11254 readonly data in sdata, if that is disabled.
11255 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11256 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11257 -mreadonly-in-sdata option.
11259 2018-03-07 Martin Sebor <msebor@redhat.com>
11261 PR tree-optimization/84468
11262 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11263 basic block when looking for nul assignment.
11265 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11268 * except.h (output_function_exception_table): Adjust prototype.
11269 * except.c (output_function_exception_table): Remove FNNAME parameter
11270 and add SECTION parameter. Ouput one part of the table at a time.
11271 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11272 the first part of the exception table and emit unwind directives.
11273 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11274 (i386_pe_seh_cold_init): Likewise.
11275 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11276 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11277 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11278 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11279 * config/i386/winnt.c: Include except.h.
11280 (struct seh_frame_state): Add reg_offset, after_prologue and
11281 in_cold_section fields.
11282 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11283 (i386_pe_seh_cold_init): New function.
11284 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11285 to seh->in_cold_section.
11286 (seh_emit_push): Record the offset of the push.
11287 (seh_emit_save): Record the offet of the save.
11288 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11289 Test seh->after_prologue to disregard the epilogue.
11290 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11291 (i386_pe_end_cold_function): New function.
11293 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11296 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11297 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11300 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11302 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11303 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11305 PR middle-end/84723
11306 * multiple_target.c: Include tree-inline.h and intl.h.
11307 (expand_target_clones): Diagnose and fail if node->definition and
11308 !tree_versionable_function_p (node->decl).
11310 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11312 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11314 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11315 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11316 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11318 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11321 * combine.c (try_combine): Use reg_or_subregno instead of handling
11322 just paradoxical SUBREGs and REGs.
11324 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11326 * config/arc/arc.c (arc_finalize_pic): Remove function.
11327 (arc_must_save_register): We use single base PIC register, remove
11328 checks to save/restore the PIC register.
11329 (arc_expand_prologue): Likewise.
11330 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11332 (arc_verify_short): Likewise.
11333 (arc_attr_type): Likewise.
11334 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11335 (walk_stores): Likewise.
11336 (arc_address_cost): Make it static.
11337 (arc_verify_short): Likewise.
11338 (branch_dest): Likewise.
11339 (arc_attr_type): Likewise.
11340 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11341 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11342 (arc_final_prescan_insn): Remove inserting the nops due to
11343 hardware hazards. It is done in reorg step.
11344 (insn_length_variant_t): Remove.
11345 (insn_length_parameters_t): Likewise.
11346 (arc_insn_length_parameters): Likewise.
11347 (arc_get_insn_variants): Likewise.
11348 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11350 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11352 PR inline-asm/84683
11353 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11356 PR tree-optimization/84687
11357 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11359 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11361 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11363 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11364 Rename to ppc_speculation_barrier.
11365 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11366 __builtin_ppc_speculation_barrier.
11368 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11371 * combine.c (combine_simplify_rtx): Don't try to simplify if
11372 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11375 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11377 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11378 to 32 bytes when compiling for POWER9.
11380 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11383 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11384 regparm >= 3 with no arg reg available also for calls with
11385 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11388 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11390 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11392 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11395 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11397 2018-03-05 Richard Biener <rguenther@suse.de>
11399 PR tree-optimization/84486
11400 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11401 When inserting a __builtin_assume_aligned call set the LHS
11402 SSA name alignment info accordingly.
11404 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11406 PR tree-optimization/84114
11407 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11408 Avoid reassociation of FLOAT_MODE addition.
11410 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11412 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11413 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11414 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11415 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11416 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11417 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11418 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11420 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11421 __builtin_ia32_wbinvd): New builtins.
11422 (SPECIAL_ARGS2): New.
11423 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11424 (SPECIAL_ARGS2): New.
11425 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11426 (ix86_valid_target_attribute_inner_p): Ditto.
11427 (ix86_init_mmx_sse_builtins): Add special_args2.
11428 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11429 TARGET_WBNOINVD_P): New.
11430 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11431 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11432 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11433 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11434 * config/i386/pconfigintrin.h: New file.
11435 * config/i386/wbnoinvdintrin.h: Ditto.
11436 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11438 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11440 2018-03-05 Richard Biener <rguenther@suse.de>
11442 PR tree-optimization/84670
11443 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11445 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11446 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11447 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11448 doesn't grow if all successors (recursively) were visited at least
11451 2018-03-05 Richard Biener <rguenther@suse.de>
11453 PR tree-optimization/84650
11454 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11455 if executed in the loop pipeline.
11457 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11459 * doc/configfiles.texi (Configuration Files): Move info about
11460 conditionalizing $target-protos.h to...
11461 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11462 differs from $target-protos.h.
11464 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11465 Chung-Ju Wu <jasonwucj@gmail.com>
11467 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11468 * config/nds32/nds32-multiple.md (setmemsi): Define.
11469 * config/nds32/nds32-memory-manipulation.c
11470 (nds32_gen_dup_4_byte_to_word_value): New.
11471 (emit_setmem_word_loop): New.
11472 (emit_setmem_byte_loop): New.
11473 (nds32_expand_setmem_loop): New.
11474 (nds32_expand_setmem_loop_v3m): New.
11475 (nds32_expand_setmem_unroll): New.
11476 (nds32_expand_setmem): New.
11478 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11479 Chung-Ju Wu <jasonwucj@gmail.com>
11481 * config/nds32/nds32-memory-manipulation.c
11482 (nds32_emit_load_store): New.
11483 (nds32_emit_post_inc_load_store): New.
11484 (nds32_emit_mem_move): New.
11485 (nds32_emit_mem_move_block): New.
11486 (nds32_expand_movmemsi_loop_unknown_size): New.
11487 (nds32_expand_movmemsi_loop_known_size): New.
11488 (nds32_expand_movmemsi_loop): New.
11489 (nds32_expand_movmemsi_unroll): New.
11490 (nds32_expand_movmemqi): Rename ...
11491 (nds32_expand_movmemsi): ... to this.
11492 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11493 (movmemsi): ... to this.
11494 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11495 (nds32_expand_movmemsi): ... to this.
11497 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11498 Monk Chiang <sh.chiang04@gmail.com>
11499 Chung-Ju Wu <jasonwucj@gmail.com>
11501 * config/nds32/nds32-protos.h
11502 (nds32_expand_load_multiple): New arguments.
11503 (nds32_expand_store_multiple): Ditto.
11504 (nds32_valid_multiple_load_store): Rename ...
11505 (nds32_valid_multiple_load_store_p): ... to this.
11506 * config/nds32/nds32-memory-manipulation.c
11507 (nds32_expand_load_multiple): Refine implementation.
11508 (nds32_expand_store_multiple): Ditto.
11509 * config/nds32/nds32-multiple.md
11510 (load_multiple): Update nds32_expand_load_multiple interface.
11511 (store_multiple): Update nds32_expand_store_multiple interface.
11512 * config/nds32/nds32-predicates.c
11513 (nds32_valid_multiple_load_store): Rename ...
11514 (nds32_valid_multiple_load_store_p): ... to this and refine
11516 * config/nds32/predicates.md
11517 (nds32_load_multiple_and_update_address_operation): New predicate.
11518 (nds32_store_multiple_and_update_address_operation): New predicate.
11520 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11521 Chung-Ju Wu <jasonwucj@gmail.com>
11523 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11524 (combo): New attribute.
11525 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11527 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11529 * config/nds32/nds32.opt: Change -mcmodel= default value.
11531 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11532 Monk Chiang <sh.chiang04@gmail.com>
11533 Chung-Ju Wu <jasonwucj@gmail.com>
11535 * config/nds32/constants.md (unspec_element): New enum.
11536 * config/nds32/constraints.md (Umw): New constraint.
11537 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11538 * config/nds32/nds32-intrinsic.md: Likewise.
11539 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11540 (nds32_valid_smw_lwm_base_p): New.
11541 (nds32_output_smw_single_word): New.
11542 (nds32_output_lmw_single_word): New.
11543 (nds32_expand_unaligned_load): New.
11544 (nds32_expand_unaligned_store): New.
11545 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11546 (nds32_output_smw_single_word): Declare.
11547 (nds32_output_lmw_single_word): Declare.
11548 (nds32_expand_unaligned_load): Declare.
11549 (nds32_expand_unaligned_store): Declare.
11550 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11551 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11552 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11553 NDS32_BUILTIN_UASTORE_DW.
11554 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11557 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11558 Kito Cheng <kito.cheng@gmail.com>
11559 Chung-Ju Wu <jasonwucj@gmail.com>
11561 * config/nds32/nds32-intrinsic.c
11562 (nds32_expand_builtin_null_ftype_reg): Delete.
11563 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11564 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11565 (nds32_read_argument): New.
11566 (nds32_legitimize_target): Ditto.
11567 (nds32_legitimize_argument): Ditto.
11568 (nds32_check_constant_argument): Ditto.
11569 (nds32_expand_unop_builtin): Ditto.
11570 (nds32_expand_unopimm_builtin): Ditto.
11571 (nds32_expand_binop_builtin): Ditto.
11572 (nds32_builtin_decl_impl): Ditto.
11573 (builtin_description): Ditto.
11574 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11575 (nds32_init_builtins_impl): Ditto.
11576 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11577 (nds32_builtin_decl): New.
11578 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11579 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11581 2018-03-02 Jeff Law <law@redhat.com>
11583 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11584 (redundant_insn, fill_simple_delay_slots): Likewise.
11585 (fill_slots_from_thread): Likewise.
11586 * resource.c (mark_referenced_resources): Likewise.
11587 (mark_set_resources, find_dead_or_set_registers): Likewise.
11589 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11591 * substring-locations.h (format_warning_va): Formatting fix for
11592 ATTRIBUTE_GCC_DIAG.
11593 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11595 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11596 * substring-locations.c: Include intl.h.
11597 (format_warning_va): Turned into small wrapper around
11598 format_warning_n_va, renamed to ...
11599 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11600 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11602 (format_warning_at_substring_n): New function.
11603 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11604 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11605 format_warning_at_substring with just a shorter name instead of
11606 const function pointer.
11607 (fmtwarn_n): New function.
11608 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11609 appropriate, get rid of all the fmtstr temporaries, move conditionals
11610 with G_() wrapped string literals directly into fmtwarn arguments,
11611 cast dir.len to (int), formatting fixes.
11613 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11615 * doc/invoke.texi: Remove "Cilk Plus" references.
11617 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11618 Richard Biener <rguenther@suse.de>
11621 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11622 for error or warning attributes if CALL_FROM_THUNK_P is set.
11625 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11628 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11629 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11632 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11633 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11635 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11636 instead of -1U in last predictors element's probability member.
11638 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11641 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11642 arguments if they are comparable.
11644 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11646 PR tree-optimization/84634
11647 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11648 masks and masked_loop_p with a single loop_masks, making sure it's
11649 null for bb vectorization.
11651 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11653 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11654 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11655 loop->force_vectorize to check whether there is no alias.
11657 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11660 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11662 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11664 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11665 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11666 instead of a loop around prev_real_insn.
11667 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11670 PR inline-asm/84625
11671 * config/i386/i386.c (ix86_print_operand): Use conditional
11672 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11675 2018-03-02 Richard Biener <rguenther@suse.de>
11677 PR tree-optimization/84427
11678 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11679 (bitmap_set_subtract_values): Rewrite to handle multiple
11682 (prune_clobbered_mems): Likewise.
11683 (phi_translate): Take edge instead of pred/phiblock.
11684 (phi_translate_1): Likewise.
11685 (phi_translate_set): Likewise. Insert all translated
11686 exprs for a value into the set, keeping possibly multiple
11687 expressions per value.
11688 (compute_antic_aux): Adjust for phi_translate changes.
11689 When intersecting union the expressions and prune those
11690 not in the final value set, keeping possibly multiple
11691 expressions per value. Do not use value-insertion
11692 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11693 all expressions. Add verification that the value-sets
11694 only shrink during iteration.
11695 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11696 (do_pre_regular_insertion): Likewise.
11697 (do_pre_partial_partial_insertion): Likewise.
11699 2018-03-02 Richard Biener <rguenther@suse.de>
11702 * config/darwin.c (saved_debug_info_level): New static global.
11703 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11704 (darwin_asm_lto_end): Restore debug info generation settings.
11706 2018-03-01 Martin Liska <mliska@suse.cz>
11709 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11710 volatile arguments.
11712 2018-03-01 Richard Biener <rguenther@suse.de>
11715 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11716 type annotation with LTO when debug was disabled at compile-time.
11718 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11720 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11722 (mips_final_postscan_insn): Likewise.
11724 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11726 PR rtl-optimization/84528
11727 * alias.c (init_alias_target): Add commentary.
11728 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11729 a unique base value if the frame pointer is not eliminated
11730 to the stack pointer.
11732 2018-03-01 Tom de Vries <tom@codesourcery.com>
11734 PR rtl-optimization/83327
11735 * lra-int.h (hard_regs_spilled_into): Declare.
11736 * lra.c (hard_regs_spilled_into): Define.
11737 (init_reg_info): Init hard_regs_spilled_into.
11738 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11739 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11740 (process_bb_lives): Handle hard_regs_spilled_into.
11741 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11742 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11744 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11746 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11747 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11748 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11749 * config/rs6000/aix72.h: New file.
11751 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11753 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11754 instead of warning_at with conditional singular and plural messages
11758 * stor-layout.c (update_alignment_for_field): For
11759 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11760 && !DECL_PACKED (field), do the alignment update, just use
11761 only desired_align instead of MAX (type_align, desired_align)
11763 (place_field): Don't do known_align < desired_align handling
11764 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11765 is non-NULL, instead do it after rli->prev_field handling and
11766 only if not within a bitfield word. For DECL_PACKED (field)
11767 use type_align of BITS_PER_UNIT.
11769 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11771 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11772 superfluous parentheses and trailing spaces.
11774 2018-02-28 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/84584
11777 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11778 SCoPs with fake exit edge.
11780 2018-02-28 Martin Liska <mliska@suse.cz>
11783 * timevar.c (timer::print): Fix format to properly print 100%
11786 2018-02-28 Richard Biener <rguenther@suse.de>
11788 PR middle-end/84607
11789 * genmatch.c (capture_info::walk_match): Do not mark
11790 captured expressions without operands as expr_p given
11791 they act more like predicates and should be subject to
11792 "lost tail" side-effect preserving.
11794 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11796 PR rtl-optimization/81611
11797 * auto-inc-dec.c (attempt_change): Move dead note from
11798 mem_insn if it's the next use of regno
11799 (find_address): Take address use of reg holding
11800 non-incremented value. Add parm to limit search to the named
11802 (merge_in_block): Attempt to use a mem insn that is the next
11803 use of the original regno.
11805 2018-02-27 Martin Sebor <msebor@redhat.com>
11808 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11809 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11811 2018-02-27 Martin Sebor <msebor@redhat.com>
11813 PR translation/84207
11814 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11815 n argument to unsigned HOST_WIDE_INT.
11816 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11817 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11818 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11819 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11821 2018-02-27 Richard Biener <rguenther@suse.de>
11823 PR tree-optimization/84512
11824 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11825 Do not use the estimate returned from record_stmt_cost for
11826 the scalar iteration cost but sum properly using add_stmt_cost.
11828 2018-02-27 Richard Biener <rguenther@suse.de>
11830 PR tree-optimization/84466
11831 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11832 Adjust last change to less strictly validate use operands.
11834 2018-02-27 Martin Liska <mliska@suse.cz>
11836 PR gcov-profile/84548
11837 * gcov.c (process_file): Allow partial overlap and consider it
11838 also as group functions.
11839 (output_lines): Properly calculate range of lines for a group.
11841 2018-02-27 Martin Liska <mliska@suse.cz>
11843 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11844 'ggc' suffixes. Change first column width.
11845 (timer::print): Fix formatting of the column.
11847 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11849 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11850 preserve inline entry blocks for the sake of debug inline
11851 entry point markers alone.
11852 (remove_unused_locals): Suggest in comments a better place to
11853 force the preservation of inline entry blocks that are
11854 otherwise unused, but do not preserve them.
11856 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11858 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11860 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11863 * config/i386/constraints.md (Bs): Replace
11864 ix86_indirect_branch_register with
11865 TARGET_INDIRECT_BRANCH_REGISTER.
11867 * config/i386/i386.md (indirect_jump): Likewise.
11868 (tablejump): Likewise.
11869 (*sibcall_memory): Likewise.
11870 (*sibcall_value_memory): Likewise.
11871 Peepholes of indirect call and jump via memory: Likewise.
11872 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11873 (*sibcall_value_GOT_32): Likewise.
11874 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11875 (GOT_memory_operand): Likewise.
11876 (call_insn_operand): Likewise.
11877 (sibcall_insn_operand): Likewise.
11878 (GOT32_symbol_operand): Likewise.
11879 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11881 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11883 PR rtl-optimization/83496
11884 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11885 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11886 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11887 redundant insn, if any.
11888 (relax_delay_slots): Likewise.
11889 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11891 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11893 PR tree-optimization/83965
11894 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11895 that grouped statements are part of a reduction chain. Return
11896 true if the statement is not marked as a reduction itself but
11897 is part of a group.
11898 (vect_recog_dot_prod_pattern): Don't check whether the statement
11899 is part of a group here.
11900 (vect_recog_sad_pattern): Likewise.
11901 (vect_recog_widen_sum_pattern): Likewise.
11903 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11906 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11908 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11911 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11913 (ix86_output_indirect_function_return): New prototype.
11914 (ix86_split_simple_return_pop_internal): Likewise.
11915 * config/i386/i386.c (indirect_return_via_cx): New.
11916 (indirect_return_via_cx_bnd): Likewise.
11917 (indirect_thunk_name): Handle return va CX_REG.
11918 (output_indirect_thunk_function): Create alias for
11919 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11920 (ix86_output_indirect_jmp): Remove the bool argument.
11921 (ix86_output_indirect_function_return): New function.
11922 (ix86_split_simple_return_pop_internal): Likewise.
11923 * config/i386/i386.md (*indirect_jump): Don't pass false
11924 to ix86_output_indirect_jmp.
11925 (*tablejump_1): Likewise.
11926 (simple_return_pop_internal): Change it to define_insn_and_split.
11927 Call ix86_split_simple_return_pop_internal to split it for
11928 -mfunction-return=.
11929 (simple_return_indirect_internal): Call
11930 ix86_output_indirect_function_return instead of
11931 ix86_output_indirect_jmp.
11933 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11936 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11937 memset and value initialization afterwards.
11939 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11941 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11943 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11946 * common/config/aarch64/aarch64-common.c
11947 (aarch_option_optimization_table[]): Switch
11948 off fomit-frame-pointer
11950 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11951 Chung-Ju Wu <jasonwucj@gmail.com>
11953 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11955 (store_multiple): Ditto.
11957 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11959 * config.gcc: Add --with-cpu support for nds32 target.
11960 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11961 * config/nds32/nds32.opt: Add -mcpu= option.
11963 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11965 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11966 isel=yes): Warn for these deprecated options.
11968 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11970 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11971 ISA_2_5_MASKS_EMBEDDED.
11973 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11975 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11976 p->max as pointers rather than using iterative_hash_expr.
11978 2018-02-23 Carl Love <cel@us.ibm.com>
11980 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11981 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11983 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11984 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11985 P8V_BUILTIN_VEC_VUNSIGNED2.
11987 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11990 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11991 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11993 * lra-constraints.c (curr_insn_transform): Set up
11994 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11996 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11997 * lra-eliminations.c (spill_pseudos): Ditto.
11998 (process_insn_for_elimination): Ditto.
11999 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12000 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12001 LRA_NON_CLOBBERED_ALT.
12003 2018-02-22 Martin Sebor <msebor@redhat.com>
12005 PR tree-optimization/84480
12006 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12007 to maybe_diag_stxncpy_trunc. Call it.
12008 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12009 from gimple_fold_builtin_strcpy. Print inlining stack.
12010 (handle_builtin_stxncpy): Print inlining stack.
12011 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12013 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12016 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12017 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12018 and -fcheck-pointer-bounds are used together.
12019 (indirect_thunk_prefix): New enum.
12020 (indirect_thunk_need_prefix): New function.
12021 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12022 "_nt" instead of "_bnd" for NOTRACK prefix.
12023 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12024 (output_indirect_thunk_function): Likewise.
12026 (ix86_code_end): Update output_indirect_thunk_function calls.
12027 (ix86_output_indirect_branch_via_reg): Replace
12028 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12029 (ix86_output_indirect_branch_via_push): Likewise.
12030 (ix86_output_function_return): Likewise.
12031 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12032 incompatible with -fcf-protection=branch and
12033 -fcheck-pointer-bounds.
12035 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12038 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12039 Change gcc_assert call to output_operand_lossage.
12041 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12043 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12045 2018-02-22 DJ Delorie <dj@redhat.com>
12046 Sebastian Perta <sebastian.perta@renesas.com>
12047 Oleg Endo <olegendo@gcc.gnu.org>
12049 * config/rx/rx.c (rx_rtx_costs): New function.
12050 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12052 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12054 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12056 2018-02-22 Martin Liska <mliska@suse.cz>
12059 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12060 Add "native" as a possible value.
12062 2018-02-22 Martin Liska <mliska@suse.cz>
12065 * config/i386/i386.c (ix86_option_override_internal):
12066 Add "native" as a possible value for -march and -mtune.
12068 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12071 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12072 to all type variants.
12074 PR tree-optimization/84503
12075 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12076 width as info->bitpos + info->bitsize - start.
12077 (merged_store_group::merge_overlapping): Simplify width computation.
12078 (check_no_overlap): New function.
12079 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12080 start + width and last_order of the group, fail if check_no_overlap
12082 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12083 to group if check_no_overlap fails.
12085 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12087 * config/rs6000/altivec.md: Delete contraint arguments to
12088 define_expand, define_split, and define_peephole2, and in
12089 define_insn_and_split if always unused.
12090 * config/rs6000/darwin.md: Ditto.
12091 * config/rs6000/dfp.md: Ditto.
12092 * config/rs6000/rs6000.md: Ditto.
12093 * config/rs6000/sync.md: Ditto.
12094 * config/rs6000/vector.md: Ditto.
12095 * config/rs6000/vsx.md: Ditto.
12097 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12099 * config/rs6000/altivec.md: Write output control strings as braced
12100 blocks instead of double-quoted strings.
12101 * config/rs6000/darwin.md: Ditto.
12102 * config/rs6000/rs6000.md: Ditto.
12103 * config/rs6000/vector.md: Ditto.
12104 * config/rs6000/vsx.md: Ditto.
12106 2018-02-21 Jason Merrill <jason@redhat.com>
12108 PR c++/84314 - ICE with templates and fastcall attribute.
12109 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12111 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12113 * ipa-cp.c (determine_versionability): Fix comment typos.
12115 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12118 * ipa-cp.c (determine_versionability): Do not version functions caling
12121 2018-02-21 Martin Liska <mliska@suse.cz>
12124 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12125 Add "native" as a possible value.
12126 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12127 the macro when native cpu detection is available.
12129 2018-02-21 Martin Liska <mliska@suse.cz>
12132 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12133 Add "native" as a possible value.
12134 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12135 when native cpu detection is available.
12137 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12138 Martin Sebor <msebor@redhat.com>
12140 PR tree-optimization/84478
12141 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12143 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12144 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12145 support which is conservatively correct, for 2 only stay conservative
12146 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12147 argument to the 2 argument get_range_strlen, adjust 6 arg
12148 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12150 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12151 (gimple_fold_builtin_strlen): Pass true as last argument to
12154 2018-02-20 Martin Sebor <msebor@redhat.com>
12156 PR middle-end/84095
12157 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12158 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12159 (builtin_memref::builtin_memref): Factor out parts into
12160 set_base_and_offset and call it.
12162 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12164 PR middle-end/84406
12165 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12166 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12167 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12168 search at the associated MODE_INT.
12170 2018-02-20 Jeff Law <law@redhat.com>
12172 PR middle-end/82123
12173 PR tree-optimization/81592
12174 PR middle-end/79257
12175 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12176 for range data rather than using global data.
12177 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12178 range data rather than using global data.
12179 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12180 pass it to children as needed.
12181 (struct directive::fmtresult): Similarly.
12182 (struct directive::set_width): Similarly.
12183 (struct directive::set_precision): Similarly.
12184 (format_integer, format_directive, parse_directive): Similarly.
12185 (format_none): Accept unnamed vr_values parameter.
12186 (format_percent, format_floating, format_character): Similarly.
12187 (format_string, format_plain): Similarly.
12188 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12189 the EVRP range analyzer for range data rather than using global data.
12190 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12191 gimple-ssa-evrp-analyze.h
12192 (class sprintf_dom_walker): Add after_dom_children member function.
12193 Add evrp_range_analyzer member.
12194 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12195 range analyzer as needed.
12196 (sprintf_dom_walker::after_dom_children): New member function.
12197 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12199 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12200 (evrp_range_analyzer::pop_to_marker): Likewise.
12202 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12204 PR tree-optimization/84419
12205 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12206 with the required type if its current type is compatible but
12209 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12211 PR middle-end/82004
12212 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12213 after vectorization.
12215 2018-02-20 Martin Liska <mliska@suse.cz>
12218 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12219 possible values if we don't have a hint.
12221 2018-02-20 Martin Liska <mliska@suse.cz>
12225 * final.c (shorten_branches): Build align_tab array with one
12227 * opts.c (finish_options): Add alignment option limit check.
12228 (MAX_CODE_ALIGN): Likewise.
12229 (MAX_CODE_ALIGN_VALUE): Likewise.
12230 * doc/invoke.texi: Document maximum allowed option value for
12231 all -falign-* options.
12233 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12236 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12237 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12238 * var-tracking.c (emit_note_insn_var_location): Remove all references
12239 to NOTE_INSN_CALL_ARG_LOCATION.
12240 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12241 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12242 Use copy_rtx_if_shared.
12243 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12244 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12245 (dwarf2out_var_location): Remove handling of
12246 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12248 * final.c (final_scan_insn): Remove all references to
12249 NOTE_INSN_CALL_ARG_LOCATION.
12250 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12251 before dumping final insns.
12252 * except.c (emit_note_eh_region_end): Remove all references to
12253 NOTE_INSN_CALL_ARG_LOCATION.
12254 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12255 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12256 * config/arc/arc.c (hwloop_optimize): Likewise.
12257 * config/arm/arm.c (create_fix_barrier): Likewise.
12258 * config/s390/s390.c (s390_chunkify_start): Likewise.
12259 * config/sh/sh.c (find_barrier): Likewise.
12260 * config/i386/i386.c (rest_of_insert_endbranch,
12261 ix86_seh_fixup_eh_fallthru): Likewise.
12262 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12263 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12264 * config/frv/frv.c (frv_function_prologue): Likewise.
12265 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12267 (note_outside_basic_block_p): Remove all references to
12268 NOTE_INSN_CALL_ARG_LOCATION.
12269 * gengtype.c (adjust_field_rtx_def): Likewise.
12270 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12272 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12273 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12276 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12279 PR tree-optimization/84452
12280 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12281 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12284 2018-02-19 Martin Liska <mliska@suse.cz>
12287 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12289 2018-02-19 Martin Liska <mliska@suse.cz>
12290 Richard Sandiford <richard.sandiford@linaro.org>
12292 PR tree-optimization/82491
12293 * gimple-fold.c (get_base_constructor): Make earlier bail out
12296 2018-02-19 Carl Love <cel@us.ibm.com>
12298 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12299 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12301 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12302 P8V_BUILTIN_VEC_NEG.
12304 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12306 * config/rl78/rl78.md (movdf): New define expand.
12308 2018-02-19 Martin Liska <mliska@suse.cz>
12311 * doc/invoke.texi: Fix typo.
12312 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12314 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12316 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12317 handle rs6000_single_float and rs6000_double_float specially for
12320 2018-02-16 Jeff Law <law@redhat.com>
12322 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12323 (pop_regs): Use it.
12325 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12328 * ipa-inline.c (inline_small_functions): Fix a typo.
12330 2018-02-16 Nathan Sidwell <nathan@acm.org>
12332 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12334 2018-02-16 Carl Love <cel@us.ibm.com>
12336 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12337 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12338 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12339 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12340 expansion to P8V_BUILTIN_VEC_FLOAT2.
12342 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12344 PR rtl-optimization/70023
12345 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12346 src_regno into account.
12348 2018-02-16 Carl Love <cel@us.ibm.com>
12350 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12351 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12352 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12353 * config/rs6000/rs6000.c: Remove case statements for
12354 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12355 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12356 and P9V_BUILTIN_VEC_VINSERT4B.
12357 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12358 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12359 * config/rs6000/vsx.md:
12360 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12363 2018-02-16 Carl Love <cel@us.ibm.com>
12365 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12367 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12369 * config/rs6000/rs6000-c.c: Add the definitions for
12370 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12371 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12372 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12373 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12374 definition for insert4b and define insn *insert3b_internal.
12375 * doc/extend.texi: Add documentation for vec_extract4b.
12377 2018-02-16 Nathan Sidwell <nathan@acm.org>
12379 * doc/extend.texi (Backwards Compatibility): Mention friend
12380 injection. Note for-scope is deprecated.
12381 * doc/invoke.texi (-ffriend-injection): Deprecate.
12383 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12385 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12386 that moved to I2, also allow destinations that are a paradoxical
12387 subreg (instead of a normal reg).
12389 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12392 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12395 2018-02-16 Richard Biener <rguenther@suse.de>
12397 PR tree-optimization/84037
12398 PR tree-optimization/84016
12400 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12401 Adjust vec_construct for the fact we need additional higher latency
12402 128bit inserts for AVX256 and AVX512 vector builds.
12403 (ix86_add_stmt_cost): Scale vector construction cost for
12406 2018-02-16 Richard Biener <rguenther@suse.de>
12408 PR tree-optimization/84417
12409 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12410 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12411 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12413 2018-02-16 Martin Liska <mliska@suse.cz>
12416 * internal-fn.def (ASAN_CHECK): Set proper flags.
12417 (ASAN_MARK): Likewise.
12419 2018-02-16 Julia Koval <julia.koval@intel.com>
12421 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12422 from PTA_CANNONLAKE.
12424 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12427 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12428 Use ++iter rather than iter++ for std::list iterators.
12429 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12430 defer deleting them until all nodes in the forest are processed. Do
12431 free even leaf nodes. Change to_process into auto_vec.
12434 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12435 * vec.h (vec_default_construct): Use memset instead of placement new
12436 if BROKEN_VALUE_INITIALIZATION is defined.
12437 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12438 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12441 PR rtl-optimization/83723
12442 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12443 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12444 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12446 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12448 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12450 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12452 PR rtl-optimization/81443
12453 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12454 from inner REGs to paradoxical SUBREGs.
12456 2018-02-16 Richard Biener <rguenther@suse.de>
12458 PR tree-optimization/84399
12459 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12460 For operands we can analyze at their definition make sure we can
12461 analyze them at each use as well.
12463 2018-02-16 Richard Biener <rguenther@suse.de>
12465 PR tree-optimization/84190
12466 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12467 volatile accesses if the decl isn't volatile.
12469 2018-02-15 Jason Merrill <jason@redhat.com>
12471 PR c++/84314 - ICE with templates and fastcall attribute.
12472 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12473 TYPE_CANONICAL on an existing type.
12475 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12477 PR tree-optimization/84383
12478 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12479 dstoff nor call operand_equal_p if dstbase is NULL.
12481 PR tree-optimization/84334
12482 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12483 also a CONSTANT_CLASS_P, punt.
12485 2018-02-14 Jim Wilson <jimw@sifive.com>
12487 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12488 first SMALL_OPERAND check. New local min_second_step. Move assert
12489 to where locals are set. Add TARGET_RVC support.
12490 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12492 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12494 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12496 2018-02-14 Martin Sebor <msebor@redhat.com>
12498 PR tree-optimization/83698
12499 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12500 arrays constrain the offset range to their bounds.
12501 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12502 (builtin_access::overlap): Avoid setting the size of overlap if it's
12504 (maybe_diag_overlap): Also consider arrays when deciding what values
12505 of offsets to include in diagnostics.
12507 2018-02-14 Martin Sebor <msebor@redhat.com>
12510 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12511 that correspond to the kind of a declaration.
12513 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12516 * config/pa/pa.md: Load address of PIC label using the linkage table
12517 if the label is nonlocal.
12519 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12521 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12522 warning message if user requests -maltivec=be.
12523 * doc/invoke.texi: Document deprecation of -maltivec=be.
12525 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12528 * config/rs6000/rs6000-c.c: Update definitions for
12529 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12530 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12532 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12535 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12536 add _get_ssp intrinsics. Remove argument from
12537 __builtin_ia32_rdssp[d|q].
12538 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12539 * config/i386/i386-builtin.def: Remove argument from
12540 __builtin_ia32_rdssp[d|q].
12541 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12542 ix86_expand_special_args_builtin for _rdssp[d|q].
12543 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12544 Clear register before usage.
12545 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12546 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12548 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12550 PR tree-optimization/84357
12551 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12552 operand 1 of an ARRAY_REF too.
12554 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12557 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12558 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12560 (set_of_reg): New struct.
12561 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12562 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12563 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12565 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12566 Split into bitclr, bitset, bitinvert patterns if appropriate.
12567 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12568 use rx_fuse_in_memory_bitop.
12569 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12570 to named insn, correct maximum insn length.
12572 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12575 * machmode.def: Define a complex mode for PARTIAL_INT.
12576 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12578 * doc/rtl.texi: Document CSPImode.
12579 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12581 (msp430_hard_regno_nregs_with_padding): Likewise.
12583 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12586 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12588 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12590 PR rtl-optimization/84169
12591 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12592 we generated a parallel as new i3 and we split that to new i2 and i3
12593 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12594 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12595 those to i2, not i1. Partially rewrite this scan code.
12597 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12600 * stor-layout.c (place_field): For variable length fields, adjust
12601 offset_align afterwards not just based on the field's alignment,
12602 but also on the size.
12604 PR middle-end/84309
12605 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12606 of exps and logs in the use_exp2 case.
12608 2018-02-13 Jeff Law <law@redhat.com>
12610 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12611 entry for "vector".
12613 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12616 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12620 * common.opt (gas-loc-support, gas-locview-support): New.
12621 (ginline-points, ginternal-reset-location-views): New.
12622 * doc/invoke.texi: Document them. Use @itemx where intended.
12623 (gvariable-location-views): Adjust.
12624 * target.def (reset_location_view): New.
12625 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12626 (TARGET_RESET_LOCATION_VIEW): New.
12627 * doc/tm.texi: Rebuilt.
12628 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12629 (dwarf2out_default_as_locview_support): New.
12630 (output_asm_line_debug_info): Use option variables.
12631 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12632 (output_loc_list): Likewise.
12633 (add_high_low_attributes): Check option variables.
12634 Don't output entry view attribute in strict mode.
12635 (gen_inlined_subroutine_die): Check option variables.
12636 (dwarf2out_inline_entry): Likewise.
12637 (init_sections_and_labels): Likewise.
12638 (dwarf2out_early_finish): Likewise.
12639 (maybe_reset_location_view): New, from...
12640 (dwarf2out_var_location): ... here. Call it.
12641 * debug.h (dwarf2out_default_as_loc_support): Declare.
12642 (dwarf2out_default_as_locview_support): Declare.
12643 * hooks.c (hook_int_rtx_insn_0): New.
12644 * hooks.h (hook_int_rtx_insn_0): Declare.
12645 * toplev.c (process_options): Take -gas-loc-support and
12646 -gas-locview-support from dwarf2out. Enable
12647 -gvariable-location-views by default only with locview
12648 assembler support. Enable -ginternal-reset-location-views by
12649 default only if the target defines the corresponding hook.
12650 Enable -ginline-points by default if location views are
12651 enabled; force it disabled if statement frontiers are
12653 * tree-inline.c (expand_call_inline): Check option variables.
12654 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12656 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12658 PR tree-optimization/84321
12659 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12660 handling. Also check whether the anti-range contains any values
12661 that satisfy the mask; switch to a VR_RANGE if not.
12663 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12666 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12668 2018-02-13 Martin Jambor <mjambor@suse.cz>
12671 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12672 of call statements, also set location of a load to a temporary.
12674 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12676 * config/rl78/rl78.c (add_vector_labels): New function.
12677 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12678 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12679 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12680 which checks that no arguments are passed.
12681 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12682 * doc/extend.texi: Documentation for the new attribute.
12684 2018-02-13 Andreas Schwab <schwab@suse.de>
12686 * config/riscv/linux.h (CPP_SPEC): Define.
12688 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12691 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12692 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12693 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12694 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12695 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12696 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12697 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12698 temporarily for AES and PCLMUL builtins.
12700 PR tree-optimization/84339
12701 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12702 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12705 PR middle-end/84309
12706 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12707 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12708 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12710 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12712 * omp-simd-clone.h: New file.
12713 * omp-simd-clone.c: Include omp-simd-clone.h.
12714 (expand_simd_clones): No longer static.
12715 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12716 cgraph.h and omp-simd-clone.h.
12717 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12718 (vect_recog_widen_shift_pattern): Formatting fix.
12719 (vect_pattern_recog_1): Don't check optab for calls.
12722 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12723 operands[2] into a REG before using gen_lowpart on it.
12725 2018-02-12 Jeff Law <law@redhat.com>
12728 * config/sh/sh.c (find_barrier): Consider a sibling call
12731 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12732 successfully back substituting a reg.
12734 2018-02-12 Richard Biener <rguenther@suse.de>
12736 PR tree-optimization/84037
12737 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12738 parameter, move visited init to caller.
12739 (vect_slp_analyze_operations): Separate cost from validity
12740 check, initialize visited once for all instances.
12741 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12743 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12745 (vect_model_promotion_demotion_cost): Likewise.
12746 (vectorizable_bswap): Guard cost modeling with !slp_node
12747 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12749 (vectorizable_call): Likewise.
12750 (vectorizable_conversion): Likewise.
12751 (vectorizable_assignment): Likewise.
12752 (vectorizable_shift): Likewise.
12753 (vectorizable_operation): Likewise.
12754 (vectorizable_store): Likewise.
12755 (vectorizable_load): Likewise.
12756 (vectorizable_condition): Likewise.
12757 (vectorizable_comparison): Likewise.
12759 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12762 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12763 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12764 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12766 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12768 PR middle-end/83665
12769 * params.def (inline-min-speedup): Increase from 8 to 15.
12770 (max-inline-insns-auto): Decrease from 40 to 30.
12771 * ipa-split.c (consider_split): Add some buffer for function to
12772 be considered inlining candidate.
12773 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12776 2018-02-12 Richard Biener <rguenther@suse.de>
12778 PR tree-optimization/84037
12779 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12780 matched stmts if we cannot swap the non-matched ones.
12782 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12784 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12785 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12786 _mm_maskz_scalef_round_ss): New intrinsics.
12787 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12788 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12789 __builtin_ia32_scalefss_round): Remove.
12790 (__builtin_ia32_scalefsd_mask_round,
12791 __builtin_ia32_scalefss_mask_round): New intrinsics.
12792 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12793 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12794 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12795 "<round_constraint>")): Changed to ...
12796 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12797 "<round_scalar_constraint>")): ... this.
12798 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12799 %0, %1, %2<round_op3>}"): Changed to ...
12800 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12801 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12802 %2<round_scalar_mask_op3>}"): ... this.
12803 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12805 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12807 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12808 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12809 (_mm_maskz_sqrt_round_ss): New intrinsics.
12810 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12811 (__builtin_ia32_sqrtsd_mask_round)
12812 (__builtin_ia32_sqrtss_mask_round): New builtins.
12813 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12814 (__builtin_ia32_sqrtss_round): Remove.
12815 (__builtin_ia32_sqrtsd_mask_round)
12816 (__builtin_ia32_sqrtss_mask_round): New builtins.
12817 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12818 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12819 ((match_operand:VF_128 1 "vector_operand"
12820 "xBm,<round_constraint>")): Changed to ...
12821 ((match_operand:VF_128 1 "vector_operand"
12822 "xBm,<round_scalar_constraint>")): ... this.
12823 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12824 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12825 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12826 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12827 %<iptr>1<round_scalar_mask_op3>}): ... this.
12828 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12829 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12831 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12834 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12835 Cast vec_cmpeq result to correct type.
12836 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12837 Cast vec_cmpgt result to correct type.
12839 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12841 * final.c (final_scan_insn_1): Renamed from...
12842 (final_scan_insn): ... this. New wrapper, to recover
12843 seen from the outermost call in recursive ones.
12844 * config/sparc/sparc.c (output_return): Drop seen from call.
12845 (output_sibcall): Likewise.
12846 * config/visium/visium.c (output_branch): Likewise.
12848 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12850 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12853 2018-02-10 Alan Modra <amodra@gmail.com>
12856 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12857 Specify LR as an input.
12859 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12862 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12863 remove_member_access_dummy_vars): New functions.
12864 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12865 lower_omp_1, execute_lower_omp): Use them.
12867 PR rtl-optimization/84308
12868 * shrink-wrap.c (spread_components): Release todo vector.
12870 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12872 PR rtl-optimization/57193
12873 * ira-color.c (struct allocno_color_data): Add member
12874 conflict_allocno_hard_prefs.
12875 (update_conflict_allocno_hard_prefs): New.
12876 (bucket_allocno_compare_func): Add a preference based on
12877 conflict_allocno_hard_prefs.
12878 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12879 (color_allocnos): Remove a dead code. Initiate
12880 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12882 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12885 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12886 constraint from =wa to wa. Avoid a subreg on the output operand,
12887 instead use a pseudo and subreg it in a move.
12888 (p9_xxbrd_<mode>): Changed to ...
12889 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12890 (p9_xxbrd_v2df): New expander.
12891 (p9_xxbrw_<mode>): Changed to ...
12892 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12893 (p9_xxbrw_v4sf): New expander.
12895 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12897 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12899 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12902 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12903 multiply in 32-bit mode.
12904 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12905 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12908 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12910 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12911 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12912 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12913 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12915 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12918 * dwarf2out.c (is_trivial_indirect_ref): New function.
12919 (dwarf2out_late_global_decl): Do not generate a location
12920 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12921 and that are not defined in the current unit.
12923 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12925 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12926 instead of a libcall for UNORDERED.
12928 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12931 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12932 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12934 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12937 * config/s390/s390.c (s390_set_current_function): Invoke
12938 s390_indirect_branch_settings also if fndecl didn't change.
12940 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12942 * config/rs6000/rs6000.md (blockage): Set length to zero.
12944 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12946 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12948 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12951 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12952 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12956 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12957 PARALLEL incoming that failed vt_get_decl_and_offset check.
12959 PR middle-end/84237
12960 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12961 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12963 (get_variable_section): For decls in named .bss* sections pass true as
12964 second argument to bss_initializer_p.
12966 2018-02-09 Marek Polacek <polacek@redhat.com>
12967 Jakub Jelinek <jakub@redhat.com>
12970 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12971 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12972 Sync some changes from cxx_fold_indirect_ref.
12974 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12976 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12978 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12979 (BLOCK_INLINE_ENTRY_LABEL): New.
12980 (dwarf2out_var_location): Disregard inline entry markers.
12981 (inline_entry_data): New struct.
12982 (inline_entry_data_hasher): New hashtable type.
12983 (inline_entry_data_hasher::hash): New.
12984 (inline_entry_data_hasher::equal): New.
12985 (inline_entry_data_table): New variable.
12986 (add_high_low_attributes): Add DW_AT_entry_pc and
12987 DW_AT_GNU_entry_view attributes if a pending entry is found
12988 in inline_entry_data_table. Add old entry_pc attribute only
12989 if debug nonbinding markers are disabled.
12990 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12991 markers are enabled.
12992 (block_within_block_p, dwarf2out_inline_entry): New.
12993 (dwarf2out_finish): Check that no entries remained in
12994 inline_entry_data_table.
12995 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12996 (final_scan_insn, notice_source_line): Likewise.
12997 (rest_of_clean_state): Skip inline entry markers.
12998 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13000 * gimple.c (gimple_build_debug_inline_entry): New.
13001 * gimple.h (enum gimple_debug_subcode): Add
13002 GIMPLE_DEBUG_INLINE_ENTRY.
13003 (gimple_build_debug_inline_entry): Declare.
13004 (gimple_debug_inline_entry_p): New.
13005 (gimple_debug_nonbind_marker_p): Adjust.
13006 * insn-notes.def (INLINE_ENTRY): New.
13007 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13008 inline entry marker notes.
13009 (print_insn): Likewise.
13010 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13011 (INSN_DEBUG_MARKER_KIND): Likewise.
13012 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13013 * tree-inline.c (expand_call_inline): Build and insert
13014 debug_inline_entry stmt.
13015 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13016 inline entry blocks early, if nonbind markers are enabled.
13017 (dump_scope_block): Dump fragment info.
13018 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13019 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13020 (gimple_build_debug_inline_entry): New.
13021 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13022 Enable/disable inline entry points too.
13023 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13024 (DEBUG_INSN): Describe inline entry markers.
13026 * common.opt (gvariable-location-views): New.
13027 (gvariable-location-views=incompat5): New.
13028 * config.in: Rebuilt.
13029 * configure: Rebuilt.
13030 * configure.ac: Test assembler for view support.
13031 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13032 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13033 * dwarf2out.c (var_loc_view): New typedef.
13034 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13035 (dwarf2out_locviews_in_attribute): New.
13036 (dwarf2out_locviews_in_loclist): New.
13037 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13038 (enum dw_line_info_opcode): Add LI_adv_address.
13039 (struct dw_line_info_table): Add view.
13040 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13041 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13042 (zero_view_p): New variable.
13043 (ZERO_VIEW_P): New macro.
13044 (output_asm_line_debug_info): New.
13045 (struct var_loc_node): Add view.
13046 (add_AT_view_list, AT_loc_list): New.
13047 (add_var_loc_to_decl): Add view param. Test it against last.
13048 (new_loc_list): Add view params. Record them.
13049 (AT_loc_list_ptr): Handle loc and view lists.
13050 (view_list_to_loc_list_val_node): New.
13051 (print_dw_val): Handle dw_val_class_view_list.
13052 (size_of_die): Likewise.
13053 (value_format): Likewise.
13054 (loc_list_has_views): New.
13055 (gen_llsym): Set vl_symbol too.
13056 (maybe_gen_llsym, skip_loc_list_entry): New.
13057 (dwarf2out_maybe_output_loclist_view_pair): New.
13058 (output_loc_list): Output view list or entries too.
13059 (output_view_list_offset): New.
13060 (output_die): Handle dw_val_class_view_list.
13061 (output_dwarf_version): New.
13062 (output_compilation_unit_header): Use it.
13063 (output_skeleton_debug_sections): Likewise.
13064 (output_rnglists, output_line_info): Likewise.
13065 (output_pubnames, output_aranges): Update version comments.
13066 (output_one_line_info_table): Output view numbers in asm comments.
13067 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13068 Call maybe_gen_llsym.
13069 (loc_list_from_tree_1): Adjust.
13070 (add_AT_location_description): Create view list attribute if
13071 needed, check it's absent otherwise.
13072 (convert_cfa_to_fb_loc_list): Adjust.
13073 (maybe_emit_file): Call output_asm_line_debug_info for test.
13074 (dwarf2out_var_location): Reset views as needed. Precompute
13075 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13076 attribute. Set view.
13077 (new_line_info_table): Reset next view.
13078 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13079 (dwarf2out_source_line): Likewise. Output view resets and labels to
13080 the assembler, or select appropriate line info opcodes.
13081 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13082 (optimize_string_length): Catch it. Adjust.
13083 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13084 dw_val_class_view_list, and remove it if no longer needed.
13085 (hash_loc_list): Hash view numbers.
13086 (loc_list_hasher::equal): Compare them.
13087 (optimize_location_lists): Check whether a view list symbol is
13088 needed, and whether the locview attribute is present, and
13089 whether they match. Remove the locview attribute if no longer
13091 (index_location_lists): Call skip_loc_list_entry for test.
13092 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13093 Use output_dwarf_version.
13094 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13095 (struct dw_val_node): Add val_view_list.
13096 * final.c (SEEN_NEXT_VIEW): New.
13097 (set_next_view_needed): New.
13098 (clear_next_view_needed): New.
13099 (maybe_output_next_view): New.
13100 (final_start_function): Rename to...
13101 (final_start_function_1): ... this. Take pointer to FIRST,
13102 add SEEN parameter. Emit param bindings in the initial view.
13103 (final_start_function): Reintroduce SEEN-less interface.
13104 (final): Rename to...
13105 (final_1): ... this. Take SEEN parameter. Output final pending
13106 next view at the end.
13107 (final): Reintroduce seen-less interface.
13108 (final_scan_insn): Output pending next view before switching
13109 sections or ending a block. Mark the next view as needed when
13110 outputting variable locations. Notify debug backend of section
13111 changes, and of location view changes.
13112 (rest_of_handle_final): Adjust.
13113 * toplev.c (process_options): Autodetect value for debug variable
13114 location views option. Warn on incompat5 without -gdwarf-5.
13115 * doc/invoke.texi (gvariable-location-views): New.
13116 (gvariable-location-views=incompat5): New.
13117 (gno-variable-location-views): New.
13119 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13121 PR tree-optimization/84136
13122 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13123 that the result of find_edge is non-NULL.
13125 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13128 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13129 storing integer register in SImode. Fix cost of 256 and 512
13130 byte aligned SSE register store.
13132 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13134 * config/i386/i386.c (ix86_multiplication_cost): Fix
13135 multiplication cost for TARGET_AVX512DQ.
13137 2018-02-08 Marek Polacek <polacek@redhat.com>
13139 PR tree-optimization/84238
13140 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13143 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13145 PR tree-optimization/84265
13146 * tree-vect-stmts.c (vectorizable_store): Don't treat
13147 VMAT_CONTIGUOUS accesses as grouped.
13148 (vectorizable_load): Likewise.
13150 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13152 PR tree-optimization/81635
13153 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13154 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13155 (test_round_for_mask): New functions.
13156 (wide_int_cc_tests): Call test_round_for_mask.
13157 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13158 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13159 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13160 range returned by get_range_info.
13162 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13165 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13166 * symtab.c: Include builtins.h
13167 (symtab_node::output_to_lto_symbol_table_p): Move here
13168 from lto-streamer-out.c:output_symbol_p.
13169 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13170 (output_symbol_p): Move all logic to symtab.c
13171 (produce_symtab): Update.
13173 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13175 * config/s390/s390-opts.h (enum indirect_branch): Define.
13176 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13177 (s390_indirect_branch_via_thunk)
13178 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13179 (enum s390_indirect_branch_type): Define.
13180 * config/s390/s390.c (struct s390_frame_layout, struct
13181 machine_function): Remove.
13182 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13183 (indirect_branch_table_label_no, indirect_branch_table_name):
13185 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13186 (enum s390_indirect_branch_option): Define.
13187 (s390_return_addr_from_memory): New function.
13188 (s390_handle_string_attribute): New function.
13189 (s390_attribute_table): Add new attribute handler.
13190 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13191 (s390_indirect_branch_via_thunk): New function.
13192 (s390_indirect_branch_via_inline_thunk): New function.
13193 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13194 sibling call optimization for non z10 compiles.
13195 (s390_emit_call): Force indirect branch target to be a single
13196 register. Add r1 clobber for non-z10 compiles.
13197 (s390_emit_epilogue): Emit return jump via return_use expander.
13198 (s390_reorg): Handle JUMP_INSNs as execute targets.
13199 (s390_option_override_internal): Perform validity checks for the
13200 new command line options.
13201 (s390_indirect_branch_attrvalue): New function.
13202 (s390_indirect_branch_settings): New function.
13203 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13204 (s390_output_indirect_thunk_function): New function.
13205 (s390_code_end): Implement target hook.
13206 (s390_case_values_threshold): Implement target hook.
13207 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13209 * config/s390/s390.h (struct s390_frame_layout)
13210 (struct machine_function): Move here from s390.c.
13211 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13212 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13213 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13214 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13215 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13216 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13217 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13218 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13219 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13220 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13221 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13222 (mnemonic attribute): Add values which aren't recognized
13224 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13225 pattern for branch conversion. Fix mnemonic attribute.
13226 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13227 indirect branch via thunk if requested.
13228 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13229 ("*indirect_jump"): Disable for branch conversion using out of
13231 ("indirect_jump_via_thunk<mode>_z10")
13232 ("indirect_jump_via_thunk<mode>")
13233 ("indirect_jump_via_inlinethunk<mode>_z10")
13234 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13235 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13236 ("casesi_jump_via_inlinethunk<mode>_z10")
13237 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13238 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13239 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13240 ("*indirect2_jump"): Disable for branch conversion.
13241 ("casesi_jump"): Turn into expander and expand patterns for branch
13243 ("return_use"): New expander.
13244 ("*return"): Emit return via thunk and rename it to ...
13245 ("*return<mode>"): ... this one.
13246 * config/s390/s390.opt: Add new options and and enum for the
13249 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13251 * lra-constraints.c (match_reload): Unconditionally use
13252 gen_lowpart_SUBREG, rather than selecting between that
13253 and equivalent gen_rtx_SUBREG code.
13255 2018-02-08 Richard Biener <rguenther@suse.de>
13257 PR tree-optimization/84233
13258 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13259 changed flag instead of boguously re-using phi_inserted.
13261 2018-02-08 Martin Jambor <mjambor@suse.cz>
13263 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13264 static local variables.
13266 2018-02-08 Richard Biener <rguenther@suse.de>
13268 PR tree-optimization/84278
13269 * tree-vect-stmts.c (vectorizable_store): When looking for
13270 smaller vector types to perform grouped strided loads/stores
13271 make sure the mode is supported by the target.
13272 (vectorizable_load): Likewise.
13274 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13276 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13277 Increase LDP/STP opportunities by adding adjacent callee-saves.
13279 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13281 PR rtl-optimization/84068
13282 PR rtl-optimization/83459
13283 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13285 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13287 PR tree-optimization/84224
13288 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13289 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13290 non-zero arguments.
13292 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13295 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13296 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13297 count, remove one USE.
13299 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13301 * doc/install.texi (Configuration): Document the
13302 --with-long-double-format={ibm,ieee} PowerPC configuration
13306 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13307 Convert from define_expand to be define_insn_and_split. Rework
13308 float/double/_Float128 conversions to QI/HI/SImode to work with
13309 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13310 conversions to QI/HImode types did a store and then a load to
13311 truncate the value. For conversions to VSX registers, don't split
13312 the insn, instead emit the code directly. Use the code iterator
13313 any_fix to combine signed and unsigned conversions.
13314 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13315 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13316 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13317 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13318 (fix_<mode>di2_hw): Likewise.
13319 (fixuns_<mode>di2_hw): Likewise.
13320 (fix_<mode>si2_hw): Likewise.
13321 (fixuns_<mode>si2_hw): Likewise.
13322 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13323 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13324 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13325 fix<uns>_trunc<SFDF:mode>si2_p8.
13326 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13327 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13328 (fix<uns>_<mode>_mem): Likewise.
13329 (fctiw<u>z_<mode>_mem): Likewise.
13330 (fix<uns>_<mode>_mem): Likewise.
13331 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13332 the register allocator from doing a direct move to the GPRs to do
13333 a store, and instead use the ISA 3.0 store byte/half-word from
13334 vector register instruction. For IEEE 128-bit floating point,
13335 also optimize stores of 32-bit ints.
13336 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13338 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13340 * genextract.c (push_pathstr_operand): New function to support
13342 (walk_rtx): Call push_pathstr_operand.
13343 (print_path): Support [a-zA-Z].
13345 2018-02-07 Richard Biener <rguenther@suse.de>
13347 PR tree-optimization/84037
13348 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13349 (cse_and_gimplify_to_preheader): Declare.
13350 (vect_get_place_in_interleaving_chain): Likewise.
13351 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13353 (_loop_vec_info::~_loop_vec_info): Delete it.
13354 (cse_and_gimplify_to_preheader): New function.
13355 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13356 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13357 (vectorizable_load): Likewise. For grouped stores always base
13358 the IV on the first element.
13359 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13360 condition before gimplifying.
13362 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13364 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13365 *DIV_EXPR and *MOD_EXPR.
13367 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13370 * config/i386/i386.c (ix86_option_override_internal): Mask out
13371 the CF_SET bit when checking -fcf-protection.
13373 2018-02-07 Tom de Vries <tom@codesourcery.com>
13376 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13379 2018-02-07 Richard Biener <rguenther@suse.de>
13381 PR tree-optimization/84204
13382 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13385 PR tree-optimization/84205
13386 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13387 special-case isl_ast_op_zdiv_r.
13389 PR tree-optimization/84223
13390 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13391 Only add conditions from within the region.
13392 (gather_bbs::after_dom_children): Adjust.
13394 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13397 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13398 * config/avr/avr.md: Only post-reload split REG-REG moves if
13399 either register is GENERAL_REG_P.
13401 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13403 PR tree-optimization/84235
13404 * tree-ssa-scopedtables.c
13405 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13406 if the subtraction is performed in floating point type where NaNs are
13407 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13408 build 1. Formatting fix.
13410 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13413 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13414 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13415 and skip it regardless of bb boundaries. Use CALL_P macro,
13416 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13417 unnecessarily, formatting fix.
13419 2018-02-06 Michael Collison <michael.collison@arm.com>
13421 * config/arm/thumb2.md:
13422 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13423 (*thumb_mov_notscc): Ditto.
13425 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13428 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13429 unsigned_fix, not "s".
13431 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13433 * configure.ac (gcc_fn_eh_frame_ro): New function.
13434 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13435 correct .eh_frame permissions.
13436 * configure: Regenerate.
13438 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13440 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13441 irrelevant options.
13443 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13445 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13446 Display warning message for -mno-speculate-indirect-jumps.
13448 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13450 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13452 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13454 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13456 PR tree-optimization/84225
13457 * tree-eh.c (find_trapping_overflow): Only call
13458 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13460 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13463 * config/i386/i386.c: Reimplement the check of possible options
13464 -mibt/-mshstk conbination. Change error messages.
13465 * doc/invoke.texi: Fix a typo: remove extra '='.
13467 2018-02-06 Marek Polacek <polacek@redhat.com>
13469 PR tree-optimization/84228
13470 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13472 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13475 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13476 emitted arch directives.
13477 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13478 __ARM_FEATURE_COPROC before changing architectures.
13480 2018-02-06 Richard Biener <rguenther@suse.de>
13482 * config/i386/i386.c (print_reg): Fix typo.
13483 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13485 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13487 * configure: Regenerate.
13489 2018-02-05 Martin Sebor <msebor@redhat.com>
13491 PR tree-optimization/83369
13492 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13495 2018-02-05 Martin Liska <mliska@suse.cz>
13497 * doc/invoke.texi: Cherry-pick upstream r323995.
13499 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13501 * ira.c (ira_init_register_move_cost): Adjust comment.
13503 2018-02-05 Martin Liska <mliska@suse.cz>
13505 PR gcov-profile/84137
13506 * doc/gcov.texi: Fix typo in documentation.
13508 2018-02-05 Martin Liska <mliska@suse.cz>
13510 PR gcov-profile/83879
13511 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13512 using dlopen functionality.
13514 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13516 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13517 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13518 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13519 _mm_maskz_range_round_ss): New intrinsics.
13520 (__builtin_ia32_rangesd128_round)
13521 (__builtin_ia32_rangess128_round): Remove.
13522 (__builtin_ia32_rangesd128_mask_round,
13523 __builtin_ia32_rangess128_mask_round): New builtins.
13524 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13525 __builtin_ia32_rangess128_round): Remove.
13526 (__builtin_ia32_rangesd128_mask_round,
13527 __builtin_ia32_rangess128_mask_round): New builtins.
13528 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13529 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13530 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13531 "<round_saeonly_constraint>")): Changed to ...
13532 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13533 "<round_saeonly_scalar_constraint>")): ... this.
13534 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13535 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13536 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13537 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13538 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13540 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13542 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13544 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13545 Remove all values except native, 8540 and 8548.
13547 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13549 * config/i386/i386.c (ix86_output_function_return): Pass
13550 INVALID_REGNUM, instead of -1, as invalid register number to
13551 indirect_thunk_name and output_indirect_thunk.
13553 2018-02-02 Julia Koval <julia.koval@intel.com>
13555 * config.gcc: Add -march=icelake.
13556 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13557 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13558 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13559 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13560 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13561 (processor_target_table): Add icelake.
13562 (ix86_option_override_internal): Handle new PTAs.
13563 (get_builtin_code_for_version): Handle icelake.
13564 (M_INTEL_COREI7_ICELAKE): New.
13565 (fold_builtin_cpu): Handle icelake.
13566 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13567 * doc/invoke.texi: Add -march=icelake.
13569 2018-02-02 Julia Koval <julia.koval@intel.com>
13571 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13572 to wide_int_bitmask.
13573 * wide-int-bitmask.h: New.
13575 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13578 * config/i386/i386.md: Replace Pmode with word_mode in
13579 builtin_setjmp_setup and builtin_longjmp to support x32.
13581 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13585 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13587 (rs6000_supported_cpu_names): New static variable.
13588 (linux_cpu_translation_table): Likewise.
13589 (elf_platform) <cpu>: Define new static variable and use it.
13590 Translate kernel AT_PLATFORM name to canonical name if needed.
13591 Error if platform name is unknown.
13593 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13596 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13598 2018-02-01 Jeff Law <law@redhat.com>
13601 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13602 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13603 the scratch if RELEASE_VIA_POP is false.
13604 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13605 If we have to save a temporary register, decrement SIZE appropriately.
13606 Pass new arguments to release_scratch_register_on_entry.
13607 (ix86_adjust_stack_and_probe): Likewise.
13608 (ix86_emit_probe_stack_range): Pass new arguments to
13609 release_scratch_register_on_entry.
13611 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13613 PR rtl-optimization/84157
13614 * combine.c (change_zero_ext): Use REG_P predicate in
13615 front of HARD_REGISTER_P predicate.
13617 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13619 * config/avr/avr.c (avr_option_override): Move disabling of
13620 -fdelete-null-pointer-checks to...
13621 * common/config/avr/avr-common.c (avr_option_optimization_table):
13624 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13626 PR tree-optimization/81635
13627 * tree-data-ref.c (split_constant_offset_1): For types that
13628 wrap on overflow, try to use range info to prove that wrapping
13631 2018-02-01 Renlin Li <renlin.li@arm.com>
13634 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13635 TAILCALL_ADDR_REGS.
13636 (aarch64_register_move_cost): Likewise.
13637 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13638 TAILCALL_ADDR_REGS.
13639 (REG_CLASS_NAMES): Likewise.
13640 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13641 TAILCALL_ADDR_REGS. Remove IP registers.
13642 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13644 2018-02-01 Richard Biener <rguenther@suse.de>
13646 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13647 for specifying RPO order and allow NULL for that.
13648 * domwalk.c (dom_walker::dom_walker): Likewise.
13649 (dom_walker::walk): Handle NULL RPO order.
13650 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13652 (rewrite_update_dom_walker): Likewise.
13653 (mark_def_dom_walker): Likewise.
13655 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13657 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13658 (aarch64_maybe_expand_sve_subreg_move): Declare.
13659 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13660 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13662 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13663 that are semantically a reverse operation.
13664 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13665 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13666 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13668 (aarch64_can_change_mode_class): For big-endian, forbid changes
13669 between two SVE modes if they have different element sizes.
13671 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13673 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13674 the TImode handling for big-endian targets.
13676 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13678 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13679 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13681 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13682 Remove BSWAP handing for big-endian targets and use the form of
13683 LD1RQ appropariate for the mode.
13685 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13687 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13688 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13689 duplicated element.
13691 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13694 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13695 check for operands that need to go through aarch64_sve_reload_be.
13697 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13699 PR tree-optimization/81661
13700 PR tree-optimization/84117
13701 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13702 * tree-eh.c: Include gimplify.h.
13703 (find_trapping_overflow, replace_trapping_overflow,
13704 rewrite_to_non_trapping_overflow): New functions.
13705 * tree-vect-loop.c: Include tree-eh.h.
13706 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13707 * tree-data-ref.c: Include tree-eh.h.
13708 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13710 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13712 PR rtl-optimization/84123
13713 * combine.c (change_zero_ext): Check if hard register satisfies
13714 can_change_dest_mode before calling gen_lowpart_SUBREG.
13716 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13719 * ira.c (ira_init_register_move_cost): Remove assert.
13721 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13723 PR rtl-optimization/84071
13724 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13725 * doc/tm.texi: Regenerate.
13727 2018-01-31 Richard Biener <rguenther@suse.de>
13729 PR tree-optimization/84132
13730 * tree-data-ref.c (analyze_miv_subscript): Properly
13731 check whether evolution_function_is_affine_multivariate_p
13732 before calling gcd_of_steps_may_divide_p.
13734 2018-01-31 Julia Koval <julia.koval@intel.com>
13737 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13738 * config/i386/i386.md (rdpid_rex64) New.
13739 (rdpid): Make 32bit only.
13741 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13744 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13745 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13747 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13750 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13752 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13754 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13756 PR rtl-optimization/84071
13757 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13758 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13760 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13762 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13763 (arc_attribute_table): Add 'aux' attribute.
13764 (arc_in_small_data_p): Consider aux like variables.
13765 (arc_is_aux_reg_p): New function.
13766 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13767 (arc_get_aux_arg): New function.
13768 (prepare_move_operands): Handle aux-register access.
13769 (arc_handle_aux_attribute): New function.
13770 * doc/extend.texi (ARC Variable attributes): Add subsection.
13772 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13774 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13775 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13776 (arc_attribute_table): Add 'uncached' attribute.
13777 (arc_print_operand): Print '.di' flag for uncached memory
13779 (arc_in_small_data_p): Do not consider for small data the uncached
13781 (arc_is_uncached_mem_p): New function.
13782 * config/arc/predicates.md (compact_store_memory_operand): Check
13783 for uncached memory accesses.
13784 (nonvol_nonimm_operand): Likewise.
13785 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13787 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13790 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13791 falign-loops=): Add Optimization flag.
13793 2018-01-30 Jeff Law <law@redhat.com>
13796 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13797 INT_REGISTERS_SAVED. Check it prior to calling
13798 get_scratch_register_on_entry.
13799 (ix86_adjust_stack_and_probe): Similarly.
13800 (ix86_emit_probe_stack_range): Similarly.
13801 (ix86_expand_prologue): Corresponding changes.
13803 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13806 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13807 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13809 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13812 * lra-constraints.c (curr_insn_transform): Process AND in the
13815 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13817 PR rtl-optimization/83986
13818 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13819 dependence against last_pending_memory_flush in addition to
13820 pending_jump_insns.
13822 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13824 PR tree-optimization/81611
13825 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13828 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13831 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13834 2018-01-30 Richard Biener <rguenther@suse.de>
13835 Jakub Jelinek <jakub@redhat.com>
13837 PR tree-optimization/84111
13838 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13839 inner loops added during recursion, as they don't have up-to-date
13842 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13845 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13846 (can_inline_edge_by_limits_p): ... here.
13847 (can_early_inline_edge_p, check_callers,
13848 update_caller_keys, update_callee_keys, recursive_inlining,
13849 add_new_edges_to_heap, speculation_useful_p,
13850 inline_small_functions,
13851 inline_small_functions, flatten_function,
13852 inline_to_all_callers_1): Update.
13854 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13856 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13859 2018-01-30 Richard Biener <rguenther@suse.de>
13861 PR tree-optimization/83008
13862 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13863 invariant and constant vector uses in stmts when they need
13864 more than one stmt.
13866 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13869 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13870 * configure: Regenerate.
13872 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13874 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13876 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13877 Use gen_rtx_REG rather than gen_lowpart.
13879 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13881 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13882 rather than 0 when creating partial subregs.
13884 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13886 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13889 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13892 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13893 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13894 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13895 flags. This restores the settings used before the 2017-07-24.
13896 Turning off pre increment/decrement/modify allows IVOPTS to
13897 optimize DF/SF loops where the index is an int.
13899 2018-01-29 Richard Biener <rguenther@suse.de>
13900 Kelvin Nilsen <kelvin@gcc.gnu.org>
13903 * tree-vect-stmts.c (vectorizable_call): Don't call
13904 targetm.vectorize_builtin_md_vectorized_function if callee is
13907 2018-01-22 Carl Love <cel@us.ibm.com>
13909 * doc/extend.tex: Fix typo in second arg in
13910 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13912 2018-01-29 Richard Biener <rguenther@suse.de>
13914 PR tree-optimization/84086
13915 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13916 (flush_ssaname_freelist): When SSA names were released reset
13917 the SCEV hash table.
13919 2018-01-29 Richard Biener <rguenther@suse.de>
13921 PR tree-optimization/84057
13922 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13923 removed paths when removing edges.
13925 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13927 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13928 -mfunction-return=@var{choice}.
13930 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13932 PR diagnostic/84034
13933 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13934 Handle CR like TAB.
13935 (layout::print_source_line): Likewise.
13936 (test_get_line_width_without_trailing_whitespace): Add test cases.
13938 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13940 PR middle-end/84040
13941 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13944 2018-01-26 Jim Wilson <jimw@sifive.com>
13946 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13948 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13951 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13953 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13954 and CMP + SUB-immediate -> SUBS.
13956 2018-01-26 Martin Sebor <msebor@redhat.com>
13958 PR tree-optimization/83896
13959 * tree-ssa-strlen.c (get_string_len): Rename...
13960 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13961 Avoid assuming length is constant.
13962 (handle_char_store): Use HOST_WIDE_INT for string length.
13964 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13967 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13968 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13970 2018-01-26 Richard Biener <rguenther@suse.de>
13972 PR rtl-optimization/84003
13973 * dse.c (record_store): Only record redundant stores when
13974 the earlier store aliases at least all accesses the later one does.
13976 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13978 PR rtl-optimization/83985
13979 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13980 REG_CFA_RESTORE insns.
13981 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13982 REG_CFA_RESTORE insns here.
13985 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13986 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13988 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13990 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13991 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13992 (arc_init): Likewise.
13993 (arc_override_options): Likewise.
13994 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13996 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13998 * config/arc/arc.h (TARGET_DBNZ): Define.
13999 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14000 properly set the tune attribute.
14001 (dbnz): Use TARGET_DBNZ guard.
14002 * config/arc/arc.opt (mtune): Add core3 option.
14004 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14006 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14007 recognize new pic like addresses.
14008 (arc_delegitimize_address): Clean up.
14010 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14012 * config/arc/arc-arches.def: Option mrf16 valid for all
14014 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14015 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14016 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14017 * config/arc/arc-tables.opt: Regenerate.
14018 * config/arc/arc.c (arc_conditional_register_usage): Handle
14019 reduced register file case.
14020 (arc_file_start): Set must have build attributes.
14021 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14022 mrf16 option value.
14023 * config/arc/arc.opt (mrf16): Add new option.
14024 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14025 * config/arc/genmultilib.awk: Handle new mrf16 option.
14026 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14027 * config/arc/t-multilib: Regenerate.
14028 * doc/invoke.texi (ARC Options): Document mrf16 option.
14030 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14032 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14033 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14034 (arc_attribute_table): Add 'secure_call' attribute.
14035 (arc_print_operand): Print secure call operand.
14036 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14038 (arc_is_secure_call_p): New function. * config/arc/arc.md
14039 (call_i): Add support for sjli instruction.
14040 (call_value_i): Likewise.
14041 * config/arc/constraints.md (Csc): New constraint.
14043 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14044 John Eric Martin <John.Martin@emmicro-us.com>
14046 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14047 * config/arc/arc.c (_arc_jli_section): New struct.
14048 (arc_jli_section): New type.
14049 (rc_jli_sections): New static variable.
14050 (arc_handle_jli_attribute): New function.
14051 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14052 (arc_file_end): New function.
14053 (TARGET_ASM_FILE_END): Define.
14054 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14055 (arc_add_jli_section): New function.
14056 (jli_call_scan): Likewise.
14057 (arc_reorg): Call jli_call_scan.
14058 (arc_output_addsi): Remove 'S' from printing asm operand.
14059 (arc_is_jli_call_p): New function.
14060 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14062 (movhi_insn): Likewise.
14063 (movsi_insn): Likewise.
14064 (movsi_set_cc_insn): Likewise.
14065 (loadqi_update): Likewise.
14066 (load_zeroextendqisi_update): Likewise.
14067 (load_signextendqisi_update): Likewise.
14068 (loadhi_update): Likewise.
14069 (load_zeroextendhisi_update): Likewise.
14070 (load_signextendhisi_update): Likewise.
14071 (loadsi_update): Likewise.
14072 (loadsf_update): Likewise.
14073 (movsicc_insn): Likewise.
14074 (bset_insn): Likewise.
14075 (bxor_insn): Likewise.
14076 (bclr_insn): Likewise.
14077 (bmsk_insn): Likewise.
14078 (bicsi3_insn): Likewise.
14079 (cmpsi_cc_c_insn): Likewise.
14080 (movsi_ne): Likewise.
14081 (movsi_cond_exec): Likewise.
14082 (clrsbsi2): Likewise.
14083 (norm_f): Likewise.
14093 (call_i): Remove 'S' asm letter, add jli instruction.
14094 (call_value_i): Likewise.
14095 * config/arc/arc.op (mjli-always): New option.
14096 * config/arc/constraints.md (Cji): New constraint.
14097 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14099 (subsf3_fpx): Likewise.
14100 (mulsf3_fpx): Likewise.
14101 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14103 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14104 function attrbutes.
14105 * doc/invoke.texi (ARC): Document mjli-always option.
14107 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14109 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14110 avoid addition with 0 and use incw and decw where possible.
14112 2018-01-26 Richard Biener <rguenther@suse.de>
14114 PR tree-optimization/81082
14115 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14116 association if it requires casting to unsigned.
14117 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14118 from fold_plusminus_mult_expr to catch important cases late when
14119 range info is available.
14121 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14123 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14124 * configure.ac (hidden_linkonce): New test.
14125 * configure: Regenerate.
14126 * config.in: Regenerate.
14128 2018-01-26 Julia Koval <julia.koval@intel.com>
14130 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14131 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14132 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14133 _mm_mask_bitshuffle_epi64_mask): Fix type.
14134 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14135 USI_FTYPE_V4DI_V4DI_USI): Remove.
14136 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14137 __builtin_ia32_vpshufbitqmb256_mask,
14138 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14139 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14140 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14142 2018-01-26 Alan Modra <amodra@gmail.com>
14145 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14146 UNSPEC_VBPERMQ. Sort other unspecs.
14148 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14150 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14152 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14154 PR middle-end/83055
14155 * predict.c (drop_profile): Do not push/pop cfun; update also
14157 (handle_missing_profiles): Fix logic looking for zero profiles.
14159 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14161 PR middle-end/83977
14162 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14163 on functions with #pragma omp declare simd or functions with simd
14165 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14166 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14167 Remove trailing \n from warning_at calls.
14169 2018-01-25 Tom de Vries <tom@codesourcery.com>
14172 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14173 for neutered workers.
14175 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14178 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14179 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14181 2018-01-24 Jeff Law <law@redhat.com>
14184 * i386.c (get_probe_interval): Move to earlier point.
14185 (ix86_compute_frame_layout): If -fstack-clash-protection and
14186 the frame is larger than the probe interval, then use pushes
14187 to save registers rather than reg->mem moves.
14188 (ix86_expand_prologue): Remove conditional for int_registers_saved
14191 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14194 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14195 min/max for never referenced object.
14197 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14199 PR middle-end/83977
14200 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14202 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14203 attributes from DECL_ATTRIBUTES (decl) without affecting
14204 DECL_ATTRIBUTES (current_function_decl).
14205 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14206 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14208 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14210 PR tree-optimization/83979
14211 * fold-const.c (fold_comparison): Use constant_boolean_node
14212 instead of boolean_{true,false}_node.
14214 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14216 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14219 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14221 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14222 Simplify the clause that sets the length attribute.
14223 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14224 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14225 clause that sets the length attribute.
14226 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14228 2018-01-24 Tom de Vries <tom@codesourcery.com>
14231 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14232 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14233 Add strict parameter.
14234 (prevent_branch_around_nothing): Insert dummy insn between branch to
14235 label and label with no ptx insn inbetween.
14236 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14238 2018-01-24 Tom de Vries <tom@codesourcery.com>
14241 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14242 for neutered threads in warp.
14243 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14245 2018-01-24 Richard Biener <rguenther@suse.de>
14247 PR tree-optimization/83176
14248 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14251 2018-01-24 Richard Biener <rguenther@suse.de>
14253 PR tree-optimization/82819
14254 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14255 code generating pluses that are no-ops in the target precision.
14257 2018-01-24 Richard Biener <rguenther@suse.de>
14259 PR middle-end/84000
14260 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14262 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14264 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14265 to merge probabilities.
14266 * predict.c (probably_never_executed): Also mark as cold functions
14267 with global 0 profile and guessed local profile.
14268 * profile-count.c (profile_probability::combine_with_count): New
14270 * profile-count.h (profile_probability::operator*,
14271 profile_probability::operator*=, profile_probability::operator/,
14272 profile_probability::operator/=): Reduce precision to adjusted
14273 and set value to guessed on contradictory divisions.
14274 (profile_probability::combine_with_freq): Remove.
14275 (profile_probability::combine_wiht_count): Declare.
14276 (profile_count::force_nonzero):: Set to adjusted.
14277 (profile_count::probability_in):: Set quality to adjusted.
14278 * tree-ssa-tail-merge.c (replace_block_by): Use
14279 combine_with_count.
14281 2018-01-23 Andrew Waterman <andrew@sifive.com>
14282 Jim Wilson <jimw@sifive.com>
14284 * config/riscv/riscv.c (riscv_stack_boundary): New.
14285 (riscv_option_override): Set riscv_stack_boundary. Handle
14286 riscv_preferred_stack_boundary_arg.
14287 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14288 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14289 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14290 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14291 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14292 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14294 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14297 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14298 of struct ix86_frame.
14299 (ix86_expand_epilogue): Likewise. Add a local variable for
14300 the reg_save_offset field in struct ix86_frame.
14302 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14304 PR tree-optimization/82604
14305 * tree-loop-distribution.c (enum partition_kind): New enum item
14306 PKIND_PARTIAL_MEMSET.
14307 (partition_builtin_p): Support above new enum item.
14308 (generate_code_for_partition): Ditto.
14309 (compute_access_range): Differentiate cases that equality can be
14310 proven at all loops, the innermost loops or no loops.
14311 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14312 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14313 (finalize_partitions, distribute_loop): Don't fuse partition of
14314 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14315 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14316 parloop is enabled.
14318 2018-01-23 Martin Liska <mliska@suse.cz>
14320 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14321 order to ignore the predictor.
14322 (PRED_POLYMORPHIC_CALL): Likewise.
14323 (PRED_RECURSIVE_CALL): Likewise.
14325 2018-01-23 Martin Liska <mliska@suse.cz>
14327 * tree-profile.c (tree_profiling): Print function header to
14328 aware reader which function we are working on.
14329 * value-prof.c (gimple_find_values_to_profile): Do not print
14330 not interesting value histograms.
14332 2018-01-23 Martin Liska <mliska@suse.cz>
14334 * profile-count.h (enum profile_quality): Add
14335 profile_uninitialized as the first value. Do not number values
14336 as they are zero based.
14337 (profile_count::verify): Update sanity check.
14338 (profile_probability::verify): Likewise.
14340 2018-01-23 Nathan Sidwell <nathan@acm.org>
14342 * doc/invoke.texi (ffor-scope): Deprecate.
14344 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14346 PR tree-optimization/83510
14347 * domwalk.c (set_all_edges_as_executable): New function.
14348 (dom_walker::dom_walker): Convert bool param
14349 "skip_unreachable_blocks" to enum reachability. Move setup of
14350 edge flags to set_all_edges_as_executable and only do it when
14351 reachability is REACHABLE_BLOCKS.
14352 * domwalk.h (enum dom_walker::reachability): New enum.
14353 (dom_walker::dom_walker): Convert bool param
14354 "skip_unreachable_blocks" to enum reachability.
14355 (set_all_edges_as_executable): New decl.
14356 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14357 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14359 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14360 but converting true to REACHABLE_BLOCKS.
14361 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14363 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14364 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14365 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14367 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14368 if check_all_array_refs will be called.
14370 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14372 * tree.c (selftest::test_location_wrappers): Add more test
14375 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14377 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14378 (selftest::test_bit_in_range): Likewise.
14380 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14383 * doc/sourcebuild.texi (vect_float): Say that the selector
14384 only describes the situation when -funsafe-math-optimizations is on.
14385 (vect_float_strict): Document.
14387 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14389 PR tree-optimization/83965
14390 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14391 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14392 instead of checking only for a reduction.
14393 (vect_recog_widen_sum_pattern): Likewise.
14395 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14397 * predict.c (probably_never_executed): Only use precise profile info.
14398 (compute_function_frequency): Skip after inlining hack since we now
14399 have quality checking.
14401 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14403 * profile-count.h (profile_probability::very_unlikely,
14404 profile_probability::unlikely, profile_probability::even): Set
14405 precision to guessed.
14407 2018-01-23 Richard Biener <rguenther@suse.de>
14409 PR tree-optimization/83963
14410 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14411 Properly terminate dominator walk when crossing the exit edge not
14412 when visiting its source block.
14414 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14417 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14418 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14420 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14422 PR tree-optimization/83957
14423 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14424 semicolon after for body surrounded by braces.
14426 PR tree-optimization/83081
14427 * profile-count.h (profile_probability::split): New method.
14428 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14429 Use profile_probability::split.
14430 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14431 when splitting a single conditional jump into 2.
14433 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14435 PR tree-optimization/69452
14436 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14441 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14442 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14443 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14445 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14447 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14449 * config/rl78/rl78.md (movdi): New define_expand.
14450 * config/rl78/rl78.c (rl78_split_movdi): New function.
14452 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14455 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14457 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14458 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14459 128-bit to produce an UNSPEC move to get the double word with the
14460 signbit and then a shift directly to do signbit.
14461 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14462 implementation with a new version that just does either a direct
14463 move or a regular move. Move memory interface to separate insns.
14464 Move insns so they are next to the expander.
14465 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14466 with signbit move. Split big and little endian case.
14467 (signbit<mode>2_dm_mem_le): Likewise.
14468 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14469 (signbit<mode>2_dm2): Likewise.
14471 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14473 * config/rl78/rl78.md (anddi3): New define_expand.
14475 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14477 * config/rl78/rl78.md (umindi3): New define_expand.
14479 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14481 * config/rl78/rl78.md (smindi3): New define_expand.
14483 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14485 * config/rl78/rl78.md (smaxdi3): New define_expand.
14487 2018-01-22 Carl Love <cel@us.ibm.com>
14489 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14490 LVX_V1TI): Add macro expansion.
14491 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14492 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14493 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14494 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14495 Change check to determine if the instruction is a byte reversing
14496 entry. Fix typo in comment.
14497 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14498 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14499 Add def_builtin calls for new builtins.
14500 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14501 Add define_insn expansion.
14503 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14505 * config/rl78/rl78.md (umaxdi3): New define_expand.
14507 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14509 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14510 for non-QImode registers.
14512 2018-01-22 Richard Biener <rguenther@suse.de>
14514 PR tree-optimization/83963
14515 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14516 including the loop exit block.
14517 (scop_detection::merge_sese): Likewise.
14518 (scop_detection::add_scop): Do it here instead.
14520 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14522 * doc/sourcebuild.texi (arm_softfloat): Document.
14524 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14527 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14528 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14529 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14531 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14532 David Edelsohn <dje.gcc@gmail.com>
14535 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14536 Change "crset eq" to "crset 2".
14537 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14538 (*call_indirect_aix<mode>_nospec): Likewise.
14539 (*call_value_indirect_aix<mode>_nospec): Likewise.
14540 (*call_indirect_elfv2<mode>_nospec): Likewise.
14541 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14542 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14543 change assembly output from . to $.
14544 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14545 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14546 (*tablejump<mode>_internal1_nospec): Likewise.
14548 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14551 * config/sh/sh_optimize_sett_clrt.cc:
14552 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14554 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14556 PR tree-optimization/83940
14557 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14558 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14559 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14560 use it to pass back the definition type.
14561 (vect_check_store_rhs): Likewise rhs_dt_out.
14562 (vect_build_gather_load_calls): Add a mask_dt argument and use
14563 it instead of a call to vect_is_simple_use.
14564 (vectorizable_store): Update calls to vect_check_load_store_mask
14565 and vect_check_store_rhs. Use the dt returned by the latter instead
14566 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14567 instead of calls to vect_is_simple_use. Pass the scalar rather
14568 than the vector operand to vect_is_simple_use when handling
14569 second and subsequent copies of an rhs value.
14570 (vectorizable_load): Update calls to vect_check_load_store_mask
14571 and vect_build_gather_load_calls. Use the cached mask_dt and
14572 gs_info.offset_dt instead of calls to vect_is_simple_use.
14574 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14576 PR middle-end/83945
14577 * tree-emutls.c: Include gimplify.h.
14578 (lower_emutls_2): New function.
14579 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14580 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14581 it before further processing.
14584 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14585 UINTVAL (trueop1) instead of INTVAL (op1).
14587 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14591 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14592 INCOMING_FRAME_SP_OFFSET if not defined.
14593 (scan_trace): Add ENTRY argument. If true and
14594 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14595 emit a note to adjust the CFA offset.
14596 (create_cfi_notes): Adjust scan_trace callers.
14597 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14598 INCOMING_FRAME_SP_OFFSET in the CIE.
14599 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14600 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14602 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14603 * doc/tm.texi: Regenerated.
14605 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14607 PR rtl-optimization/83147
14608 * lra-constraints.c (remove_inheritance_pseudos): Use
14609 lra_substitute_pseudo_within_insn.
14611 2018-01-19 Tom de Vries <tom@codesourcery.com>
14612 Cesar Philippidis <cesar@codesourcery.com>
14615 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14617 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14620 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14621 spaces for function labels.
14623 2018-01-19 Martin Liska <mliska@suse.cz>
14625 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14626 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14627 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14628 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14629 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14630 (PRED_CONST_RETURN): Change from 69 to 65.
14631 (PRED_NULL_RETURN): Change from 91 to 71.
14632 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14633 (PRED_LOOP_GUARD): Change from 66 to 73.
14635 2018-01-19 Martin Liska <mliska@suse.cz>
14637 * predict.c (predict_insn_def): Add new assert.
14638 (struct branch_predictor): Change type to signed integer.
14639 (test_prediction_value_range): Amend test to cover
14640 PROB_UNINITIALIZED.
14641 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14642 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14643 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14644 (PRED_LOOP_IV_COMPARE): Likewise.
14645 * predict.h (PROB_UNINITIALIZED): Define new constant.
14647 2018-01-19 Martin Liska <mliska@suse.cz>
14649 * predict.c (dump_prediction): Add new format for
14650 analyze_brprob.py script which is enabled with -details
14652 * profile-count.h (precise_p): New function.
14654 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14656 PR tree-optimization/83922
14657 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14658 there are no statements that need masking.
14659 (vect_active_double_reduction_p): New function.
14660 (vect_analyze_loop_operations): Use it when handling phis that
14661 are not in the loop header.
14663 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14665 PR tree-optimization/83914
14666 * tree-vect-loop.c (vectorizable_induction): Don't convert
14667 init_expr or apply the peeling adjustment for inductions
14668 that are nested within the vectorized loop.
14670 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14672 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14675 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14679 * function.h (gimplify_parameters): Add gimple_seq * argument.
14680 * function.c: Include gimple.h and options.h.
14681 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14682 for the added local temporaries if needed.
14683 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14684 if there are any parameter cleanups, wrap whole body into a
14685 try/finally with the cleanups.
14687 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14690 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14691 Use GET_MODE_CLASS for scalar floating point.
14693 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14697 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14698 Fix call of call_cgraph_insertion_hooks.
14700 2018-01-18 Martin Sebor <msebor@redhat.com>
14702 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14704 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14707 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14710 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14713 * common.opt: (-ffile-prefix-map): New option.
14714 * opts.c (common_handle_option): Defer it.
14715 * opts-global.c (handle_common_deferred_options): Handle it.
14716 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14717 * file-prefix-map.h: New file.
14718 (remap_debug_filename, add_debug_prefix_map): ...here.
14719 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14720 * final.c (debug_prefix_map, add_debug_prefix_map
14721 remap_debug_filename): Move to...
14722 * file-prefix-map.c: New file.
14723 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14724 generalize, get rid of alloca(), use strrchr() instead of strchr().
14725 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14726 Implement in terms of add_prefix_map().
14727 (remap_macro_filename, remap_debug_filename): Implement in term of
14729 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14730 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14731 * dbxout.c: Include file-prefix-map.h.
14732 * varasm.c: Likewise.
14733 * vmsdbgout.c: Likewise.
14734 * xcoffout.c: Likewise.
14735 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14736 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14737 * doc/invoke.texi (-ffile-prefix-map): Document.
14738 (-fdebug-prefix-map): Update description.
14740 2018-01-18 Martin Liska <mliska@suse.cz>
14742 * config/i386/i386.c (indirect_thunk_name): Document that also
14744 (output_indirect_thunk): Document why both instructions
14745 (pause and lfence) are generated.
14747 2018-01-18 Richard Biener <rguenther@suse.de>
14749 PR tree-optimization/83887
14750 * graphite-scop-detection.c
14751 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14752 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14753 (scop_detection::merge_sese): Re-implement with a flood-fill
14754 algorithm that properly finds a SESE region if it exists.
14756 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14759 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14760 pointer_diff optimizations use view_convert instead of convert.
14762 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14764 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14765 Generate different code for -mno-speculate-indirect-jumps.
14766 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14767 (*call_indirect_aix<mode>): Disable for
14768 -mno-speculate-indirect-jumps.
14769 (*call_indirect_aix<mode>_nospec): New define_insn.
14770 (*call_value_indirect_aix<mode>): Disable for
14771 -mno-speculate-indirect-jumps.
14772 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14773 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14774 -mno-speculate-indirect-jumps.
14775 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14777 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14779 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14780 long double type, set the flags for noting the default long double
14781 type, even if we don't pass or return a long double type.
14783 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14786 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14789 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14791 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14792 support for merge[hl].
14793 (fold_mergehl_helper): New helper function.
14794 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14795 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14796 (altivec_vmrglw_direct): Add xxmrglw insn.
14798 2018-01-17 Andrew Waterman <andrew@sifive.com>
14800 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14801 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14803 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14806 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14807 call the lto_location_cache before reading the
14808 DECL_SOURCE_LOCATION of the types.
14810 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14811 Richard Sandiford <richard.sandiford@linaro.org>
14813 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14814 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14815 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14816 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14817 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14819 * config/aarch64/constraints.md (aarch64_movti_operand):
14821 * config/aarch64/predicates.md (Uti): Add new constraint.
14823 2018-01-17 Carl Love <cel@us.ibm.com>
14825 * config/rs6000/vsx.md (define_expand xl_len_r,
14826 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14827 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14829 (define_expand, define_insn): Move the shift left from the
14830 define_insn to the define_expand for lxvl and stxvl instructions.
14831 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14832 and XL_LEN_R definitions to PURE.
14834 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14836 * config/i386/i386.c (indirect_thunk_name): Declare regno
14837 as unsigned int. Compare regno with INVALID_REGNUM.
14838 (output_indirect_thunk): Ditto.
14839 (output_indirect_thunk_function): Ditto.
14840 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14841 in the call to output_indirect_thunk_function.
14843 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14845 PR middle-end/83884
14846 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14847 rather than the size of inner_type to determine the stack slot size
14848 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14850 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14853 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14856 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14858 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14859 endian Linux systems to optionally enable multilibs for selecting
14860 the long double type if the user configured an explicit type.
14861 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14862 have no long double multilibs if not defined.
14863 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14864 warn if the user used -mabi={ieee,ibm}longdouble and we built
14865 multilibs for long double.
14866 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14867 appropriate multilib option.
14868 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14870 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14871 for building long double multilibs.
14872 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14874 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14876 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14879 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14881 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14884 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14887 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14890 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14892 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14893 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14895 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14897 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14898 different rtl trees depending on TARGET_64BIT.
14899 (rs6000_gen_lvx): Likewise.
14901 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14903 * config/visium/visium.md (nop): Tweak comment.
14904 (hazard_nop): Likewise.
14906 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14908 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14909 -mspeculate-indirect-jumps.
14910 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14911 for -mno-speculate-indirect-jumps.
14912 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14913 (*call_value_indirect_elfv2<mode>): Disable for
14914 -mno-speculate-indirect-jumps.
14915 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14916 (indirect_jump): Emit different RTL for
14917 -mno-speculate-indirect-jumps.
14918 (*indirect_jump<mode>): Disable for
14919 -mno-speculate-indirect-jumps.
14920 (*indirect_jump<mode>_nospec): New define_insn.
14921 (tablejump): Emit different RTL for
14922 -mno-speculate-indirect-jumps.
14923 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14924 (tablejumpsi_nospec): New define_expand.
14925 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14926 (tablejumpdi_nospec): New define_expand.
14927 (*tablejump<mode>_internal1): Disable for
14928 -mno-speculate-indirect-jumps.
14929 (*tablejump<mode>_internal1_nospec): New define_insn.
14930 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14933 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14935 * caller-save.c (insert_save): Drop unnecessary parameter. All
14938 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14939 Richard Biener <rguenth@suse.de>
14942 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14943 return early, inline manually is_gimple_sizepos. Make sure if we
14944 call gimplify_expr we don't end up with a gimple constant.
14945 * tree.c (variably_modified_type_p): Don't return true for
14946 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14947 * gimplify.h (is_gimple_sizepos): Remove.
14949 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14951 PR tree-optimization/83857
14952 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14953 vectorizable_live_operation for pure SLP statements.
14954 (vectorizable_live_operation): Handle PHIs.
14956 2018-01-16 Richard Biener <rguenther@suse.de>
14958 PR tree-optimization/83867
14959 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14960 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14962 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14965 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14966 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14967 If off is not INTEGER_CST, issue a may not be aligned warning
14968 rather than isn't aligned. Use isn%'t rather than isn't.
14969 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14971 <case MULT_EXPR>: Improve the case when bottom and one of the
14972 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14973 operand, in that case check if the other operand is multiple of
14974 bottom divided by the INTEGER_CST operand.
14976 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14979 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14980 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14981 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14982 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14983 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14984 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14985 (pa_function_arg_size): New function.
14987 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14989 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14990 in a separate statement.
14992 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14994 PR tree-optimization/83847
14995 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14996 group gathers and scatters.
14998 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15000 PR rtl-optimization/86620
15001 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15003 PR rtl-optimization/83213
15004 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15005 to last if both are JUMP_INSNs.
15007 PR tree-optimization/83843
15008 * gimple-ssa-store-merging.c
15009 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15010 store_immediate_info for bswap/nop orig_stores.
15012 2018-01-15 Andrew Waterman <andrew@sifive.com>
15014 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15016 <UDIV>: Increase cost if !TARGET_DIV.
15018 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15020 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15021 (define_attr "cr_logical_3op"): New.
15022 (cceq_ior_compare): Adjust.
15023 (cceq_ior_compare_complement): Adjust.
15024 (*cceq_rev_compare): Adjust.
15025 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15026 (is_cracked_insn): Adjust.
15027 (insn_must_be_first_in_group): Adjust.
15028 * config/rs6000/40x.md: Adjust.
15029 * config/rs6000/440.md: Adjust.
15030 * config/rs6000/476.md: Adjust.
15031 * config/rs6000/601.md: Adjust.
15032 * config/rs6000/603.md: Adjust.
15033 * config/rs6000/6xx.md: Adjust.
15034 * config/rs6000/7450.md: Adjust.
15035 * config/rs6000/7xx.md: Adjust.
15036 * config/rs6000/8540.md: Adjust.
15037 * config/rs6000/cell.md: Adjust.
15038 * config/rs6000/e300c2c3.md: Adjust.
15039 * config/rs6000/e500mc.md: Adjust.
15040 * config/rs6000/e500mc64.md: Adjust.
15041 * config/rs6000/e5500.md: Adjust.
15042 * config/rs6000/e6500.md: Adjust.
15043 * config/rs6000/mpc.md: Adjust.
15044 * config/rs6000/power4.md: Adjust.
15045 * config/rs6000/power5.md: Adjust.
15046 * config/rs6000/power6.md: Adjust.
15047 * config/rs6000/power7.md: Adjust.
15048 * config/rs6000/power8.md: Adjust.
15049 * config/rs6000/power9.md: Adjust.
15050 * config/rs6000/rs64.md: Adjust.
15051 * config/rs6000/titan.md: Adjust.
15053 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15055 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15056 ix86_indirect_branch_register logic.
15058 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15060 * config/i386/constraints.md (Bs): Update
15061 ix86_indirect_branch_register check. Don't check
15062 ix86_indirect_branch_register with GOT_memory_operand.
15064 * config/i386/predicates.md (GOT_memory_operand): Don't check
15065 ix86_indirect_branch_register here.
15066 (GOT32_symbol_operand): Likewise.
15068 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15070 * config/i386/predicates.md (constant_call_address_operand):
15071 Rewrite ix86_indirect_branch_register logic.
15072 (sibcall_insn_operand): Likewise.
15074 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15076 * config/i386/constraints.md (Bs): Replace
15077 ix86_indirect_branch_thunk_register with
15078 ix86_indirect_branch_register.
15080 * config/i386/i386.md (indirect_jump): Likewise.
15081 (tablejump): Likewise.
15082 (*sibcall_memory): Likewise.
15083 (*sibcall_value_memory): Likewise.
15084 Peepholes of indirect call and jump via memory: Likewise.
15085 * config/i386/i386.opt: Likewise.
15086 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15087 (GOT_memory_operand): Likewise.
15088 (call_insn_operand): Likewise.
15089 (sibcall_insn_operand): Likewise.
15090 (GOT32_symbol_operand): Likewise.
15092 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15094 PR middle-end/83837
15095 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15096 type rather than type addr's type points to.
15097 (expand_omp_atomic_mutex): Likewise.
15098 (expand_omp_atomic): Likewise.
15100 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15103 * config/i386/i386.c (output_indirect_thunk_function): Use
15104 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15105 for __x86_return_thunk.
15107 2018-01-15 Richard Biener <rguenther@suse.de>
15109 PR middle-end/83850
15110 * expmed.c (extract_bit_field_1): Fix typo.
15112 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15115 * config/arm/iterators.md (VF): New mode iterator.
15116 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15117 Remove integer-related logic from pattern.
15118 (neon_vabd<mode>_3): Likewise.
15120 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15122 PR middle-end/82694
15123 * common.opt (fstrict-overflow): No longer an alias.
15124 (fwrapv-pointer): New option.
15125 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15126 also for pointer types based on flag_wrapv_pointer.
15127 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15128 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15129 opts->x_flag_wrapv got set.
15130 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15131 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15132 POINTER_TYPE_OVERFLOW_UNDEFINED.
15133 * match.pd: Likewise in address comparison pattern.
15134 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15136 2018-01-15 Richard Biener <rguenther@suse.de>
15139 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15140 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15141 Reset type names to their identifier if their TYPE_DECL doesn't
15142 have linkage (and thus is used for ODR and devirt).
15143 (save_debug_info_for_decl): Remove.
15144 (save_debug_info_for_type): Likewise.
15145 (add_tree_to_fld_list): Adjust.
15146 * tree-pretty-print.c (dump_generic_node): Make dumping of
15147 type names more robust.
15149 2018-01-15 Richard Biener <rguenther@suse.de>
15151 * BASE-VER: Bump to 8.0.1.
15153 2018-01-14 Martin Sebor <msebor@redhat.com>
15156 * builtins.c (check_access): Avoid warning when the no-warning bit
15159 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15161 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15162 * ira-color (allocno_hard_regs_compare): Likewise.
15164 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15167 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15168 Use .pushsection/.popsection.
15170 2018-01-14 Martin Sebor <msebor@redhat.com>
15173 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15175 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15177 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15178 entry from extra_headers.
15179 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15180 extra_headers, make the list bitwise identical to the i?86-*-* one.
15182 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15184 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15185 -mcmodel=large with -mindirect-branch=thunk,
15186 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15187 -mfunction-return=thunk-extern.
15188 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15189 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15190 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15192 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15194 * config/i386/i386.c (print_reg): Print the name of the full
15195 integer register without '%'.
15196 (ix86_print_operand): Handle 'V'.
15197 * doc/extend.texi: Document 'V' modifier.
15199 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15201 * config/i386/constraints.md (Bs): Disallow memory operand for
15202 -mindirect-branch-register.
15204 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15205 (GOT_memory_operand): Likewise.
15206 (call_insn_operand): Likewise.
15207 (sibcall_insn_operand): Likewise.
15208 (GOT32_symbol_operand): Likewise.
15209 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15210 for -mindirect-branch-register.
15211 (tablejump): Likewise.
15212 (*sibcall_memory): Likewise.
15213 (*sibcall_value_memory): Likewise.
15214 Disallow peepholes of indirect call and jump via memory for
15215 -mindirect-branch-register.
15216 (*call_pop): Replace m with Bw.
15217 (*call_value_pop): Likewise.
15218 (*sibcall_pop_memory): Replace m with Bs.
15219 * config/i386/i386.opt (mindirect-branch-register): New option.
15220 * doc/invoke.texi: Document -mindirect-branch-register option.
15222 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15224 * config/i386/i386-protos.h (ix86_output_function_return): New.
15225 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15226 set function_return_type.
15227 (indirect_thunk_name): Add ret_p to indicate thunk for function
15229 (output_indirect_thunk_function): Pass false to
15230 indirect_thunk_name.
15231 (ix86_output_indirect_branch_via_reg): Likewise.
15232 (ix86_output_indirect_branch_via_push): Likewise.
15233 (output_indirect_thunk_function): Create alias for function
15234 return thunk if regno < 0.
15235 (ix86_output_function_return): New function.
15236 (ix86_handle_fndecl_attribute): Handle function_return.
15237 (ix86_attribute_table): Add function_return.
15238 * config/i386/i386.h (machine_function): Add
15239 function_return_type.
15240 * config/i386/i386.md (simple_return_internal): Use
15241 ix86_output_function_return.
15242 (simple_return_internal_long): Likewise.
15243 * config/i386/i386.opt (mfunction-return=): New option.
15244 (indirect_branch): Mention -mfunction-return=.
15245 * doc/extend.texi: Document function_return function attribute.
15246 * doc/invoke.texi: Document -mfunction-return= option.
15248 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15250 * config/i386/i386-opts.h (indirect_branch): New.
15251 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15252 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15253 with local indirect jump when converting indirect call and jump.
15254 (ix86_set_indirect_branch_type): New.
15255 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15256 (indirectlabelno): New.
15257 (indirect_thunk_needed): Likewise.
15258 (indirect_thunk_bnd_needed): Likewise.
15259 (indirect_thunks_used): Likewise.
15260 (indirect_thunks_bnd_used): Likewise.
15261 (INDIRECT_LABEL): Likewise.
15262 (indirect_thunk_name): Likewise.
15263 (output_indirect_thunk): Likewise.
15264 (output_indirect_thunk_function): Likewise.
15265 (ix86_output_indirect_branch_via_reg): Likewise.
15266 (ix86_output_indirect_branch_via_push): Likewise.
15267 (ix86_output_indirect_branch): Likewise.
15268 (ix86_output_indirect_jmp): Likewise.
15269 (ix86_code_end): Call output_indirect_thunk_function if needed.
15270 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15272 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15273 (ix86_attribute_table): Add indirect_branch.
15274 * config/i386/i386.h (machine_function): Add indirect_branch_type
15275 and has_local_indirect_jump.
15276 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15278 (tablejump): Likewise.
15279 (*indirect_jump): Use ix86_output_indirect_jmp.
15280 (*tablejump_1): Likewise.
15281 (simple_return_indirect_internal): Likewise.
15282 * config/i386/i386.opt (mindirect-branch=): New option.
15283 (indirect_branch): New.
15286 (thunk-inline): Likewise.
15287 (thunk-extern): Likewise.
15288 * doc/extend.texi: Document indirect_branch function attribute.
15289 * doc/invoke.texi: Document -mindirect-branch= option.
15291 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15294 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15296 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15298 * ipa-inline.c (want_inline_small_function_p): Return false if
15299 inlining has already failed with CIF_FINAL_ERROR.
15300 (update_caller_keys): Call want_inline_small_function_p before
15302 (update_callee_keys): Likewise.
15304 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15306 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15308 (rs6000_quadword_masked_address_p): Likewise.
15309 (quad_aligned_load_p): Likewise.
15310 (quad_aligned_store_p): Likewise.
15311 (const_load_sequence_p): Add comment to describe the outer-most loop.
15312 (mimic_memory_attributes_and_flags): New function.
15313 (rs6000_gen_stvx): Likewise.
15314 (replace_swapped_aligned_store): Likewise.
15315 (rs6000_gen_lvx): Likewise.
15316 (replace_swapped_aligned_load): Likewise.
15317 (replace_swapped_load_constant): Capitalize argument name in
15318 comment describing this function.
15319 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15320 and stores that access quad-word aligned addresses and replace
15321 with stvx or lvx instructions when appropriate.
15322 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15323 New function prototype.
15324 (rs6000_quadword_masked_address_p): Likewise.
15325 (rs6000_gen_lvx): Likewise.
15326 (rs6000_gen_stvx): Likewise.
15327 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15328 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15329 when memory address is aligned.
15330 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15331 this split to select lvx instruction when memory address is aligned.
15332 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15333 instruction when memory address is aligned.
15334 (*vsx_le_perm_load_v16qi): Likewise.
15335 (four unnamed splitters): Modify to select the stvx instruction
15336 when memory is aligned.
15338 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15340 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15341 which appears in the queue multiple times.
15343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15344 Alan Hayward <alan.hayward@arm.com>
15345 David Sherwood <david.sherwood@arm.com>
15347 * tree-vectorizer.h (vec_lower_bound): New structure.
15348 (_loop_vec_info): Add check_nonzero and lower_bounds.
15349 (LOOP_VINFO_CHECK_NONZERO): New macro.
15350 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15351 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15352 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15353 fields. Make seg_len the distance travelled, not including the
15355 (dr_direction_indicator): Declare.
15356 (dr_zero_step_indicator): Likewise.
15357 (dr_known_forward_stride_p): Likewise.
15358 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15360 (runtime_alias_check_p): Allow runtime alias checks with
15362 (operator ==): Compare access_size and align.
15363 (prune_runtime_alias_test_list): Rework for new distinction between
15364 the access_size and seg_len.
15365 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15367 (get_segment_min_max): New function.
15368 (create_intersect_range_checks): Use it.
15369 (dr_step_indicator): New function.
15370 (dr_direction_indicator): Likewise.
15371 (dr_zero_step_indicator): Likewise.
15372 (dr_known_forward_stride_p): Likewise.
15373 * tree-loop-distribution.c (data_ref_segment_size): Return
15374 DR_STEP * (niters - 1).
15375 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15377 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15378 (vect_preserves_scalar_order_p): New function, split out from...
15379 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15380 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15381 (vect_vfa_access_size): New function.
15382 (vect_vfa_align): Likewise.
15383 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15384 (dump_lower_bound): New function.
15385 (vect_check_lower_bound): Likewise.
15386 (vect_small_gap_p): Likewise.
15387 (vectorizable_with_step_bound_p): Likewise.
15388 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15389 depencies if the vectorization factor is 1. Convert the checks
15390 for nonzero steps into checks on the bounds of DR_STEP. Try using
15391 a bunds check for variable steps if the minimum required step is
15392 relatively small. Update calls to the dr_with_seg_len
15393 constructor and to vect_compile_time_alias.
15394 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15396 (vect_loop_versioning): Call it.
15397 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15399 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15402 Alan Hayward <alan.hayward@arm.com>
15403 David Sherwood <david.sherwood@arm.com>
15405 * doc/sourcebuild.texi (vect_scatter_store): Document.
15406 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15408 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15410 * genopinit.c (main): Add supports_vec_scatter_store and
15411 supports_vec_scatter_store_cached to target_optabs.
15412 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15413 IFN_MASK_SCATTER_STORE.
15414 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15416 * internal-fn.h (internal_store_fn_p): Declare.
15417 (internal_fn_stored_value_index): Likewise.
15418 * internal-fn.c (scatter_store_direct): New macro.
15419 (expand_scatter_store_optab_fn): New function.
15420 (direct_scatter_store_optab_supported_p): New macro.
15421 (internal_store_fn_p): New function.
15422 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15423 IFN_MASK_SCATTER_STORE.
15424 (internal_fn_mask_index): Likewise.
15425 (internal_fn_stored_value_index): New function.
15426 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15427 for scatter stores.
15428 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15429 * optabs-query.c (supports_vec_scatter_store_p): New function.
15430 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15431 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15432 true for scatter stores.
15433 (vect_gather_scatter_fn_p): Handle scatter stores too.
15434 (vect_check_gather_scatter): Consider using scatter stores if
15435 supports_vec_scatter_store_p.
15436 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15437 scatter stores too.
15438 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15439 internal_fn_stored_value_index.
15440 (check_load_store_masking): Handle scatter stores too.
15441 (vect_get_store_rhs): Make public.
15442 (vectorizable_call): Use internal_store_fn_p.
15443 (vectorizable_store): Handle scatter store internal functions.
15444 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15445 when deciding whether the end of the group has been reached.
15446 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15447 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15448 (mask_scatter_store<mode>): New insns.
15450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15451 Alan Hayward <alan.hayward@arm.com>
15452 David Sherwood <david.sherwood@arm.com>
15454 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15455 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15456 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15458 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15459 Use vect_truncate_gather_scatter_offset if we can't treat the
15460 operation as a normal gather load or scatter store.
15461 (get_group_load_store_type): Take the gather_scatter_info
15462 as argument. Try using a gather load or scatter store for
15463 single-element groups.
15464 (get_load_store_type): Update calls to get_group_load_store_type
15465 and vect_use_strided_gather_scatters_p.
15467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15468 Alan Hayward <alan.hayward@arm.com>
15469 David Sherwood <david.sherwood@arm.com>
15471 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15472 optional tree argument.
15473 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15475 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15476 but continue to use the current value as a fallback.
15477 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15478 to compare the updates.
15479 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15480 (get_load_store_type): Use it when handling a strided access.
15481 (vect_get_strided_load_store_ops): New function.
15482 (vect_get_data_ptr_increment): Likewise.
15483 (vectorizable_load): Handle strided gather loads. Always pass
15484 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15487 Alan Hayward <alan.hayward@arm.com>
15488 David Sherwood <david.sherwood@arm.com>
15490 * doc/md.texi (gather_load@var{m}): Document.
15491 (mask_gather_load@var{m}): Likewise.
15492 * genopinit.c (main): Add supports_vec_gather_load and
15493 supports_vec_gather_load_cached to target_optabs.
15494 * optabs-tree.c (init_tree_optimization_optabs): Use
15495 ggc_cleared_alloc to allocate target_optabs.
15496 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15497 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15499 * internal-fn.h (internal_load_fn_p): Declare.
15500 (internal_gather_scatter_fn_p): Likewise.
15501 (internal_fn_mask_index): Likewise.
15502 (internal_gather_scatter_fn_supported_p): Likewise.
15503 * internal-fn.c (gather_load_direct): New macro.
15504 (expand_gather_load_optab_fn): New function.
15505 (direct_gather_load_optab_supported_p): New macro.
15506 (direct_internal_fn_optab): New function.
15507 (internal_load_fn_p): Likewise.
15508 (internal_gather_scatter_fn_p): Likewise.
15509 (internal_fn_mask_index): Likewise.
15510 (internal_gather_scatter_fn_supported_p): Likewise.
15511 * optabs-query.c (supports_at_least_one_mode_p): New function.
15512 (supports_vec_gather_load_p): Likewise.
15513 * optabs-query.h (supports_vec_gather_load_p): Declare.
15514 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15515 and memory_type field.
15516 (NUM_PATTERNS): Bump to 15.
15517 * tree-vect-data-refs.c: Include internal-fn.h.
15518 (vect_gather_scatter_fn_p): New function.
15519 (vect_describe_gather_scatter_call): Likewise.
15520 (vect_check_gather_scatter): Try using internal functions for
15521 gather loads. Recognize existing calls to a gather load function.
15522 (vect_analyze_data_refs): Consider using gather loads if
15523 supports_vec_gather_load_p.
15524 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15525 (vect_get_gather_scatter_offset_type): Likewise.
15526 (vect_convert_mask_for_vectype): Likewise.
15527 (vect_add_conversion_to_patterm): Likewise.
15528 (vect_try_gather_scatter_pattern): Likewise.
15529 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15530 (vect_vect_recog_func_ptrs): Add it.
15531 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15532 internal_fn_mask_index and internal_gather_scatter_fn_p.
15533 (check_load_store_masking): Take the gather_scatter_info as an
15534 argument and handle gather loads.
15535 (vect_get_gather_scatter_ops): New function.
15536 (vectorizable_call): Check internal_load_fn_p.
15537 (vectorizable_load): Likewise. Handle gather load internal
15539 (vectorizable_store): Update call to check_load_store_masking.
15540 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15541 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15542 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15543 (aarch64_gather_scale_operand_d): New predicates.
15544 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15545 (mask_gather_load<mode>): New insns.
15547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15548 Alan Hayward <alan.hayward@arm.com>
15549 David Sherwood <david.sherwood@arm.com>
15551 * optabs.def (fold_left_plus_optab): New optab.
15552 * doc/md.texi (fold_left_plus_@var{m}): Document.
15553 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15554 * internal-fn.c (fold_left_direct): Define.
15555 (expand_fold_left_optab_fn): Likewise.
15556 (direct_fold_left_optab_supported_p): Likewise.
15557 * fold-const-call.c (fold_const_fold_left): New function.
15558 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15559 * tree-parloops.c (valid_reduction_p): New function.
15560 (gather_scalar_reductions): Use it.
15561 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15562 (vect_finish_replace_stmt): Declare.
15563 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15564 (needs_fold_left_reduction_p): New function, split out from...
15565 (vect_is_simple_reduction): ...here. Accept reductions that
15566 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15567 (vect_force_simple_reduction): Also store the reduction type in
15568 the assignment's STMT_VINFO_REDUC_TYPE.
15569 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15570 (merge_with_identity): New function.
15571 (vect_expand_fold_left): Likewise.
15572 (vectorize_fold_left_reduction): Likewise.
15573 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15574 scalar phi in place for it. Check for target support and reject
15575 cases that would reassociate the operation. Defer the transform
15576 phase to vectorize_fold_left_reduction.
15577 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15578 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15579 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15583 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15584 call to ifc_temp_var.
15586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15587 Alan Hayward <alan.hayward@arm.com>
15588 David Sherwood <david.sherwood@arm.com>
15590 * target.def (legitimize_address_displacement): Take the original
15591 offset as a poly_int.
15592 * targhooks.h (default_legitimize_address_displacement): Update
15594 * targhooks.c (default_legitimize_address_displacement): Likewise.
15595 * doc/tm.texi: Regenerate.
15596 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15597 as an argument, moving assert of ad->disp == ad->disp_term to...
15598 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15599 Try calling targetm.legitimize_address_displacement before expanding
15600 the address rather than afterwards, and adjust for the new interface.
15601 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15602 Match the new hook interface. Handle SVE addresses.
15603 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15604 new hook interface.
15606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15608 * Makefile.in (OBJS): Add early-remat.o.
15609 * target.def (select_early_remat_modes): New hook.
15610 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15611 * doc/tm.texi: Regenerate.
15612 * targhooks.h (default_select_early_remat_modes): Declare.
15613 * targhooks.c (default_select_early_remat_modes): New function.
15614 * timevar.def (TV_EARLY_REMAT): New timevar.
15615 * passes.def (pass_early_remat): New pass.
15616 * tree-pass.h (make_pass_early_remat): Declare.
15617 * early-remat.c: New file.
15618 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15620 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15623 Alan Hayward <alan.hayward@arm.com>
15624 David Sherwood <david.sherwood@arm.com>
15626 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15627 vfm1 with a bound_epilog parameter.
15628 (vect_do_peeling): Update calls accordingly, and move the prologue
15629 call earlier in the function. Treat the base bound_epilog as 0 for
15630 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15631 this base when peeling for gaps.
15632 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15633 with fully-masked loops.
15634 (vect_estimate_min_profitable_iters): Handle the single peeled
15635 iteration in that case.
15637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15638 Alan Hayward <alan.hayward@arm.com>
15639 David Sherwood <david.sherwood@arm.com>
15641 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15642 single-element interleaving even if the size is not a power of 2.
15643 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15644 accesses for single-element interleaving if the group size is
15647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15648 Alan Hayward <alan.hayward@arm.com>
15649 David Sherwood <david.sherwood@arm.com>
15651 * doc/md.texi (fold_extract_last_@var{m}): Document.
15652 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15653 * optabs.def (fold_extract_last_optab): New optab.
15654 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15655 * internal-fn.c (fold_extract_direct): New macro.
15656 (expand_fold_extract_optab_fn): Likewise.
15657 (direct_fold_extract_optab_supported_p): Likewise.
15658 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15659 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15660 EXTRACT_LAST_REDUCTION.
15661 (get_initial_def_for_reduction): Do not create an initial vector
15662 for EXTRACT_LAST_REDUCTION reductions.
15663 (vectorizable_reduction): Leave the scalar phi in place for
15664 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15665 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15666 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15667 transform phase to vectorizable_condition.
15668 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15670 (vect_finish_stmt_generation): ...here.
15671 (vect_finish_replace_stmt): New function.
15672 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15673 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15675 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15678 Alan Hayward <alan.hayward@arm.com>
15679 David Sherwood <david.sherwood@arm.com>
15681 * doc/md.texi (extract_last_@var{m}): Document.
15682 * optabs.def (extract_last_optab): New optab.
15683 * internal-fn.def (EXTRACT_LAST): New internal function.
15684 * internal-fn.c (cond_unary_direct): New macro.
15685 (expand_cond_unary_optab_fn): Likewise.
15686 (direct_cond_unary_optab_supported_p): Likewise.
15687 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15688 loops using EXTRACT_LAST.
15689 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15690 (extract_last_<mode>): ...this optab.
15691 (vec_extract<mode><Vel>): Update accordingly.
15693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15694 Alan Hayward <alan.hayward@arm.com>
15695 David Sherwood <david.sherwood@arm.com>
15697 * target.def (empty_mask_is_expensive): New hook.
15698 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15699 * doc/tm.texi: Regenerate.
15700 * targhooks.h (default_empty_mask_is_expensive): Declare.
15701 * targhooks.c (default_empty_mask_is_expensive): New function.
15702 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15703 if the target says that empty masks are expensive.
15704 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15706 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15709 Alan Hayward <alan.hayward@arm.com>
15710 David Sherwood <david.sherwood@arm.com>
15712 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15713 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15714 (vect_use_loop_mask_for_alignment_p): New function.
15715 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15716 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15717 niters_skip argument. Make sure that the first niters_skip elements
15718 of the first iteration are inactive.
15719 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15720 Update call to vect_set_loop_masks_directly.
15721 (get_misalign_in_elems): New function, split out from...
15722 (vect_gen_prolog_loop_niters): ...here.
15723 (vect_update_init_of_dr): Take a code argument that specifies whether
15724 the adjustment should be added or subtracted.
15725 (vect_update_init_of_drs): Likewise.
15726 (vect_prepare_for_masked_peels): New function.
15727 (vect_do_peeling): Skip prologue peeling if we're using a mask
15728 instead. Update call to vect_update_inits_of_drs.
15729 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15731 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15732 alignment. Do not include the number of peeled iterations in
15733 the minimum threshold in that case.
15734 (vectorizable_induction): Adjust the start value down by
15735 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15736 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15737 Take the number of skipped iterations into account when calculating
15739 * tree-vect-stmts.c (vect_gen_while_not): New function.
15741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15742 Alan Hayward <alan.hayward@arm.com>
15743 David Sherwood <david.sherwood@arm.com>
15745 * doc/sourcebuild.texi (vect_fully_masked): Document.
15746 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15747 default value to 0.
15748 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15750 (vect_analyze_loop_2): ...here. Don't check the vectorization
15751 factor against the number of loop iterations if the loop is
15754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15755 Alan Hayward <alan.hayward@arm.com>
15756 David Sherwood <david.sherwood@arm.com>
15758 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15759 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15760 (dump_groups): Update accordingly.
15761 (iv_use::mem_type): New member variable.
15762 (address_p): New function.
15763 (record_use): Add a mem_type argument and initialize the new
15765 (record_group_use): Add a mem_type argument. Use address_p.
15766 Remove obsolete null checks of base_object. Update call to record_use.
15767 (find_interesting_uses_op): Update call to record_group_use.
15768 (find_interesting_uses_cond): Likewise.
15769 (find_interesting_uses_address): Likewise.
15770 (get_mem_type_for_internal_fn): New function.
15771 (find_address_like_use): Likewise.
15772 (find_interesting_uses_stmt): Try find_address_like_use before
15773 calling find_interesting_uses_op.
15774 (addr_offset_valid_p): Use the iv mem_type field as the type
15775 of the addressed memory.
15776 (add_autoinc_candidates): Likewise.
15777 (get_address_cost): Likewise.
15778 (split_small_address_groups_p): Use address_p.
15779 (split_address_groups): Likewise.
15780 (add_iv_candidate_for_use): Likewise.
15781 (autoinc_possible_for_pair): Likewise.
15782 (rewrite_groups): Likewise.
15783 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15784 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15785 (get_alias_ptr_type_for_ptr_address): New function.
15786 (rewrite_use_address): Rewrite address uses in calls that were
15787 identified by find_address_like_use.
15789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15790 Alan Hayward <alan.hayward@arm.com>
15791 David Sherwood <david.sherwood@arm.com>
15793 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15795 * gimple-expr.h (is_gimple_addressable: Likewise.
15796 * gimple-expr.c (is_gimple_address): Likewise.
15797 * internal-fn.c (expand_call_mem_ref): New function.
15798 (expand_mask_load_optab_fn): Use it.
15799 (expand_mask_store_optab_fn): Likewise.
15801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15802 Alan Hayward <alan.hayward@arm.com>
15803 David Sherwood <david.sherwood@arm.com>
15805 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15806 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15807 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15808 (cond_umax@var{mode}): Document.
15809 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15810 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15811 (cond_umin_optab, cond_umax_optab): New optabs.
15812 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15813 (COND_IOR, COND_XOR): New internal functions.
15814 * internal-fn.h (get_conditional_internal_fn): Declare.
15815 * internal-fn.c (cond_binary_direct): New macro.
15816 (expand_cond_binary_optab_fn): Likewise.
15817 (direct_cond_binary_optab_supported_p): Likewise.
15818 (get_conditional_internal_fn): New function.
15819 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15820 Cope with reduction statements that are vectorized as calls rather
15822 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15823 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15824 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15825 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15826 (UNSPEC_COND_EOR): New unspecs.
15827 (optab): Add mappings for them.
15828 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15829 (sve_int_op, sve_fp_op): New int attributes.
15831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15832 Alan Hayward <alan.hayward@arm.com>
15833 David Sherwood <david.sherwood@arm.com>
15835 * optabs.def (while_ult_optab): New optab.
15836 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15837 * internal-fn.def (WHILE_ULT): New internal function.
15838 * internal-fn.h (direct_internal_fn_supported_p): New override
15839 that takes two types as argument.
15840 * internal-fn.c (while_direct): New macro.
15841 (expand_while_optab_fn): New function.
15842 (convert_optab_supported_p): Likewise.
15843 (direct_while_optab_supported_p): New macro.
15844 * wide-int.h (wi::udiv_ceil): New function.
15845 * tree-vectorizer.h (rgroup_masks): New structure.
15846 (vec_loop_masks): New typedef.
15847 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15848 and fully_masked_p.
15849 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15850 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15851 (vect_max_vf): New function.
15852 (slpeel_make_loop_iterate_ntimes): Delete.
15853 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15854 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15855 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15856 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15857 internal-fn.h, stor-layout.h and optabs-query.h.
15858 (vect_set_loop_mask): New function.
15859 (add_preheader_seq): Likewise.
15860 (add_header_seq): Likewise.
15861 (interleave_supported_p): Likewise.
15862 (vect_maybe_permute_loop_masks): Likewise.
15863 (vect_set_loop_masks_directly): Likewise.
15864 (vect_set_loop_condition_masked): Likewise.
15865 (vect_set_loop_condition_unmasked): New function, split out from
15866 slpeel_make_loop_iterate_ntimes.
15867 (slpeel_make_loop_iterate_ntimes): Rename to..
15868 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15869 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15870 (vect_do_peeling): Update call accordingly.
15871 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15873 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15874 mask_compare_type, can_fully_mask_p and fully_masked_p.
15875 (release_vec_loop_masks): New function.
15876 (_loop_vec_info): Use it to free the loop masks.
15877 (can_produce_all_loop_masks_p): New function.
15878 (vect_get_max_nscalars_per_iter): Likewise.
15879 (vect_verify_full_masking): Likewise.
15880 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15881 retries, and free the mask rgroups before retrying. Check loop-wide
15882 reasons for disallowing fully-masked loops. Make the final decision
15883 about whether use a fully-masked loop or not.
15884 (vect_estimate_min_profitable_iters): Do not assume that peeling
15885 for the number of iterations will be needed for fully-masked loops.
15886 (vectorizable_reduction): Disable fully-masked loops.
15887 (vectorizable_live_operation): Likewise.
15888 (vect_halve_mask_nunits): New function.
15889 (vect_double_mask_nunits): Likewise.
15890 (vect_record_loop_mask): Likewise.
15891 (vect_get_loop_mask): Likewise.
15892 (vect_transform_loop): Handle the case in which the final loop
15893 iteration might handle a partial vector. Call vect_set_loop_condition
15894 instead of slpeel_make_loop_iterate_ntimes.
15895 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15896 (check_load_store_masking): New function.
15897 (prepare_load_store_mask): Likewise.
15898 (vectorizable_store): Handle fully-masked loops.
15899 (vectorizable_load): Likewise.
15900 (supportable_widening_operation): Use vect_halve_mask_nunits for
15902 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15903 (vect_gen_while): New function.
15904 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15905 (aarch64_uqdec<mode>): New insn.
15907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15908 Alan Hayward <alan.hayward@arm.com>
15909 David Sherwood <david.sherwood@arm.com>
15911 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15912 (reduc_xor_scal_optab): New optabs.
15913 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15914 (reduc_xor_scal_@var{m}): Document.
15915 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15916 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15917 internal functions.
15918 * fold-const-call.c (fold_const_call): Handle them.
15919 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15920 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15921 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15922 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15923 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15924 (UNSPEC_XORV): New unspecs.
15925 (optab): Add entries for them.
15926 (BITWISEV): New int iterator.
15927 (bit_reduc_op): New int attributes.
15929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15930 Alan Hayward <alan.hayward@arm.com>
15931 David Sherwood <david.sherwood@arm.com>
15933 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15934 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15935 * optabs.def (vec_shl_insert_optab): New optab.
15936 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15937 (duplicate_and_interleave): Likewise.
15938 * tree-vect-loop.c: Include internal-fn.h.
15939 (neutral_op_for_slp_reduction): New function, split out from
15940 get_initial_defs_for_reduction.
15941 (get_initial_def_for_reduction): Handle option 2 for variable-length
15942 vectors by loading the neutral value into a vector and then shifting
15943 the initial value into element 0.
15944 (get_initial_defs_for_reduction): Replace the code argument with
15945 the neutral value calculated by neutral_op_for_slp_reduction.
15946 Use gimple_build_vector for constant-length vectors.
15947 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15948 but the first group_size elements have a neutral value.
15949 Use duplicate_and_interleave otherwise.
15950 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15951 Update call to get_initial_defs_for_reduction. Handle SLP
15952 reductions for variable-length vectors by creating one vector
15953 result for each scalar result, with the elements associated
15954 with other scalar results stubbed out with the neutral value.
15955 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15956 Require IFN_VEC_SHL_INSERT for double reductions on
15957 variable-length vectors, or SLP reductions that have
15958 a neutral value. Require can_duplicate_and_interleave_p
15959 support for variable-length unchained SLP reductions if there
15960 is no neutral value, such as for MIN/MAX reductions. Also require
15961 the number of vector elements to be a multiple of the number of
15962 SLP statements when doing variable-length unchained SLP reductions.
15963 Update call to vect_create_epilog_for_reduction.
15964 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15965 and remove initial values.
15966 (duplicate_and_interleave): Make public.
15967 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15968 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15971 Alan Hayward <alan.hayward@arm.com>
15972 David Sherwood <david.sherwood@arm.com>
15974 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15975 (can_duplicate_and_interleave_p): New function.
15976 (vect_get_and_check_slp_defs): Take the vector of statements
15977 rather than just the current one. Remove excess parentheses.
15978 Restriction rejectinon of vect_constant_def and vect_external_def
15979 for variable-length vectors to boolean types, or types for which
15980 can_duplicate_and_interleave_p is false.
15981 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15982 (duplicate_and_interleave): New function.
15983 (vect_get_constant_vectors): Use gimple_build_vector for
15984 constant-length vectors and suitable variable-length constant
15985 vectors. Use duplicate_and_interleave for other variable-length
15986 vectors. Don't defer the update when inserting new statements.
15988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15989 Alan Hayward <alan.hayward@arm.com>
15990 David Sherwood <david.sherwood@arm.com>
15992 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15993 min_profitable_iters doesn't go negative.
15995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15996 Alan Hayward <alan.hayward@arm.com>
15997 David Sherwood <david.sherwood@arm.com>
15999 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16000 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16001 * optabs.def (vec_mask_load_lanes_optab): New optab.
16002 (vec_mask_store_lanes_optab): Likewise.
16003 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16004 (MASK_STORE_LANES): Likewise.
16005 * internal-fn.c (mask_load_lanes_direct): New macro.
16006 (mask_store_lanes_direct): Likewise.
16007 (expand_mask_load_optab_fn): Handle masked operations.
16008 (expand_mask_load_lanes_optab_fn): New macro.
16009 (expand_mask_store_optab_fn): Handle masked operations.
16010 (expand_mask_store_lanes_optab_fn): New macro.
16011 (direct_mask_load_lanes_optab_supported_p): Likewise.
16012 (direct_mask_store_lanes_optab_supported_p): Likewise.
16013 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16015 (vect_load_lanes_supported): Likewise.
16016 * tree-vect-data-refs.c (strip_conversion): New function.
16017 (can_group_stmts_p): Likewise.
16018 (vect_analyze_data_ref_accesses): Use it instead of checking
16019 for a pair of assignments.
16020 (vect_store_lanes_supported): Take a masked_p parameter.
16021 (vect_load_lanes_supported): Likewise.
16022 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16023 vect_store_lanes_supported and vect_load_lanes_supported.
16024 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16025 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16026 parameter. Don't allow gaps for masked accesses.
16027 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16028 and vect_load_lanes_supported.
16029 (get_load_store_type): Take a masked_p parameter and update
16030 call to get_group_load_store_type.
16031 (vectorizable_store): Update call to get_load_store_type.
16032 Handle IFN_MASK_STORE_LANES.
16033 (vectorizable_load): Update call to get_load_store_type.
16034 Handle IFN_MASK_LOAD_LANES.
16036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16037 Alan Hayward <alan.hayward@arm.com>
16038 David Sherwood <david.sherwood@arm.com>
16040 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16042 * config/aarch64/aarch64-protos.h
16043 (aarch64_sve_struct_memory_operand_p): Declare.
16044 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16045 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16046 (VPRED, vpred): Handle SVE structure modes.
16047 * config/aarch64/constraints.md (Utx): New constraint.
16048 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16049 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16050 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16051 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16052 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16053 structure modes. Split into pieces after RA.
16054 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16055 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16057 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16058 SVE structure modes.
16059 (aarch64_classify_address): Likewise.
16060 (sizetochar): Move earlier in file.
16061 (aarch64_print_operand): Handle SVE register lists.
16062 (aarch64_array_mode): New function.
16063 (aarch64_sve_struct_memory_operand_p): Likewise.
16064 (TARGET_ARRAY_MODE): Redefine.
16066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16067 Alan Hayward <alan.hayward@arm.com>
16068 David Sherwood <david.sherwood@arm.com>
16070 * target.def (array_mode): New target hook.
16071 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16072 * doc/tm.texi: Regenerate.
16073 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16074 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16075 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16076 targetm.array_mode.
16077 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16081 Alan Hayward <alan.hayward@arm.com>
16082 David Sherwood <david.sherwood@arm.com>
16084 * fold-const.c (fold_binary_loc): Check the argument types
16085 rather than the result type when testing for a vector operation.
16087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16089 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16090 * doc/tm.texi: Regenerate.
16092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16093 Alan Hayward <alan.hayward@arm.com>
16094 David Sherwood <david.sherwood@arm.com>
16096 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16097 (sve): Document new AArch64 extension.
16098 * doc/md.texi (w): Extend the description of the AArch64
16099 constraint to include SVE vectors.
16100 (Upl, Upa): Document new AArch64 predicate constraints.
16101 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16103 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16104 (msve-vector-bits=): New option.
16105 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16106 SVE when these are disabled.
16107 (sve): New extension.
16108 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16109 modes. Adjust their number of units based on aarch64_sve_vg.
16110 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16111 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16112 aarch64_addr_query_type.
16113 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16114 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16115 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16116 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16117 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16118 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16119 (aarch64_simd_imm_zero_p): Delete.
16120 (aarch64_check_zero_based_sve_index_immediate): Declare.
16121 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16122 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16123 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16124 (aarch64_sve_float_mul_immediate_p): Likewise.
16125 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16126 rather than an rtx.
16127 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16128 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16129 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16130 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16131 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16132 (aarch64_regmode_natural_size): Likewise.
16133 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16134 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16136 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16137 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16138 for VG and the SVE predicate registers.
16139 (V_ALIASES): Add a "z"-prefixed alias.
16140 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16141 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16142 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16143 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16144 (REG_CLASS_NAMES): Add entries for them.
16145 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16146 and the predicate registers.
16147 (aarch64_sve_vg): Declare.
16148 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16149 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16150 (REGMODE_NATURAL_SIZE): Define.
16151 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16153 * config/aarch64/aarch64.c: Include cfgrtl.h.
16154 (simd_immediate_info): Add a constructor for series vectors,
16155 and an associated step field.
16156 (aarch64_sve_vg): New variable.
16157 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16158 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16159 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16160 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16161 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16162 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16163 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16164 (aarch64_get_mask_mode): New functions.
16165 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16166 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16167 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16168 predicate modes and predicate registers. Explicitly restrict
16169 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16170 to store a vector mode if it is recognized by
16171 aarch64_classify_vector_mode.
16172 (aarch64_regmode_natural_size): New function.
16173 (aarch64_hard_regno_caller_save_mode): Return the original mode
16175 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16176 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16177 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16178 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16180 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16181 does not overlap dest if the function is frame-related. Handle
16183 (aarch64_split_add_offset): New function.
16184 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16185 them aarch64_add_offset.
16186 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16187 and update call to aarch64_sub_sp.
16188 (aarch64_add_cfa_expression): New function.
16189 (aarch64_expand_prologue): Pass extra temporary registers to the
16190 functions above. Handle the case in which we need to emit new
16191 DW_CFA_expressions for registers that were originally saved
16192 relative to the stack pointer, but now have to be expressed
16193 relative to the frame pointer.
16194 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16196 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16197 IP0 and IP1 values for SVE frames.
16198 (aarch64_expand_vec_series): New function.
16199 (aarch64_expand_sve_widened_duplicate): Likewise.
16200 (aarch64_expand_sve_const_vector): Likewise.
16201 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16202 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16203 into the register, rather than emitting a SET directly.
16204 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16205 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16206 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16207 (offset_9bit_signed_scaled_p): New functions.
16208 (aarch64_replicate_bitmask_imm): New function.
16209 (aarch64_bitmask_imm): Use it.
16210 (aarch64_cannot_force_const_mem): Reject expressions involving
16211 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16212 (aarch64_classify_index): Handle SVE indices, by requiring
16213 a plain register index with a scale that matches the element size.
16214 (aarch64_classify_address): Handle SVE addresses. Assert that
16215 the mode of the address is VOIDmode or an integer mode.
16216 Update call to aarch64_classify_symbol.
16217 (aarch64_classify_symbolic_expression): Update call to
16218 aarch64_classify_symbol.
16219 (aarch64_const_vec_all_in_range_p): New function.
16220 (aarch64_print_vector_float_operand): Likewise.
16221 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16222 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16223 and the FP immediates 1.0 and 0.5.
16224 (aarch64_print_address_internal): Handle SVE addresses.
16225 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16226 (aarch64_regno_regclass): Handle predicate registers.
16227 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16229 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16230 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16231 (aarch64_convert_sve_vector_bits): New function.
16232 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16233 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16234 rather than an rtx.
16235 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16236 Handle SVE vector and predicate modes. Accept VL-based constants
16237 that need only one temporary register, and VL offsets that require
16238 no temporary registers.
16239 (aarch64_conditional_register_usage): Mark the predicate registers
16240 as fixed if SVE isn't available.
16241 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16242 Return true for SVE vector and predicate modes.
16243 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16244 rather than an unsigned int. Handle SVE modes.
16245 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16247 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16249 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16250 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16251 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16252 (aarch64_sve_float_mul_immediate_p): New functions.
16253 (aarch64_sve_valid_immediate): New function.
16254 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16255 Explicitly reject structure modes. Check for INDEX constants.
16256 Handle PTRUE and PFALSE constants.
16257 (aarch64_check_zero_based_sve_index_immediate): New function.
16258 (aarch64_simd_imm_zero_p): Delete.
16259 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16260 vector modes. Accept constants in the range of CNT[BHWD].
16261 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16262 ask for an Advanced SIMD mode.
16263 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16264 (aarch64_simd_vector_alignment): Handle SVE predicates.
16265 (aarch64_vectorize_preferred_vector_alignment): New function.
16266 (aarch64_simd_vector_alignment_reachable): Use it instead of
16268 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16269 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16271 (MAX_VECT_LEN): Delete.
16272 (expand_vec_perm_d): Add a vec_flags field.
16273 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16274 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16275 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16277 (aarch64_evpc_rev): Rename to...
16278 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16279 (aarch64_evpc_rev_global): New function.
16280 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16281 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16283 (aarch64_evpc_sve_tbl): New function.
16284 (aarch64_expand_vec_perm_const_1): Update after rename of
16285 aarch64_evpc_rev. Handle SVE permutes too, trying
16286 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16287 than aarch64_evpc_tbl.
16288 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16289 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16290 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16291 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16292 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16293 (aarch64_expand_sve_vcond): New functions.
16294 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16295 of aarch64_vector_mode_p.
16296 (aarch64_dwarf_poly_indeterminate_value): New function.
16297 (aarch64_compute_pressure_classes): Likewise.
16298 (aarch64_can_change_mode_class): Likewise.
16299 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16300 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16301 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16302 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16303 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16304 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16305 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16306 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16308 (Dn, Dl, Dr): Accept const as well as const_vector.
16309 (Dz): Likewise. Compare against CONST0_RTX.
16310 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16311 of "vector" where appropriate.
16312 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16313 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16314 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16315 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16316 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16317 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16318 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16319 (v_int_equiv): Extend to SVE modes.
16320 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16322 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16323 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16324 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16325 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16326 (SVE_COND_FP_CMP): New int iterators.
16327 (perm_hilo): Handle the new unpack unspecs.
16328 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16330 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16331 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16332 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16333 (aarch64_equality_operator, aarch64_constant_vector_operand)
16334 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16335 (aarch64_sve_nonimmediate_operand): Likewise.
16336 (aarch64_sve_general_operand): Likewise.
16337 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16338 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16339 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16340 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16341 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16342 (aarch64_sve_float_arith_immediate): Likewise.
16343 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16344 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16345 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16346 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16347 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16348 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16349 (aarch64_sve_float_arith_operand): Likewise.
16350 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16351 (aarch64_sve_float_mul_operand): Likewise.
16352 (aarch64_sve_vec_perm_operand): Likewise.
16353 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16354 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16355 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16356 as well as const_vector.
16357 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16358 in file. Use CONST0_RTX and CONSTM1_RTX.
16359 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16360 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16361 Use aarch64_simd_imm_zero.
16362 * config/aarch64/aarch64-sve.md: New file.
16363 * config/aarch64/aarch64.md: Include it.
16364 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16365 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16366 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16367 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16368 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16369 (sve): New attribute.
16370 (enabled): Disable instructions with the sve attribute unless
16372 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16373 aarch64_expand_mov_immediate.
16374 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16375 CNT[BHSD] immediates.
16376 (movti): Split CONST_POLY_INT moves into two halves.
16377 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16378 Split additions that need a temporary here if the destination
16379 is the stack pointer.
16380 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16381 (*add<mode>3_poly_1): New instruction.
16382 (set_clobber_cc): New expander.
16384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16386 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16387 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16388 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16389 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16390 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16391 Change innermode from fixed_mode_size to machine_mode.
16392 (simplify_subreg): Update call accordingly. Handle a constant-sized
16393 subreg of a variable-length CONST_VECTOR.
16395 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16396 Alan Hayward <alan.hayward@arm.com>
16397 David Sherwood <david.sherwood@arm.com>
16399 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16400 (add_offset_to_base): New function, split out from...
16401 (create_mem_ref): ...here. When handling a scale other than 1,
16402 check first whether the address is valid without the offset.
16403 Add it into the base if so, leaving the index and scale as-is.
16405 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16408 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16409 fold_for_warn before checking if arg2 is INTEGER_CST.
16411 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16413 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16414 (store_multiple_operation): Delete.
16415 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16416 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16417 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16418 guarded by TARGET_STRING.
16419 (rs6000_output_load_multiple): Delete.
16420 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16421 OPTION_MASK_STRING / TARGET_STRING handling.
16422 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16423 (const rs6000_opt_masks) <"string">: Change mask to 0.
16424 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16425 (MASK_STRING): Delete.
16426 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16428 (load_multiple): Delete.
16435 (store_multiple): Delete.
16442 (movmemsi_8reg): Delete.
16443 (corresponding unnamed define_insn): Delete.
16444 (movmemsi_6reg): Delete.
16445 (corresponding unnamed define_insn): Delete.
16446 (movmemsi_4reg): Delete.
16447 (corresponding unnamed define_insn): Delete.
16448 (movmemsi_2reg): Delete.
16449 (corresponding unnamed define_insn): Delete.
16450 (movmemsi_1reg): Delete.
16451 (corresponding unnamed define_insn): Delete.
16452 * config/rs6000/rs6000.opt (mno-string): New.
16453 (mstring): Replace by deprecation warning stub.
16454 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16456 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16458 * regrename.c (regrename_do_replace): If replacing the same
16459 reg multiple times, try to reuse last created gen_raw_REG.
16462 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16463 main to workaround a bug in GDB.
16465 2018-01-12 Tom de Vries <tom@codesourcery.com>
16468 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16470 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16472 PR rtl-optimization/80481
16473 * ira-color.c (get_cap_member): New function.
16474 (allocnos_conflict_by_live_ranges_p): Use it.
16475 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16476 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16478 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16481 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16482 (*saddl_se_1): Ditto.
16483 (*ssubsi_1): Ditto.
16484 (*ssubl_se_1): Ditto.
16486 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16488 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16489 rather than wi::to_widest for DR_INITs.
16490 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16491 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16492 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16494 (vect_analyze_group_access_1): Note that here.
16496 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16498 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16499 polynomial type sizes.
16501 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16503 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16504 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16505 (gimple_add_tmp_var): Likewise.
16507 2018-01-12 Martin Liska <mliska@suse.cz>
16509 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16510 (gimple_alloc_sizes): Likewise.
16511 (dump_gimple_statistics): Use PRIu64 in printf format.
16512 * gimple.h: Change uint64_t to int.
16514 2018-01-12 Martin Liska <mliska@suse.cz>
16516 * tree-core.h: Use uint64_t instead of int.
16517 * tree.c (tree_node_counts): Likewise.
16518 (tree_node_sizes): Likewise.
16519 (dump_tree_statistics): Use PRIu64 in printf format.
16521 2018-01-12 Martin Liska <mliska@suse.cz>
16523 * Makefile.in: As qsort_chk is implemented in vec.c, add
16524 vec.o to linkage of gencfn-macros.
16525 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16526 passing the info to record_node_allocation_statistics.
16527 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16529 * ggc-common.c (struct ggc_usage): Add operator== and use
16530 it in operator< and compare function.
16531 * mem-stats.h (struct mem_usage): Likewise.
16532 * vec.c (struct vec_usage): Remove operator< and compare
16533 function. Can be simply inherited.
16535 2018-01-12 Martin Jambor <mjambor@suse.cz>
16538 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16539 * tree-ssa-math-opts.c: Include domwalk.h.
16540 (convert_mult_to_fma_1): New function.
16541 (fma_transformation_info): New type.
16542 (fma_deferring_state): Likewise.
16543 (cancel_fma_deferring): New function.
16544 (result_of_phi): Likewise.
16545 (last_fma_candidate_feeds_initial_phi): Likewise.
16546 (convert_mult_to_fma): Added deferring logic, split actual
16547 transformation to convert_mult_to_fma_1.
16548 (math_opts_dom_walker): New type.
16549 (math_opts_dom_walker::after_dom_children): New method, body moved
16550 here from pass_optimize_widening_mul::execute, added deferring logic
16552 (pass_optimize_widening_mul::execute): Moved most of code to
16553 math_opts_dom_walker::after_dom_children.
16554 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16555 * config/i386/i386.c (ix86_option_override_internal): Added
16556 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16558 2018-01-12 Richard Biener <rguenther@suse.de>
16561 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16562 inline instance vars.
16564 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16567 * config/rx/rx.c (rx_is_restricted_memory_address):
16568 Handle SUBREG case.
16570 2018-01-12 Richard Biener <rguenther@suse.de>
16572 PR tree-optimization/80846
16573 * target.def (split_reduction): New target hook.
16574 * targhooks.c (default_split_reduction): New function.
16575 * targhooks.h (default_split_reduction): Declare.
16576 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16577 target requests first reduce vectors by combining low and high
16579 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16580 (get_vectype_for_scalar_type_and_size): Export.
16581 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16582 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16583 * doc/tm.texi: Regenerate.
16584 * config/i386/i386.c (ix86_split_reduction): Implement
16585 TARGET_VECTORIZE_SPLIT_REDUCTION.
16587 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16590 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16591 in PIC mode except for TARGET_VXWORKS_RTP.
16592 * config/sparc/sparc.c: Include cfgrtl.h.
16593 (TARGET_INIT_PIC_REG): Define.
16594 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16595 (sparc_pic_register_p): New predicate.
16596 (sparc_legitimate_address_p): Use it.
16597 (sparc_legitimize_pic_address): Likewise.
16598 (sparc_delegitimize_address): Likewise.
16599 (sparc_mode_dependent_address_p): Likewise.
16600 (gen_load_pcrel_sym): Remove 4th parameter.
16601 (load_got_register): Adjust call to above. Remove obsolete stuff.
16602 (sparc_expand_prologue): Do not call load_got_register here.
16603 (sparc_flat_expand_prologue): Likewise.
16604 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16605 (sparc_use_pseudo_pic_reg): New function.
16606 (sparc_init_pic_reg): Likewise.
16607 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16608 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16610 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16612 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16613 Add item for branch_cost.
16615 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16617 PR rtl-optimization/83565
16618 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16619 not extend the result to a larger mode for rotate operations.
16620 (num_sign_bit_copies1): Likewise.
16622 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16625 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16627 Use values-Xc.o for -pedantic.
16628 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16630 2018-01-12 Martin Liska <mliska@suse.cz>
16633 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16635 (possible_polymorphic_call_targets): Use it.
16636 (ipa_devirt): Likewise.
16638 2018-01-12 Martin Liska <mliska@suse.cz>
16640 * profile-count.h (enum profile_quality): Use 0 as invalid
16641 enum value of profile_quality.
16643 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16645 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16646 -mext-string options.
16648 2018-01-12 Richard Biener <rguenther@suse.de>
16650 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16651 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16652 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16654 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16656 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16658 * configure.ac (--with-long-double-format): Add support for the
16659 configuration option to change the default long double format on
16661 * config.gcc (powerpc*-linux*-*): Likewise.
16662 * configure: Regenerate.
16663 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16664 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16665 used without modification.
16667 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16669 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16670 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16671 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16672 MISC_BUILTIN_SPEC_BARRIER.
16673 (rs6000_init_builtins): Likewise.
16674 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16676 (speculation_barrier): New define_insn.
16677 * doc/extend.texi: Document __builtin_speculation_barrier.
16679 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16682 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16683 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16684 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16686 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16687 integral modes instead of "ss" and "sd".
16688 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16689 vectors with 32-bit and 64-bit elements.
16690 (vecdupssescalarmodesuffix): New mode attribute.
16691 (vec_dup<mode>): Use it.
16693 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16696 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16697 frame if argument is passed on stack.
16699 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16702 * ree.c (combine_reaching_defs): Optimize also
16703 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16704 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16706 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16708 PR middle-end/83189
16709 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16711 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16713 PR middle-end/83718
16714 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16715 after they are computed.
16717 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16719 PR tree-optimization/83695
16720 * gimple-loop-linterchange.cc
16721 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16722 reset cached scev information after interchange.
16723 (pass_linterchange::execute): Remove call to scev_reset_htab.
16725 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16727 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16728 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16729 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16730 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16731 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16732 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16733 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16734 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16735 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16736 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16737 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16738 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16739 (V_lane_reg): Likewise.
16740 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16742 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16743 (vfmal_lane_low<mode>_intrinsic,
16744 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16745 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16746 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16747 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16748 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16749 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16751 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16753 * config/arm/arm-cpus.in (fp16fml): New feature.
16754 (ALL_SIMD): Add fp16fml.
16755 (armv8.2-a): Add fp16fml as an option.
16756 (armv8.3-a): Likewise.
16757 (armv8.4-a): Add fp16fml as part of fp16.
16758 * config/arm/arm.h (TARGET_FP16FML): Define.
16759 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16761 * config/arm/arm-modes.def (V2HF): Define.
16762 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16763 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16764 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16765 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16766 vfmsl_low, vfmsl_high): New set of builtins.
16767 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16768 (vfml_op): New code attribute.
16769 (VFMLHALVES): New int iterator.
16770 (VFML, VFMLSEL): New mode attributes.
16771 (V_reg): Define mapping for V2HF.
16772 (V_hi, V_lo): New mode attributes.
16773 (VF_constraint): Likewise.
16774 (vfml_half, vfml_half_selector): New int attributes.
16775 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16777 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16778 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16780 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16781 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16782 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16783 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16785 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16786 Document new effective target and option set.
16788 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16790 * config/arm/arm-cpus.in (armv8_4): New feature.
16791 (ARMv8_4a): New fgroup.
16792 (armv8.4-a): New arch.
16793 * config/arm/arm-tables.opt: Regenerate.
16794 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16795 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16796 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16797 Add matching rules for -march=armv8.4-a and extensions.
16798 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16800 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16803 * config/rx/rx.md (BW): New mode attribute.
16804 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16806 2018-01-11 Richard Biener <rguenther@suse.de>
16808 PR tree-optimization/83435
16809 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16810 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16811 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16813 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16814 Alan Hayward <alan.hayward@arm.com>
16815 David Sherwood <david.sherwood@arm.com>
16817 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16819 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16820 (aarch64_print_address_internal): Use it to check for a zero offset.
16822 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16823 Alan Hayward <alan.hayward@arm.com>
16824 David Sherwood <david.sherwood@arm.com>
16826 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16827 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16828 Return a poly_int64 rather than a HOST_WIDE_INT.
16829 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16830 rather than a HOST_WIDE_INT.
16831 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16832 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16833 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16834 final_offset from HOST_WIDE_INT to poly_int64.
16835 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16836 to_constant when getting the number of units in an Advanced SIMD
16838 (aarch64_builtin_vectorized_function): Check for a constant number
16840 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16842 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16843 attribute instead of GET_MODE_NUNITS.
16844 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16845 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16846 GET_MODE_SIZE for fixed-size registers.
16847 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16848 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16849 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16850 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16851 (aarch64_print_operand, aarch64_print_address_internal)
16852 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16853 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16854 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16855 Handle polynomial GET_MODE_SIZE.
16856 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16857 wider than SImode without modification.
16858 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16859 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16860 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16861 passing and returning SVE modes.
16862 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16863 rather than GEN_INT.
16864 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16865 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16866 (aarch64_allocate_and_probe_stack_space): Likewise.
16867 (aarch64_layout_frame): Cope with polynomial offsets.
16868 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16869 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16870 polynomial offsets.
16871 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16872 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16873 poly_int64 rather than a HOST_WIDE_INT.
16874 (aarch64_get_separate_components, aarch64_process_components)
16875 (aarch64_expand_prologue, aarch64_expand_epilogue)
16876 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16877 (aarch64_anchor_offset): New function, split out from...
16878 (aarch64_legitimize_address): ...here.
16879 (aarch64_builtin_vectorization_cost): Handle polynomial
16880 TYPE_VECTOR_SUBPARTS.
16881 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16883 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16884 number of elements from the PARALLEL rather than the mode.
16885 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16886 rather than GET_MODE_BITSIZE.
16887 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16888 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16889 (aarch64_expand_vec_perm_const_1): Handle polynomial
16890 d->perm.length () and d->perm elements.
16891 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16892 Apply to_constant to d->perm elements.
16893 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16894 polynomial CONST_VECTOR_NUNITS.
16895 (aarch64_move_pointer): Take amount as a poly_int64 rather
16897 (aarch64_progress_pointer): Avoid temporary variable.
16898 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16899 the mode attribute instead of GET_MODE.
16901 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16902 Alan Hayward <alan.hayward@arm.com>
16903 David Sherwood <david.sherwood@arm.com>
16905 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16906 x exists before using it.
16907 (aarch64_add_constant_internal): Rename to...
16908 (aarch64_add_offset_1): ...this. Replace regnum with separate
16909 src and dest rtxes. Handle the case in which they're different,
16910 including when the offset is zero. Replace scratchreg with an rtx.
16911 Use 2 additions if there is no spare register into which we can
16912 move a 16-bit constant.
16913 (aarch64_add_constant): Delete.
16914 (aarch64_add_offset): Replace reg with separate src and dest
16915 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16916 Use aarch64_add_offset_1.
16917 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16918 an rtx rather than an int. Take the delta as a poly_int64
16919 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16920 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16921 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16922 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16923 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16924 and aarch64_add_sp.
16925 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16926 aarch64_add_constant.
16928 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16930 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16931 Use scalar_float_mode.
16933 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16935 * config/aarch64/aarch64-simd.md
16936 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16937 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16938 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16939 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16940 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16941 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16942 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16943 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16944 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16945 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16947 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16950 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16951 targ_options->x_arm_arch_string is non NULL.
16953 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16955 * config/aarch64/aarch64.h
16956 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16958 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16961 * expmed.c (emit_store_flag_force): Swap if const op0
16962 and change VOIDmode to mode of op0.
16964 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16966 PR rtl-optimization/83761
16967 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16968 than bytes to mode_for_size.
16970 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16972 PR middle-end/83189
16973 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16974 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16977 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16979 PR middle-end/83575
16980 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16981 when in layout mode.
16982 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16983 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16986 2018-01-10 Michael Collison <michael.collison@arm.com>
16988 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16989 * config/aarch64/aarch64-option-extension.def: Add
16990 AARCH64_OPT_EXTENSION of 'fp16fml'.
16991 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16992 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16993 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16994 * config/aarch64/constraints.md (Ui7): New constraint.
16995 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16996 (VFMLA_SEL_W): Ditto.
16999 (VFMLA16_LOW): New int iterator.
17000 (VFMLA16_HIGH): Ditto.
17001 (UNSPEC_FMLAL): New unspec.
17002 (UNSPEC_FMLSL): Ditto.
17003 (UNSPEC_FMLAL2): Ditto.
17004 (UNSPEC_FMLSL2): Ditto.
17005 (f16mac): New code attribute.
17006 * config/aarch64/aarch64-simd-builtins.def
17007 (aarch64_fmlal_lowv2sf): Ditto.
17008 (aarch64_fmlsl_lowv2sf): Ditto.
17009 (aarch64_fmlalq_lowv4sf): Ditto.
17010 (aarch64_fmlslq_lowv4sf): Ditto.
17011 (aarch64_fmlal_highv2sf): Ditto.
17012 (aarch64_fmlsl_highv2sf): Ditto.
17013 (aarch64_fmlalq_highv4sf): Ditto.
17014 (aarch64_fmlslq_highv4sf): Ditto.
17015 (aarch64_fmlal_lane_lowv2sf): Ditto.
17016 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17017 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17018 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17019 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17020 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17021 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17022 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17023 (aarch64_fmlal_lane_highv2sf): Ditto.
17024 (aarch64_fmlsl_lane_highv2sf): Ditto.
17025 (aarch64_fmlal_laneq_highv2sf): Ditto.
17026 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17027 (aarch64_fmlalq_lane_highv4sf): Ditto.
17028 (aarch64_fmlsl_lane_highv4sf): Ditto.
17029 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17030 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17031 * config/aarch64/aarch64-simd.md:
17032 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17033 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17034 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17035 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17036 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17037 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17038 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17039 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17040 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17041 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17042 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17043 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17044 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17045 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17046 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17047 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17048 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17049 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17050 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17051 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17052 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17053 (vfmlsl_low_u32): Ditto.
17054 (vfmlalq_low_u32): Ditto.
17055 (vfmlslq_low_u32): Ditto.
17056 (vfmlal_high_u32): Ditto.
17057 (vfmlsl_high_u32): Ditto.
17058 (vfmlalq_high_u32): Ditto.
17059 (vfmlslq_high_u32): Ditto.
17060 (vfmlal_lane_low_u32): Ditto.
17061 (vfmlsl_lane_low_u32): Ditto.
17062 (vfmlal_laneq_low_u32): Ditto.
17063 (vfmlsl_laneq_low_u32): Ditto.
17064 (vfmlalq_lane_low_u32): Ditto.
17065 (vfmlslq_lane_low_u32): Ditto.
17066 (vfmlalq_laneq_low_u32): Ditto.
17067 (vfmlslq_laneq_low_u32): Ditto.
17068 (vfmlal_lane_high_u32): Ditto.
17069 (vfmlsl_lane_high_u32): Ditto.
17070 (vfmlal_laneq_high_u32): Ditto.
17071 (vfmlsl_laneq_high_u32): Ditto.
17072 (vfmlalq_lane_high_u32): Ditto.
17073 (vfmlslq_lane_high_u32): Ditto.
17074 (vfmlalq_laneq_high_u32): Ditto.
17075 (vfmlslq_laneq_high_u32): Ditto.
17076 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17077 (AARCH64_FL_FOR_ARCH8_4): New.
17078 (AARCH64_ISA_F16FML): New ISA flag.
17079 (TARGET_F16FML): New feature flag for fp16fml.
17080 (doc/invoke.texi): Document new fp16fml option.
17082 2018-01-10 Michael Collison <michael.collison@arm.com>
17084 * config/aarch64/aarch64-builtins.c:
17085 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17086 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17087 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17088 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17089 (AARCH64_ISA_SHA3): New ISA flag.
17090 (TARGET_SHA3): New feature flag for sha3.
17091 * config/aarch64/iterators.md (sha512_op): New int attribute.
17092 (CRYPTO_SHA512): New int iterator.
17093 (UNSPEC_SHA512H): New unspec.
17094 (UNSPEC_SHA512H2): Ditto.
17095 (UNSPEC_SHA512SU0): Ditto.
17096 (UNSPEC_SHA512SU1): Ditto.
17097 * config/aarch64/aarch64-simd-builtins.def
17098 (aarch64_crypto_sha512hqv2di): New builtin.
17099 (aarch64_crypto_sha512h2qv2di): Ditto.
17100 (aarch64_crypto_sha512su0qv2di): Ditto.
17101 (aarch64_crypto_sha512su1qv2di): Ditto.
17102 (aarch64_eor3qv8hi): Ditto.
17103 (aarch64_rax1qv2di): Ditto.
17104 (aarch64_xarqv2di): Ditto.
17105 (aarch64_bcaxqv8hi): Ditto.
17106 * config/aarch64/aarch64-simd.md:
17107 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17108 (aarch64_crypto_sha512su0qv2di): Ditto.
17109 (aarch64_crypto_sha512su1qv2di): Ditto.
17110 (aarch64_eor3qv8hi): Ditto.
17111 (aarch64_rax1qv2di): Ditto.
17112 (aarch64_xarqv2di): Ditto.
17113 (aarch64_bcaxqv8hi): Ditto.
17114 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17115 (vsha512h2q_u64): Ditto.
17116 (vsha512su0q_u64): Ditto.
17117 (vsha512su1q_u64): Ditto.
17118 (veor3q_u16): Ditto.
17119 (vrax1q_u64): Ditto.
17120 (vxarq_u64): Ditto.
17121 (vbcaxq_u16): Ditto.
17122 * config/arm/types.md (crypto_sha512): New type attribute.
17123 (crypto_sha3): Ditto.
17124 (doc/invoke.texi): Document new sha3 option.
17126 2018-01-10 Michael Collison <michael.collison@arm.com>
17128 * config/aarch64/aarch64-builtins.c:
17129 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17130 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17131 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17132 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17133 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17134 (AARCH64_ISA_SM4): New ISA flag.
17135 (TARGET_SM4): New feature flag for sm4.
17136 * config/aarch64/aarch64-simd-builtins.def
17137 (aarch64_sm3ss1qv4si): Ditto.
17138 (aarch64_sm3tt1aq4si): Ditto.
17139 (aarch64_sm3tt1bq4si): Ditto.
17140 (aarch64_sm3tt2aq4si): Ditto.
17141 (aarch64_sm3tt2bq4si): Ditto.
17142 (aarch64_sm3partw1qv4si): Ditto.
17143 (aarch64_sm3partw2qv4si): Ditto.
17144 (aarch64_sm4eqv4si): Ditto.
17145 (aarch64_sm4ekeyqv4si): Ditto.
17146 * config/aarch64/aarch64-simd.md:
17147 (aarch64_sm3ss1qv4si): Ditto.
17148 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17149 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17150 (aarch64_sm4eqv4si): Ditto.
17151 (aarch64_sm4ekeyqv4si): Ditto.
17152 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17153 (sm3part_op): Ditto.
17154 (CRYPTO_SM3TT): Ditto.
17155 (CRYPTO_SM3PART): Ditto.
17156 (UNSPEC_SM3SS1): New unspec.
17157 (UNSPEC_SM3TT1A): Ditto.
17158 (UNSPEC_SM3TT1B): Ditto.
17159 (UNSPEC_SM3TT2A): Ditto.
17160 (UNSPEC_SM3TT2B): Ditto.
17161 (UNSPEC_SM3PARTW1): Ditto.
17162 (UNSPEC_SM3PARTW2): Ditto.
17163 (UNSPEC_SM4E): Ditto.
17164 (UNSPEC_SM4EKEY): Ditto.
17165 * config/aarch64/constraints.md (Ui2): New constraint.
17166 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17167 * config/arm/types.md (crypto_sm3): New type attribute.
17168 (crypto_sm4): Ditto.
17169 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17170 (vsm3tt1aq_u32): Ditto.
17171 (vsm3tt1bq_u32): Ditto.
17172 (vsm3tt2aq_u32): Ditto.
17173 (vsm3tt2bq_u32): Ditto.
17174 (vsm3partw1q_u32): Ditto.
17175 (vsm3partw2q_u32): Ditto.
17176 (vsm4eq_u32): Ditto.
17177 (vsm4ekeyq_u32): Ditto.
17178 (doc/invoke.texi): Document new sm4 option.
17180 2018-01-10 Michael Collison <michael.collison@arm.com>
17182 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17183 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17184 (AARCH64_FL_FOR_ARCH8_4): New.
17185 (AARCH64_FL_V8_4): New flag.
17186 (doc/invoke.texi): Document new armv8.4-a option.
17188 2018-01-10 Michael Collison <michael.collison@arm.com>
17190 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17191 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17192 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17193 * config/aarch64/aarch64-option-extension.def: Add
17194 AARCH64_OPT_EXTENSION of 'sha2'.
17195 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17196 (crypto): Disable sha2 and aes if crypto disabled.
17197 (crypto): Enable aes and sha2 if enabled.
17198 (simd): Disable sha2 and aes if simd disabled.
17199 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17201 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17202 (TARGET_SHA2): New feature flag for sha2.
17203 (TARGET_AES): New feature flag for aes.
17204 * config/aarch64/aarch64-simd.md:
17205 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17206 conditional on TARGET_AES.
17207 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17208 (aarch64_crypto_sha1hsi): Make pattern conditional
17210 (aarch64_crypto_sha1hv4si): Ditto.
17211 (aarch64_be_crypto_sha1hv4si): Ditto.
17212 (aarch64_crypto_sha1su1v4si): Ditto.
17213 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17214 (aarch64_crypto_sha1su0v4si): Ditto.
17215 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17216 (aarch64_crypto_sha256su0v4si): Ditto.
17217 (aarch64_crypto_sha256su1v4si): Ditto.
17218 (doc/invoke.texi): Document new aes and sha2 options.
17220 2018-01-10 Martin Sebor <msebor@redhat.com>
17222 PR tree-optimization/83781
17223 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17226 2018-01-11 Martin Sebor <msebor@gmail.com>
17227 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17229 PR tree-optimization/83501
17230 PR tree-optimization/81703
17232 * tree-ssa-strlen.c (get_string_cst): Rename...
17233 (get_string_len): ...to this. Handle global constants.
17234 (handle_char_store): Adjust.
17236 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17237 Jim Wilson <jimw@sifive.com>
17239 * config/riscv/riscv-protos.h (riscv_output_return): New.
17240 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17241 (riscv_attribute_table, riscv_output_return),
17242 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17243 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17244 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17246 (riscv_expand_prologue): Add early return for naked function.
17247 (riscv_expand_epilogue): Likewise.
17248 (riscv_function_ok_for_sibcall): Return false for naked function.
17249 (riscv_set_current_function): New.
17250 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17251 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17252 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17253 * doc/extend.texi (RISC-V Function Attributes): New.
17255 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17257 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17258 check for 128-bit long double before checking TCmode.
17259 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17260 128-bit long doubles before checking TFmode or TCmode.
17261 (FLOAT128_IBM_P): Likewise.
17263 2018-01-10 Martin Sebor <msebor@redhat.com>
17265 PR tree-optimization/83671
17266 * builtins.c (c_strlen): Unconditionally return zero for the empty
17268 Use -Warray-bounds for warnings.
17269 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17270 for non-constant array indices with COMPONENT_REF, arrays of
17271 arrays, and pointers to arrays.
17272 (gimple_fold_builtin_strlen): Determine and set length range for
17273 non-constant character arrays.
17275 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17277 PR middle-end/81897
17278 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17281 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17283 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17285 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17288 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17289 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17290 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17291 indexed_or_indirect_operand predicate.
17292 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17293 (*vsx_le_perm_load_v8hi): Likewise.
17294 (*vsx_le_perm_load_v16qi): Likewise.
17295 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17296 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17297 (*vsx_le_perm_store_v8hi): Likewise.
17298 (*vsx_le_perm_store_v16qi): Likewise.
17299 (eight unnamed splitters): Likewise.
17301 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17303 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17304 * config/rs6000/emmintrin.h: Likewise.
17305 * config/rs6000/mmintrin.h: Likewise.
17306 * config/rs6000/xmmintrin.h: Likewise.
17308 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17311 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17313 * tree.c (tree_nop_conversion): Return true for location wrapper
17315 (maybe_wrap_with_location): New function.
17316 (selftest::check_strip_nops): New function.
17317 (selftest::test_location_wrappers): New function.
17318 (selftest::tree_c_tests): Call it.
17319 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17320 (maybe_wrap_with_location): New decl.
17321 (EXPR_LOCATION_WRAPPER_P): New macro.
17322 (location_wrapper_p): New inline function.
17323 (tree_strip_any_location_wrapper): New inline function.
17325 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17328 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17329 stack_realign_offset for the largest alignment of stack slot
17331 (ix86_find_max_used_stack_alignment): New function.
17332 (ix86_finalize_stack_frame_flags): Use it. Set
17333 max_used_stack_alignment if we don't realign stack.
17334 * config/i386/i386.h (machine_function): Add
17335 max_used_stack_alignment.
17337 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17339 * config/arm/arm.opt (-mbranch-cost): New option.
17340 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17343 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17346 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17347 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17349 2018-01-10 Richard Biener <rguenther@suse.de>
17352 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17353 early out so it also covers the case where we have a non-NULL
17356 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17358 PR tree-optimization/83753
17359 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17360 for non-strided grouped accesses if the number of elements is 1.
17362 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17365 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17366 * i386.h (TARGET_USE_GATHER): Define.
17367 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17369 2018-01-10 Martin Liska <mliska@suse.cz>
17372 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17373 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17375 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17376 CLEANUP_NO_PARTITIONING is not set.
17378 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17380 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17381 for vectors, as a partial revert of r254296.
17382 * rtl.h (const_vec_p): Delete.
17383 (const_vec_duplicate_p): Don't test for vector CONSTs.
17384 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17385 * expmed.c (make_tree): Likewise.
17388 * common.md (E, F): Use CONSTANT_P instead of checking for
17390 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17391 checking for CONST_VECTOR.
17393 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17395 PR middle-end/83575
17396 * predict.c (force_edge_cold): Handle in more sane way edges
17397 with no prediction.
17399 2018-01-09 Carl Love <cel@us.ibm.com>
17401 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17403 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17404 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17405 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17406 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17407 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17408 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17409 * config/rs6000/rs6000-protos.h: Add extern defition for
17410 rs6000_generate_float2_double_code.
17411 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17413 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17414 (float2_v2df): Add define_expand.
17416 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17419 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17420 op_mode in the force_to_mode call.
17422 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17424 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17425 instead of checking each element individually.
17426 (aarch64_evpc_uzp): Likewise.
17427 (aarch64_evpc_zip): Likewise.
17428 (aarch64_evpc_ext): Likewise.
17429 (aarch64_evpc_rev): Likewise.
17430 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17431 instead of checking each element individually. Return true without
17433 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17434 whether all selected elements come from the same input, instead of
17435 checking each element individually. Remove calls to gen_rtx_REG,
17436 start_sequence and end_sequence and instead assert that no rtl is
17439 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17441 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17442 order of HIGH and CONST checks.
17444 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17446 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17447 if the destination isn't an SSA_NAME.
17449 2018-01-09 Richard Biener <rguenther@suse.de>
17451 PR tree-optimization/83668
17452 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17454 (canonicalize_loop_form): ... here, renamed from ...
17455 (canonicalize_loop_closed_ssa_form): ... this and amended to
17456 swap successor edges for loop exit blocks to make us use
17457 the RPO order we need for initial schedule generation.
17459 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17461 PR tree-optimization/64811
17462 * match.pd: When optimizing comparisons with Inf, avoid
17463 introducing or losing exceptions from comparisons with NaN.
17465 2018-01-09 Martin Liska <mliska@suse.cz>
17468 * asan.c (shadow_mem_size): Add gcc_assert.
17470 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17472 Don't save registers in main().
17475 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17476 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17477 * config/avr/avr.c (avr_set_current_function): Don't error if
17478 naked, OS_task or OS_main are specified at the same time.
17479 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17481 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17483 * common/config/avr/avr-common.c (avr_option_optimization_table):
17484 Switch on -mmain-is-OS_task for optimizing compilations.
17486 2018-01-09 Richard Biener <rguenther@suse.de>
17488 PR tree-optimization/83572
17489 * graphite.c: Include cfganal.h.
17490 (graphite_transform_loops): Connect infinite loops to exit
17491 and remove fake edges at the end.
17493 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17495 * ipa-inline.c (edge_badness): Revert accidental checkin.
17497 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17500 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17501 symbols; not inline clones.
17503 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17506 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17507 hard registers. Formatting fixes.
17509 PR preprocessor/83722
17510 * gcc.c (try_generate_repro): Pass
17511 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17512 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17515 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17516 Kito Cheng <kito.cheng@gmail.com>
17518 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17519 (riscv_leaf_function_p): Delete.
17520 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17522 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17524 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17526 (do_ifelse): New function.
17527 (do_isel): New function.
17528 (do_sub3): New function.
17529 (do_add3): New function.
17530 (do_load_mask_compare): New function.
17531 (do_overlap_load_compare): New function.
17532 (expand_compare_loop): New function.
17533 (expand_block_compare): Call expand_compare_loop() when appropriate.
17534 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17535 option description.
17536 (-mblock-compare-inline-loop-limit): New option.
17538 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17541 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17542 Reverse order of second and third operands in first alternative.
17543 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17544 of first and second elements in UNSPEC_VPERMR vector.
17545 (altivec_expand_vec_perm_le): Likewise.
17547 2018-01-08 Jeff Law <law@redhat.com>
17549 PR rtl-optimizatin/81308
17550 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17551 (process_switch): If group_case_labels makes a change, then set
17553 (pass_convert_switch::execute): If a switch is converted, then
17554 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17556 PR rtl-optimization/81308
17557 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17560 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17562 PR target/83663 - Revert r255946
17563 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17564 generation for cases where splatting a value is not useful.
17565 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17566 across a vec_duplicate and a paradoxical subreg forming a vector
17567 mode to a vec_concat.
17569 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17571 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17572 -march=armv8.3-a variants.
17573 * config/arm/t-multilib: Likewise.
17574 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17576 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17578 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17580 (cceq_ior_compare_complement): Give it a name so I can use it, and
17581 change boolean_or_operator predicate to boolean_operator so it can
17582 be used to generate a crand.
17583 (eqne): New code iterator.
17584 (bd/bd_neg): New code_attrs.
17585 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17586 a single define_insn.
17587 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17588 decrement (bdnzt/bdnzf/bdzt/bdzf).
17589 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17590 with the new names of the branch decrement patterns, and added the
17591 names of the branch decrement conditional patterns.
17593 2018-01-08 Richard Biener <rguenther@suse.de>
17595 PR tree-optimization/83563
17596 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17599 2018-01-08 Richard Biener <rguenther@suse.de>
17601 PR middle-end/83713
17602 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17604 2018-01-08 Richard Biener <rguenther@suse.de>
17606 PR tree-optimization/83685
17607 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17608 references to abnormals.
17610 2018-01-08 Richard Biener <rguenther@suse.de>
17613 * dwarf2out.c (output_indirect_strings): Handle empty
17614 skeleton_debug_str_hash.
17615 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17617 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17619 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17620 (emit_store_direct): Likewise.
17621 (arc_trampoline_adjust_address): Likewise.
17622 (arc_asm_trampoline_template): New function.
17623 (arc_initialize_trampoline): Use asm_trampoline_template.
17624 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17625 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17626 * config/arc/arc.md (flush_icache): Delete pattern.
17628 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17630 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17631 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17634 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17637 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17638 by not USED_FOR_TARGET.
17639 (make_pass_resolve_sw_modes): Likewise.
17641 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17643 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17646 2018-01-08 Richard Biener <rguenther@suse.de>
17648 PR middle-end/83580
17649 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17651 2018-01-08 Richard Biener <rguenther@suse.de>
17653 PR middle-end/83517
17654 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17656 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17658 PR middle-end/81897
17659 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17660 basic blocks with a small number of successors.
17661 (convert_control_dep_chain_into_preds): Improve handling of
17663 (dump_predicates): Split apart into...
17664 (dump_pred_chain): ...here...
17665 (dump_pred_info): ...and here.
17666 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17667 (can_chain_union_be_invalidated_p): Improve check for invalidation
17669 (uninit_uses_cannot_happen): Avoid unnecessary if
17670 convert_control_dep_chain_into_preds yielded nothing.
17672 2018-01-06 Martin Sebor <msebor@redhat.com>
17674 PR tree-optimization/83640
17675 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17676 subtracting negative offset from size.
17677 (builtin_access::overlap): Adjust offset bounds of the access to fall
17678 within the size of the object if possible.
17680 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17682 PR rtl-optimization/83699
17683 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17684 extract_bit_field_as_subreg to cases in which the extracted
17685 value is also a vector.
17687 * lra-constraints.c (process_alt_operands): Test for the equivalence
17688 substitutions when detecting a possible reload cycle.
17690 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17693 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17694 by default if flag_selective_schedling{,2}. Formatting fixes.
17696 PR rtl-optimization/83682
17697 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17698 if it has non-VECTOR_MODE element mode.
17699 (vec_duplicate_p): Likewise.
17701 PR middle-end/83694
17702 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17703 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17705 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17708 * config/i386/i386-builtin.def
17709 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17710 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17711 Require also OPTION_MASK_ISA_AVX512F in addition to
17712 OPTION_MASK_ISA_GFNI.
17713 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17714 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17715 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17716 to OPTION_MASK_ISA_GFNI.
17717 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17718 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17719 OPTION_MASK_ISA_AVX512BW.
17720 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17721 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17722 addition to OPTION_MASK_ISA_GFNI.
17723 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17724 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17725 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17726 to OPTION_MASK_ISA_GFNI.
17727 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17728 a requirement for all ISAs rather than any of them with a few
17730 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17732 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17733 bitmasks to be enabled with 3 exceptions, instead of requiring any
17734 enabled ISA with lots of exceptions.
17735 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17736 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17737 Change avx512bw in isa attribute to avx512f.
17738 * config/i386/sgxintrin.h: Add license boilerplate.
17739 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17740 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17741 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17742 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17744 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17745 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17746 temporarily sse2 rather than sse if not enabled already.
17749 * config/i386/sse.md (VI248_VLBW): Rename to ...
17750 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17751 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17752 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17753 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17754 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17755 mode iterator instead of VI248_VLBW.
17757 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17759 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17760 (record_modified): Skip clobbers; add debug output.
17761 (param_change_prob): Use sreal frequencies.
17763 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17765 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17766 punt for user-aligned variables.
17768 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17770 * tree-chrec.c (chrec_contains_symbols): Return true for
17773 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17776 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17777 of (x|y) == x for BICS pattern.
17779 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17781 PR tree-optimization/83605
17782 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17783 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17786 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17788 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17789 * config/epiphany/rtems.h: New file.
17791 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17792 Uros Bizjak <ubizjak@gmail.com>
17795 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17796 QIreg_operand instead of register_operand predicate.
17797 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17798 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17799 comments instead of -fmitigate[-_]rop.
17801 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17804 * cgraphunit.c (symbol_table::compile): Switch to text_section
17805 before calling assembly_start debug hook.
17806 * run-rtl-passes.c (run_rtl_passes): Likewise.
17809 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17811 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17812 range_int_cst_p rather than !symbolic_range_p before calling
17813 extract_range_from_multiplicative_op_1.
17815 2018-01-04 Jeff Law <law@redhat.com>
17817 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17818 redundant test in assertion.
17820 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17822 * doc/rtl.texi: Document machine_mode wrapper classes.
17824 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17826 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17827 using tree_to_uhwi.
17829 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17831 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17832 the VEC_PERM_EXPR fold to fail.
17834 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17837 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17838 to switched_sections.
17840 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17843 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17844 test for d.testing.
17846 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17849 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17850 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17852 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17855 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17856 is BLKmode and bitpos not zero or mode change is needed.
17858 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17861 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17864 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17867 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17868 instead of MULT rtx. Update all corresponding splitters.
17869 (*saddl_se): Ditto.
17870 (*ssub<modesuffix>): Ditto.
17871 (*ssubl_se): Ditto.
17872 (*cmp_sadd_di): Update split patterns.
17873 (*cmp_sadd_si): Ditto.
17874 (*cmp_sadd_sidi): Ditto.
17875 (*cmp_ssub_di): Ditto.
17876 (*cmp_ssub_si): Ditto.
17877 (*cmp_ssub_sidi): Ditto.
17878 * config/alpha/predicates.md (const23_operand): New predicate.
17879 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17880 Look for ASHIFT, not MULT inner operand.
17881 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17883 2018-01-04 Martin Liska <mliska@suse.cz>
17885 PR gcov-profile/83669
17886 * gcov.c (output_intermediate_file): Add version to intermediate
17888 * doc/gcov.texi: Document new field 'version' in intermediate
17889 file format. Fix location of '-k' option of gcov command.
17891 2018-01-04 Martin Liska <mliska@suse.cz>
17894 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17896 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17898 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17900 2018-01-03 Martin Sebor <msebor@redhat.com>
17902 PR tree-optimization/83655
17903 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17904 checking calls with invalid arguments.
17906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17908 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17909 (vectorizable_mask_load_store): Delete.
17910 (vectorizable_call): Return false for masked loads and stores.
17911 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17912 instead of gimple_assign_rhs1.
17913 (vectorizable_load): Handle IFN_MASK_LOAD.
17914 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17918 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17920 (vectorizable_mask_load_store): ...here.
17921 (vectorizable_load): ...and here.
17923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17925 * tree-vect-stmts.c (vect_build_all_ones_mask)
17926 (vect_build_zero_merge_argument): New functions, split out from...
17927 (vectorizable_load): ...here.
17929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17931 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17933 (vectorizable_mask_load_store): ...here.
17934 (vectorizable_store): ...and here.
17936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17938 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17940 (vectorizable_mask_load_store): ...here.
17942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17944 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17945 (vect_model_store_cost): Take a vec_load_store_type instead of a
17947 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17948 (vect_model_store_cost): Take a vec_load_store_type instead of a
17950 (vectorizable_mask_load_store): Update accordingly.
17951 (vectorizable_store): Likewise.
17952 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17956 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17957 IFN_MASK_LOAD calls here rather than...
17958 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17961 Alan Hayward <alan.hayward@arm.com>
17962 David Sherwood <david.sherwood@arm.com>
17964 * expmed.c (extract_bit_field_1): For vector extracts,
17965 fall back to extract_bit_field_as_subreg if vec_extract
17968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17969 Alan Hayward <alan.hayward@arm.com>
17970 David Sherwood <david.sherwood@arm.com>
17972 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17973 they are variable or constant sized.
17974 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17975 slots for constant-sized data.
17977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17978 Alan Hayward <alan.hayward@arm.com>
17979 David Sherwood <david.sherwood@arm.com>
17981 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17982 handling COND_EXPRs with boolean comparisons, try to find a better
17983 basis for the mask type than the boolean itself.
17985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17987 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17988 is calculated and how it can be overridden.
17989 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17990 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17992 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17996 Alan Hayward <alan.hayward@arm.com>
17997 David Sherwood <david.sherwood@arm.com>
17999 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18000 Remove the mode argument.
18001 (aarch64_simd_valid_immediate): Remove the mode and inverse
18003 * config/aarch64/iterators.md (bitsize): New iterator.
18004 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18005 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18006 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18007 aarch64_simd_valid_immediate.
18008 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18009 (aarch64_reg_or_bic_imm): Likewise.
18010 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18011 with an insn_type enum and msl with a modifier_type enum.
18012 Replace element_width with a scalar_mode. Change the shift
18013 to unsigned int. Add constructors for scalar_float_mode and
18014 scalar_int_mode elements.
18015 (aarch64_vect_float_const_representable_p): Delete.
18016 (aarch64_can_const_movi_rtx_p)
18017 (aarch64_simd_scalar_immediate_valid_for_move)
18018 (aarch64_simd_make_constant): Update call to
18019 aarch64_simd_valid_immediate.
18020 (aarch64_advsimd_valid_immediate_hs): New function.
18021 (aarch64_advsimd_valid_immediate): Likewise.
18022 (aarch64_simd_valid_immediate): Remove mode and inverse
18023 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18024 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18025 and aarch64_float_const_representable_p on the result.
18026 (aarch64_output_simd_mov_immediate): Remove mode argument.
18027 Update call to aarch64_simd_valid_immediate and use of
18028 simd_immediate_info.
18029 (aarch64_output_scalar_simd_mov_immediate): Update call
18032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18033 Alan Hayward <alan.hayward@arm.com>
18034 David Sherwood <david.sherwood@arm.com>
18036 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18037 (mode_nunits): Likewise CONST_MODE_NUNITS.
18038 * machmode.def (ADJUST_NUNITS): Document.
18039 * genmodes.c (mode_data::need_nunits_adj): New field.
18040 (blank_mode): Update accordingly.
18041 (adj_nunits): New variable.
18042 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18044 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18045 listed in adj_nunits.
18046 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18047 listed in adj_nunits. Don't emit case statements for such modes.
18048 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18049 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18050 nothing if adj_nunits is nonnull.
18051 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18052 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18053 (emit_mode_fbit): Update use of print_maybe_const_decl.
18054 (emit_move_size): Likewise. Treat the array as non-const
18056 (emit_mode_adjustments): Handle adj_nunits.
18058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18060 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18061 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18062 (VECTOR_MODES): Use it.
18063 (make_vector_modes): Take the prefix as an argument.
18065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18066 Alan Hayward <alan.hayward@arm.com>
18067 David Sherwood <david.sherwood@arm.com>
18069 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18070 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18071 for MODE_VECTOR_BOOL.
18072 * machmode.def (VECTOR_BOOL_MODE): Document.
18073 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18074 (make_vector_bool_mode): New function.
18075 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18077 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18078 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18080 * stor-layout.c (int_mode_for_mode): Likewise.
18081 * tree.c (build_vector_type_for_mode): Likewise.
18082 * varasm.c (output_constant_pool_2): Likewise.
18083 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18084 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18085 for MODE_VECTOR_BOOL.
18086 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18087 of mode class checks.
18088 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18089 instead of a list of mode class checks.
18090 (expand_vector_scalar_condition): Likewise.
18091 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18094 Alan Hayward <alan.hayward@arm.com>
18095 David Sherwood <david.sherwood@arm.com>
18097 * machmode.h (mode_size): Change from unsigned short to
18099 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18100 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18101 or if measurement_type is not polynomial.
18102 (fixed_size_mode::includes_p): Check for constant-sized modes.
18103 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18104 return a poly_uint16 rather than an unsigned short.
18105 (emit_mode_size): Change the type of mode_size from unsigned short
18106 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18107 (emit_mode_adjustments): Cope with polynomial vector sizes.
18108 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18110 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18112 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18113 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18114 * caller-save.c (setup_save_areas): Likewise.
18115 (replace_reg_with_saved_mem): Likewise.
18116 * calls.c (emit_library_call_value_1): Likewise.
18117 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18118 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18119 (gen_lowpart_for_combine): Likewise.
18120 * convert.c (convert_to_integer_1): Likewise.
18121 * cse.c (equiv_constant, cse_insn): Likewise.
18122 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18123 (cselib_subst_to_values): Likewise.
18124 * dce.c (word_dce_process_block): Likewise.
18125 * df-problems.c (df_word_lr_mark_ref): Likewise.
18126 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18127 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18128 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18129 (rtl_for_decl_location): Likewise.
18130 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18131 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18132 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18133 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18134 (expand_expr_real_1): Likewise.
18135 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18136 (pad_below): Likewise.
18137 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18138 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18139 * ira.c (get_subreg_tracking_sizes): Likewise.
18140 * ira-build.c (ira_create_allocno_objects): Likewise.
18141 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18142 (ira_sort_regnos_for_alter_reg): Likewise.
18143 * ira-costs.c (record_operand_costs): Likewise.
18144 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18145 (resolve_simple_move): Likewise.
18146 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18147 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18148 (lra_constraints): Likewise.
18149 (CONST_POOL_OK_P): Reject variable-sized modes.
18150 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18151 (add_pseudo_to_slot, lra_spill): Likewise.
18152 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18153 * optabs-query.c (get_best_extraction_insn): Likewise.
18154 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18155 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18156 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18157 * recog.c (offsettable_address_addr_space_p): Likewise.
18158 * regcprop.c (maybe_mode_change): Likewise.
18159 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18160 * regrename.c (build_def_use): Likewise.
18161 * regstat.c (dump_reg_info): Likewise.
18162 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18163 (find_reloads, find_reloads_subreg_address): Likewise.
18164 * reload1.c (eliminate_regs_1): Likewise.
18165 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18166 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18167 (simplify_binary_operation_1, simplify_subreg): Likewise.
18168 * targhooks.c (default_function_arg_padding): Likewise.
18169 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18170 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18171 (verify_gimple_assign_ternary): Likewise.
18172 * tree-inline.c (estimate_move_cost): Likewise.
18173 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18174 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18175 (get_address_cost_ainc): Likewise.
18176 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18177 (vect_supportable_dr_alignment): Likewise.
18178 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18179 (vectorizable_reduction): Likewise.
18180 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18181 (vectorizable_operation, vectorizable_load): Likewise.
18182 * tree.c (build_same_sized_truth_vector_type): Likewise.
18183 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18184 * var-tracking.c (emit_note_insn_var_location): Likewise.
18185 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18186 (ADDR_VEC_ALIGN): Likewise.
18188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18189 Alan Hayward <alan.hayward@arm.com>
18190 David Sherwood <david.sherwood@arm.com>
18192 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18194 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18195 or if measurement_type is polynomial.
18196 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18197 * combine.c (make_extraction): Likewise.
18198 * dse.c (find_shift_sequence): Likewise.
18199 * dwarf2out.c (mem_loc_descriptor): Likewise.
18200 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18201 (extract_bit_field, extract_low_bits): Likewise.
18202 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18203 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18204 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18205 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18206 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18207 * reload.c (find_reloads): Likewise.
18208 * reload1.c (alter_reg): Likewise.
18209 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18210 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18211 * tree-if-conv.c (predicate_mem_writes): Likewise.
18212 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18213 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18214 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18215 * valtrack.c (dead_debug_insert_temp): Likewise.
18216 * varasm.c (mergeable_constant_section): Likewise.
18217 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18220 Alan Hayward <alan.hayward@arm.com>
18221 David Sherwood <david.sherwood@arm.com>
18223 * expr.c (expand_assignment): Cope with polynomial mode sizes
18224 when assigning to a CONCAT.
18226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18227 Alan Hayward <alan.hayward@arm.com>
18228 David Sherwood <david.sherwood@arm.com>
18230 * machmode.h (mode_precision): Change from unsigned short to
18232 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18234 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18235 or if measurement_type is not polynomial.
18236 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18237 in which the mode is already known to be a scalar_int_mode.
18238 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18239 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18241 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18242 for GET_MODE_PRECISION.
18243 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18244 for GET_MODE_PRECISION.
18245 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18247 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18248 (expand_field_assignment, make_extraction): Likewise.
18249 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18250 (get_last_value): Likewise.
18251 * convert.c (convert_to_integer_1): Likewise.
18252 * cse.c (cse_insn): Likewise.
18253 * expr.c (expand_expr_real_1): Likewise.
18254 * lra-constraints.c (simplify_operand_subreg): Likewise.
18255 * optabs-query.c (can_atomic_load_p): Likewise.
18256 * optabs.c (expand_atomic_load): Likewise.
18257 (expand_atomic_store): Likewise.
18258 * ree.c (combine_reaching_defs): Likewise.
18259 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18260 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18261 * tree.h (type_has_mode_precision_p): Likewise.
18262 * ubsan.c (instrument_si_overflow): Likewise.
18264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18265 Alan Hayward <alan.hayward@arm.com>
18266 David Sherwood <david.sherwood@arm.com>
18268 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18269 polynomial numbers of units.
18270 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18271 (valid_vector_subparts_p): New function.
18272 (build_vector_type): Remove temporary shim and take the number
18273 of units as a poly_uint64 rather than an int.
18274 (build_opaque_vector_type): Take the number of units as a
18275 poly_uint64 rather than an int.
18276 * tree.c (build_vector_from_ctor): Handle polynomial
18277 TYPE_VECTOR_SUBPARTS.
18278 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18279 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18280 (build_vector_from_val): If the number of units is variable,
18281 use build_vec_duplicate_cst for constant operands and
18282 VEC_DUPLICATE_EXPR otherwise.
18283 (make_vector_type): Remove temporary is_constant ().
18284 (build_vector_type, build_opaque_vector_type): Take the number of
18285 units as a poly_uint64 rather than an int.
18286 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18288 * cfgexpand.c (expand_debug_expr): Likewise.
18289 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18290 (store_constructor, expand_expr_real_1): Likewise.
18291 (const_scalar_mask_from_tree): Likewise.
18292 * fold-const-call.c (fold_const_reduction): Likewise.
18293 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18294 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18295 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18296 (fold_relational_const): Likewise.
18297 (native_interpret_vector): Likewise. Change the size from an
18298 int to an unsigned int.
18299 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18300 TYPE_VECTOR_SUBPARTS.
18301 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18302 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18303 duplicating a non-constant operand into a variable-length vector.
18304 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18305 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18306 * ipa-icf.c (sem_variable::equals): Likewise.
18307 * match.pd: Likewise.
18308 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18309 * print-tree.c (print_node): Likewise.
18310 * stor-layout.c (layout_type): Likewise.
18311 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18312 * tree-cfg.c (verify_gimple_comparison): Likewise.
18313 (verify_gimple_assign_binary): Likewise.
18314 (verify_gimple_assign_ternary): Likewise.
18315 (verify_gimple_assign_single): Likewise.
18316 * tree-pretty-print.c (dump_generic_node): Likewise.
18317 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18318 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18319 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18320 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18321 (vect_shift_permute_load_chain): Likewise.
18322 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18323 (expand_vector_condition, optimize_vector_constructor): Likewise.
18324 (lower_vec_perm, get_compute_type): Likewise.
18325 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18326 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18327 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18328 (vect_recog_mask_conversion_pattern): Likewise.
18329 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18330 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18331 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18332 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18333 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18334 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18335 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18336 (supportable_widening_operation): Likewise.
18337 (supportable_narrowing_operation): Likewise.
18338 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18340 * varasm.c (output_constant): Likewise.
18342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18343 Alan Hayward <alan.hayward@arm.com>
18344 David Sherwood <david.sherwood@arm.com>
18346 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18347 so that both the length == 3 and length != 3 cases set up their
18348 own permute vectors. Add comments explaining why we know the
18349 number of elements is constant.
18350 (vect_permute_load_chain): Likewise.
18352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18353 Alan Hayward <alan.hayward@arm.com>
18354 David Sherwood <david.sherwood@arm.com>
18356 * machmode.h (mode_nunits): Change from unsigned char to
18358 (ONLY_FIXED_SIZE_MODES): New macro.
18359 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18360 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18361 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18363 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18364 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18365 or if measurement_type is not polynomial.
18366 * genmodes.c (ZERO_COEFFS): New macro.
18367 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18369 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18370 Use ZERO_COEFFS when emitting initializers.
18371 * data-streamer.h (bp_pack_poly_value): New function.
18372 (bp_unpack_poly_value): Likewise.
18373 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18374 for GET_MODE_NUNITS.
18375 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18376 for GET_MODE_NUNITS.
18377 * tree.c (make_vector_type): Remove temporary shim and make
18378 the real function take the number of units as a poly_uint64
18379 rather than an int.
18380 (build_vector_type_for_mode): Handle polynomial nunits.
18381 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18382 * emit-rtl.c (const_vec_series_p_1): Likewise.
18383 (gen_rtx_CONST_VECTOR): Likewise.
18384 * fold-const.c (test_vec_duplicate_folding): Likewise.
18385 * genrecog.c (validate_pattern): Likewise.
18386 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18387 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18388 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18389 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18390 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18391 * rtlanal.c (subreg_get_info): Likewise.
18392 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18393 (vect_grouped_load_supported): Likewise.
18394 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18395 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18396 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18397 (simplify_const_unary_operation, simplify_binary_operation_1)
18398 (simplify_const_binary_operation, simplify_ternary_operation)
18399 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18400 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18401 instead of CONST_VECTOR_NUNITS.
18402 * varasm.c (output_constant_pool_2): Likewise.
18403 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18404 explicit-encoded elements in the XVEC for variable-length vectors.
18406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18408 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18411 Alan Hayward <alan.hayward@arm.com>
18412 David Sherwood <david.sherwood@arm.com>
18414 * coretypes.h (fixed_size_mode): Declare.
18415 (fixed_size_mode_pod): New typedef.
18416 * builtins.h (target_builtins::x_apply_args_mode)
18417 (target_builtins::x_apply_result_mode): Change type to
18418 fixed_size_mode_pod.
18419 * builtins.c (apply_args_size, apply_result_size, result_vector)
18420 (expand_builtin_apply_args_1, expand_builtin_apply)
18421 (expand_builtin_return): Update accordingly.
18423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18425 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18426 * cselib.c (cselib_hash_rtx): Likewise.
18427 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18428 CONST_VECTOR encoding.
18430 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18431 Jeff Law <law@redhat.com>
18434 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18435 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18436 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18437 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18440 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18441 explicitly probe *sp in a noreturn function if there were any callee
18442 register saves or frame pointer is needed.
18444 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18447 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18448 BLKmode for ternary, binary or unary expressions.
18451 * var-tracking.c (delete_vta_debug_insn): New inline function.
18452 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18453 insns from get_insns () to NULL instead of each bb separately.
18454 Use delete_vta_debug_insn. No longer static.
18455 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18456 delete_vta_debug_insns callers.
18457 * rtl.h (delete_vta_debug_insns): Declare.
18458 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18459 instead of variable_tracking_main.
18461 2018-01-03 Martin Sebor <msebor@redhat.com>
18463 PR tree-optimization/83603
18464 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18465 arguments past the endof the argument list in functions declared
18466 without a prototype.
18467 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18468 Avoid checking when arguments are null.
18470 2018-01-03 Martin Sebor <msebor@redhat.com>
18473 * doc/extend.texi (attribute const): Fix a typo.
18474 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18475 issuing -Wsuggest-attribute for void functions.
18477 2018-01-03 Martin Sebor <msebor@redhat.com>
18479 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18480 offset_int::from instead of wide_int::to_shwi.
18481 (maybe_diag_overlap): Remove assertion.
18482 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18483 * gimple-ssa-sprintf.c (format_directive): Same.
18484 (parse_directive): Same.
18485 (sprintf_dom_walker::compute_format_length): Same.
18486 (try_substitute_return_value): Same.
18488 2018-01-03 Jeff Law <law@redhat.com>
18490 PR middle-end/83654
18491 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18492 non-constant residual for zero at runtime and avoid probing in
18493 that case. Reorganize code for trailing problem to mirror handling
18496 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18498 PR tree-optimization/83501
18499 * tree-ssa-strlen.c (get_string_cst): New.
18500 (handle_char_store): Call get_string_cst.
18502 2018-01-03 Martin Liska <mliska@suse.cz>
18504 PR tree-optimization/83593
18505 * tree-ssa-strlen.c: Include tree-cfg.h.
18506 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18507 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18508 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18510 (strlen_dom_walker::before_dom_children): Call
18511 gimple_purge_dead_eh_edges. Dump tranformation with details
18513 (strlen_dom_walker::before_dom_children): Update call by adding
18514 new argument cleanup_eh.
18515 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18517 2018-01-03 Martin Liska <mliska@suse.cz>
18520 * cif-code.def (VARIADIC_THUNK): New enum value.
18521 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18524 2018-01-03 Jan Beulich <jbeulich@suse.com>
18526 * sse.md (mov<mode>_internal): Tighten condition for when to use
18527 vmovdqu<ssescalarsize> for TI and OI modes.
18529 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18531 Update copyright years.
18533 2018-01-03 Martin Liska <mliska@suse.cz>
18536 * ipa-visibility.c (function_and_variable_visibility): Skip
18537 functions with noipa attribure.
18539 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18541 * gcc.c (process_command): Update copyright notice dates.
18542 * gcov-dump.c (print_version): Ditto.
18543 * gcov.c (print_version): Ditto.
18544 * gcov-tool.c (print_version): Ditto.
18545 * gengtype.c (create_file): Ditto.
18546 * doc/cpp.texi: Bump @copying's copyright year.
18547 * doc/cppinternals.texi: Ditto.
18548 * doc/gcc.texi: Ditto.
18549 * doc/gccint.texi: Ditto.
18550 * doc/gcov.texi: Ditto.
18551 * doc/install.texi: Ditto.
18552 * doc/invoke.texi: Ditto.
18554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18556 * vector-builder.h (vector_builder::m_full_nelts): Change from
18557 unsigned int to poly_uint64.
18558 (vector_builder::full_nelts): Update prototype accordingly.
18559 (vector_builder::new_vector): Likewise.
18560 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18561 (vector_builder::operator ==): Likewise.
18562 (vector_builder::finalize): Likewise.
18563 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18564 Take the number of elements as a poly_uint64 rather than an
18566 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18567 from unsigned int to poly_uint64.
18568 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18569 (vec_perm_indices::new_vector): Likewise.
18570 (vec_perm_indices::length): Likewise.
18571 (vec_perm_indices::nelts_per_input): Likewise.
18572 (vec_perm_indices::input_nelts): Likewise.
18573 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18574 number of elements per input as a poly_uint64 rather than an
18575 unsigned int. Use the original encoding for variable-length
18576 vectors, rather than clamping each individual element.
18577 For the second and subsequent elements in each pattern,
18578 clamp the step and base before clamping their sum.
18579 (vec_perm_indices::series_p): Handle polynomial element counts.
18580 (vec_perm_indices::all_in_range_p): Likewise.
18581 (vec_perm_indices_to_tree): Likewise.
18582 (vec_perm_indices_to_rtx): Likewise.
18583 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18584 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18585 (tree_vector_builder::new_binary_operation): Handle polynomial
18586 element counts. Return false if we need to know the number
18587 of elements at compile time.
18588 * fold-const.c (fold_vec_perm): Punt if the number of elements
18589 isn't known at compile time.
18591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18593 * vec-perm-indices.h (vec_perm_builder): Change element type
18594 from HOST_WIDE_INT to poly_int64.
18595 (vec_perm_indices::element_type): Update accordingly.
18596 (vec_perm_indices::clamp): Handle polynomial element_types.
18597 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18598 (vec_perm_indices::all_in_range_p): Likewise.
18599 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18601 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18602 polynomial vec_perm_indices element types.
18603 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18604 * fold-const.c (fold_vec_perm): Likewise.
18605 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18606 * tree-vect-generic.c (lower_vec_perm): Likewise.
18607 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18608 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18609 element type to HOST_WIDE_INT.
18611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18612 Alan Hayward <alan.hayward@arm.com>
18613 David Sherwood <david.sherwood@arm.com>
18615 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18616 rather than an int. Use plus_constant.
18617 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18618 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18621 Alan Hayward <alan.hayward@arm.com>
18622 David Sherwood <david.sherwood@arm.com>
18624 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18625 a HOST_WIDE_INT to a poly_int64.
18627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18628 Alan Hayward <alan.hayward@arm.com>
18629 David Sherwood <david.sherwood@arm.com>
18631 * calls.c (load_register_parameters): Cope with polynomial
18632 mode sizes. Require a constant size for BLKmode parameters
18633 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18634 forces a parameter to be padded at the lsb end in order to
18635 fill a complete number of words, require the parameter size
18636 to be ordered wrt UNITS_PER_WORD.
18638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18639 Alan Hayward <alan.hayward@arm.com>
18640 David Sherwood <david.sherwood@arm.com>
18642 * reload1.c (spill_stack_slot_width): Change element type
18643 from unsigned int to poly_uint64_pod.
18644 (alter_reg): Treat mode sizes as polynomial.
18646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18647 Alan Hayward <alan.hayward@arm.com>
18648 David Sherwood <david.sherwood@arm.com>
18650 * reload.c (complex_word_subreg_p): New function.
18651 (reload_inner_reg_of_subreg, push_reload): Use it.
18653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18654 Alan Hayward <alan.hayward@arm.com>
18655 David Sherwood <david.sherwood@arm.com>
18657 * lra-constraints.c (process_alt_operands): Reject matched
18658 operands whose sizes aren't ordered.
18659 (match_reload): Refer to this check here.
18661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18662 Alan Hayward <alan.hayward@arm.com>
18663 David Sherwood <david.sherwood@arm.com>
18665 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18666 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18673 Use plus_constant instead of gen_rtx_PLUS.
18675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18676 Alan Hayward <alan.hayward@arm.com>
18677 David Sherwood <david.sherwood@arm.com>
18679 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18680 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18681 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18682 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18683 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18684 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18685 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18686 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18687 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18688 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18690 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18691 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18692 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18693 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18694 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18695 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18696 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18697 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18698 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18699 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18701 * expr.c (emit_move_resolve_push): Treat the input and result
18702 of PUSH_ROUNDING as a poly_int64.
18703 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18704 (emit_push_insn): Likewise.
18705 * lra-eliminations.c (mark_not_eliminable): Likewise.
18706 * recog.c (push_operand): Likewise.
18707 * reload1.c (elimination_effects): Likewise.
18708 * rtlanal.c (nonzero_bits1): Likewise.
18709 * calls.c (store_one_arg): Likewise. Require the padding to be
18710 known at compile time.
18712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18713 Alan Hayward <alan.hayward@arm.com>
18714 David Sherwood <david.sherwood@arm.com>
18716 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18717 Use plus_constant instead of gen_rtx_PLUS.
18719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18720 Alan Hayward <alan.hayward@arm.com>
18721 David Sherwood <david.sherwood@arm.com>
18723 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18724 rather than an int.
18726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18727 Alan Hayward <alan.hayward@arm.com>
18728 David Sherwood <david.sherwood@arm.com>
18730 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18731 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18732 via stack temporaries. Treat the mode size as polynomial too.
18734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18735 Alan Hayward <alan.hayward@arm.com>
18736 David Sherwood <david.sherwood@arm.com>
18738 * expr.c (expand_expr_real_2): When handling conversions involving
18739 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18740 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18741 as a poly_uint64 too.
18743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18744 Alan Hayward <alan.hayward@arm.com>
18745 David Sherwood <david.sherwood@arm.com>
18747 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18750 Alan Hayward <alan.hayward@arm.com>
18751 David Sherwood <david.sherwood@arm.com>
18753 * combine.c (can_change_dest_mode): Handle polynomial
18754 REGMODE_NATURAL_SIZE.
18755 * expmed.c (store_bit_field_1): Likewise.
18756 * expr.c (store_constructor): Likewise.
18757 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18758 and polynomial REGMODE_NATURAL_SIZE.
18759 (gen_lowpart_common): Likewise.
18760 * reginfo.c (record_subregs_of_mode): Likewise.
18761 * rtlanal.c (read_modify_subreg_p): Likewise.
18763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18764 Alan Hayward <alan.hayward@arm.com>
18765 David Sherwood <david.sherwood@arm.com>
18767 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18768 numbers of elements.
18770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18771 Alan Hayward <alan.hayward@arm.com>
18772 David Sherwood <david.sherwood@arm.com>
18774 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18781 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
18788 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18789 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18796 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18797 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18798 (expand_vector_condition, vector_element): Likewise.
18799 (subparts_gt): New function.
18800 (get_compute_type): Use subparts_gt.
18801 (count_type_subparts): Delete.
18802 (expand_vector_operations_1): Use subparts_gt instead of
18803 count_type_subparts.
18805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18806 Alan Hayward <alan.hayward@arm.com>
18807 David Sherwood <david.sherwood@arm.com>
18809 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18810 (vect_compile_time_alias): ...this new function. Do the calculation
18811 on poly_ints rather than trees.
18812 (vect_prune_runtime_alias_test_list): Update call accordingly.
18814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18815 Alan Hayward <alan.hayward@arm.com>
18816 David Sherwood <david.sherwood@arm.com>
18818 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18820 (vect_schedule_slp_instance): Likewise.
18822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18823 Alan Hayward <alan.hayward@arm.com>
18824 David Sherwood <david.sherwood@arm.com>
18826 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18827 constant and extern definitions for variable-length vectors.
18828 (vect_get_constant_vectors): Note that the number of units
18829 is known to be constant.
18831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18832 Alan Hayward <alan.hayward@arm.com>
18833 David Sherwood <david.sherwood@arm.com>
18835 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18836 of units as polynomial. Choose between WIDE and NARROW based
18839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18840 Alan Hayward <alan.hayward@arm.com>
18841 David Sherwood <david.sherwood@arm.com>
18843 * tree-vect-stmts.c (simd_clone_subparts): New function.
18844 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18847 Alan Hayward <alan.hayward@arm.com>
18848 David Sherwood <david.sherwood@arm.com>
18850 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18851 vectors as polynomial. Use build_index_vector for
18852 IFN_GOMP_SIMD_LANE.
18854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18855 Alan Hayward <alan.hayward@arm.com>
18856 David Sherwood <david.sherwood@arm.com>
18858 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18859 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18860 for variable-length vectors.
18861 (vectorizable_mask_load_store): Treat the number of units as
18862 polynomial, asserting that it is constant if the condition has
18863 already been enforced.
18864 (vectorizable_store, vectorizable_load): Likewise.
18866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18867 Alan Hayward <alan.hayward@arm.com>
18868 David Sherwood <david.sherwood@arm.com>
18870 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18871 of units as polynomial. Punt if we can't tell at compile time
18872 which vector contains the final result.
18874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18875 Alan Hayward <alan.hayward@arm.com>
18876 David Sherwood <david.sherwood@arm.com>
18878 * tree-vect-loop.c (vectorizable_induction): Treat the number
18879 of units as polynomial. Punt on SLP inductions. Use an integer
18880 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18881 cast of such a series for variable-length floating-point
18884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18885 Alan Hayward <alan.hayward@arm.com>
18886 David Sherwood <david.sherwood@arm.com>
18888 * tree.h (build_index_vector): Declare.
18889 * tree.c (build_index_vector): New function.
18890 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18891 of units as polynomial, forcibly converting it to a constant if
18892 vectorizable_reduction has already enforced the condition.
18893 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18894 to create a {1,2,3,...} vector.
18895 (vectorizable_reduction): Treat the number of units as polynomial.
18896 Choose vectype_in based on the largest scalar element size rather
18897 than the smallest number of units. Enforce the restrictions
18900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18901 Alan Hayward <alan.hayward@arm.com>
18902 David Sherwood <david.sherwood@arm.com>
18904 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18905 number of units as polynomial.
18907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18908 Alan Hayward <alan.hayward@arm.com>
18909 David Sherwood <david.sherwood@arm.com>
18911 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18912 * target.def (autovectorize_vector_sizes): Return the vector sizes
18913 by pointer, using vector_sizes rather than a bitmask.
18914 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18915 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18916 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18918 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18919 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18920 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18921 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18922 * omp-general.c (omp_max_vf): Likewise.
18923 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18924 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18925 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18926 * tree-vect-slp.c (vect_slp_bb): Likewise.
18927 * doc/tm.texi: Regenerate.
18928 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18930 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18931 the vector size as a poly_uint64 rather than an unsigned int.
18932 (current_vector_size): Change from an unsigned int to a poly_uint64.
18933 (get_vectype_for_scalar_type): Update accordingly.
18934 * tree.h (build_truth_vector_type): Take the size and number of
18935 units as a poly_uint64 rather than an unsigned int.
18936 (build_vector_type): Add a temporary overload that takes
18937 the number of units as a poly_uint64 rather than an unsigned int.
18938 * tree.c (make_vector_type): Likewise.
18939 (build_truth_vector_type): Take the number of units as a poly_uint64
18940 rather than an unsigned int.
18942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18943 Alan Hayward <alan.hayward@arm.com>
18944 David Sherwood <david.sherwood@arm.com>
18946 * target.def (get_mask_mode): Take the number of units and length
18947 as poly_uint64s rather than unsigned ints.
18948 * targhooks.h (default_get_mask_mode): Update accordingly.
18949 * targhooks.c (default_get_mask_mode): Likewise.
18950 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18951 * doc/tm.texi: Regenerate.
18953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18954 Alan Hayward <alan.hayward@arm.com>
18955 David Sherwood <david.sherwood@arm.com>
18957 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18958 * omp-general.c (omp_max_vf): Likewise.
18959 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18960 (expand_omp_simd): Handle polynomial safelen.
18961 * omp-low.c (omplow_simd_context): Add a default constructor.
18962 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18963 (lower_rec_simd_input_clauses): Update accordingly.
18964 (lower_rec_input_clauses): Likewise.
18966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18967 Alan Hayward <alan.hayward@arm.com>
18968 David Sherwood <david.sherwood@arm.com>
18970 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18971 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18972 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18973 (vect_analyze_slp_cost): Likewise.
18974 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18975 (vect_model_load_cost): Likewise.
18977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18978 Alan Hayward <alan.hayward@arm.com>
18979 David Sherwood <david.sherwood@arm.com>
18981 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18982 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18983 from an unsigned int * to a poly_uint64_pod *.
18984 (calculate_unrolling_factor): New function.
18985 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18988 Alan Hayward <alan.hayward@arm.com>
18989 David Sherwood <david.sherwood@arm.com>
18991 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18992 from an unsigned int to a poly_uint64.
18993 (_loop_vec_info::slp_unrolling_factor): Likewise.
18994 (_loop_vec_info::vectorization_factor): Change from an int
18996 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18997 (vect_get_num_vectors): New function.
18998 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18999 (vect_get_num_copies): Use vect_get_num_vectors.
19000 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19001 to an unsigned int *.
19002 (vect_analyze_data_refs): Change min_vf from an int * to a
19004 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19005 than an unsigned HOST_WIDE_INT.
19006 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19007 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19008 to an unsigned int *.
19009 (vect_analyze_data_ref_dependences): Likewise.
19010 (vect_compute_data_ref_alignment): Handle polynomial vf.
19011 (vect_enhance_data_refs_alignment): Likewise.
19012 (vect_prune_runtime_alias_test_list): Likewise.
19013 (vect_shift_permute_load_chain): Likewise.
19014 (vect_supportable_dr_alignment): Likewise.
19015 (dependence_distance_ge_vf): Take the vectorization factor as a
19016 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19017 (vect_analyze_data_refs): Change min_vf from an int * to a
19019 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19020 vfm1 as a poly_uint64 rather than an int. Make the same change
19021 for the returned bound_scalar.
19022 (vect_gen_vector_loop_niters): Handle polynomial vf.
19023 (vect_do_peeling): Likewise. Update call to
19024 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19025 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19027 * tree-vect-loop.c (vect_determine_vectorization_factor)
19028 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19029 (vect_get_known_peeling_cost): Likewise.
19030 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19031 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19032 (vect_transform_loop): Likewise. Use the lowest possible VF when
19033 updating the upper bounds of the loop.
19034 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19035 rather than an int.
19036 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19037 polynomial unroll factors.
19038 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19039 (vect_make_slp_decision): Likewise.
19040 (vect_supported_load_permutation_p): Likewise, and polynomial
19042 (vect_analyze_slp_cost): Handle polynomial vf.
19043 (vect_slp_analyze_node_operations): Likewise.
19044 (vect_slp_analyze_bb_1): Likewise.
19045 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19046 than an unsigned HOST_WIDE_INT.
19047 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19048 (vectorizable_load): Handle polynomial vf.
19049 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19051 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19054 Alan Hayward <alan.hayward@arm.com>
19055 David Sherwood <david.sherwood@arm.com>
19057 * match.pd: Handle bit operations involving three constants
19058 and try to fold one pair.
19060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19062 * tree-vect-loop-manip.c: Include gimple-fold.h.
19063 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19064 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19065 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19066 Add a path that uses a step of VF instead of 1, but disable it
19068 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19069 and niters_no_overflow parameters. Update calls to
19070 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19071 Create a new SSA name if the latter choses to use a ste other
19072 than zero, and return it via niters_vector_mult_vf_var.
19073 * tree-vect-loop.c (vect_transform_loop): Update calls to
19074 vect_do_peeling, vect_gen_vector_loop_niters and
19075 slpeel_make_loop_iterate_ntimes.
19076 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19077 (vect_gen_vector_loop_niters): Update declarations after above changes.
19079 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19081 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19082 128-bit round to integer instructions.
19083 (ceil<mode>2): Likewise.
19084 (btrunc<mode>2): Likewise.
19085 (round<mode>2): Likewise.
19087 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19089 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19090 unaligned VSX load/store on P8/P9.
19091 (expand_block_clear): Allow the use of unaligned VSX
19092 load/store on P8/P9.
19094 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19096 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19098 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19099 swap associated with both a load and a store.
19101 2018-01-02 Andrew Waterman <andrew@sifive.com>
19103 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19104 * config/riscv/riscv.md (clear_cache): Use it.
19106 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19108 * web.c: Remove out-of-date comment.
19110 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19112 * expr.c (fixup_args_size_notes): Check that any existing
19113 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19114 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19115 (emit_single_push_insn): ...here.
19117 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19119 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19120 (const_vector_encoded_nelts): New function.
19121 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19122 (const_vector_int_elt, const_vector_elt): Declare.
19123 * emit-rtl.c (const_vector_int_elt_1): New function.
19124 (const_vector_elt): Likewise.
19125 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19126 of CONST_VECTOR_ELT.
19128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19130 * expr.c: Include rtx-vector-builder.h.
19131 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19132 directly on the tree encoding.
19133 (const_vector_from_tree): Likewise.
19134 * optabs.c: Include rtx-vector-builder.h.
19135 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19136 sequence of "u" values.
19137 * vec-perm-indices.c: Include rtx-vector-builder.h.
19138 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19139 directly on the vec_perm_indices encoding.
19141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19143 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19144 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19145 * rtx-vector-builder.h: New file.
19146 * rtx-vector-builder.c: Likewise.
19147 * rtl.h (rtx_def::u2): Add a const_vector field.
19148 (CONST_VECTOR_NPATTERNS): New macro.
19149 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19150 (CONST_VECTOR_DUPLICATE_P): Likewise.
19151 (CONST_VECTOR_STEPPED_P): Likewise.
19152 (CONST_VECTOR_ENCODED_ELT): Likewise.
19153 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19154 (unwrap_const_vec_duplicate): Likewise.
19155 (const_vec_series_p): Check for a non-duplicated vector encoding.
19156 Say that the function only returns true for integer vectors.
19157 * emit-rtl.c: Include rtx-vector-builder.h.
19158 (gen_const_vec_duplicate_1): Delete.
19159 (gen_const_vector): Call gen_const_vec_duplicate instead of
19160 gen_const_vec_duplicate_1.
19161 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19162 (gen_const_vec_duplicate): Use rtx_vector_builder.
19163 (gen_const_vec_series): Likewise.
19164 (gen_rtx_CONST_VECTOR): Likewise.
19165 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19166 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19167 Build a new vector rather than modifying a CONST_VECTOR in-place.
19168 (handle_special_swappables): Update call accordingly.
19169 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19170 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19171 Build a new vector rather than modifying a CONST_VECTOR in-place.
19172 (handle_special_swappables): Update call accordingly.
19174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19176 * simplify-rtx.c (simplify_const_binary_operation): Use
19177 CONST_VECTOR_ELT instead of XVECEXP.
19179 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19181 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19182 the selector elements to be different from the data elements
19183 if the selector is a VECTOR_CST.
19184 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19185 ssizetype for the selector.
19187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19189 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19190 before testing each element individually.
19191 * tree-vect-generic.c (lower_vec_perm): Likewise.
19193 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19195 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19196 * selftest-run-tests.c (selftest::run_tests): Call it.
19197 * vector-builder.h (vector_builder::operator ==): New function.
19198 (vector_builder::operator !=): Likewise.
19199 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19200 (vec_perm_indices::all_from_input_p): New function.
19201 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19202 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19203 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19204 instead of reading the VECTOR_CST directly. Detect whether both
19205 vector inputs are the same before constructing the vec_perm_indices,
19206 and update the number of inputs argument accordingly. Use the
19207 utility functions added above. Only construct sel2 if we need to.
19209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19211 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19212 the broadcast of the low byte.
19213 (expand_mult_highpart): Use an explicit encoding for the permutes.
19214 * optabs-query.c (can_mult_highpart_p): Likewise.
19215 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19216 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19217 (vectorizable_bswap): Likewise.
19218 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19219 explicit encoding for the power-of-2 permutes.
19220 (vect_permute_store_chain): Likewise.
19221 (vect_grouped_load_supported): Likewise.
19222 (vect_permute_load_chain): Likewise.
19224 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19226 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19227 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19228 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19229 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19230 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19231 (vect_gen_perm_mask_any): Likewise.
19233 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19235 * int-vector-builder.h: New file.
19236 * vec-perm-indices.h: Include int-vector-builder.h.
19237 (vec_perm_indices): Redefine as an int_vector_builder.
19238 (auto_vec_perm_indices): Delete.
19239 (vec_perm_builder): Redefine as a stand-alone class.
19240 (vec_perm_indices::vec_perm_indices): New function.
19241 (vec_perm_indices::clamp): Likewise.
19242 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19243 (vec_perm_indices::new_vector): New function.
19244 (vec_perm_indices::new_expanded_vector): Update for new
19245 vec_perm_indices class.
19246 (vec_perm_indices::rotate_inputs): New function.
19247 (vec_perm_indices::all_in_range_p): Operate directly on the
19248 encoded form, without computing elided elements.
19249 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19250 encoding. Update for new vec_perm_indices class.
19251 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19252 the given vec_perm_builder.
19253 (expand_vec_perm_var): Update vec_perm_builder constructor.
19254 (expand_mult_highpart): Use vec_perm_builder instead of
19255 auto_vec_perm_indices.
19256 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19257 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19258 or double series encoding as appropriate.
19259 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19260 vec_perm_indices instead of auto_vec_perm_indices.
19261 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19262 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19263 (vect_permute_store_chain): Likewise.
19264 (vect_grouped_load_supported): Likewise.
19265 (vect_permute_load_chain): Likewise.
19266 (vect_shift_permute_load_chain): Likewise.
19267 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19268 (vect_transform_slp_perm_load): Likewise.
19269 (vect_schedule_slp_instance): Likewise.
19270 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19271 (vectorizable_mask_load_store): Likewise.
19272 (vectorizable_bswap): Likewise.
19273 (vectorizable_store): Likewise.
19274 (vectorizable_load): Likewise.
19275 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19276 vec_perm_indices instead of auto_vec_perm_indices. Use
19277 tree_to_vec_perm_builder to read the vector from a tree.
19278 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19279 vec_perm_builder instead of a vec_perm_indices.
19280 (have_whole_vector_shift): Use vec_perm_builder and
19281 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19282 truncation to calc_vec_perm_mask_for_shift.
19283 (vect_create_epilog_for_reduction): Likewise.
19284 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19285 from auto_vec_perm_indices to vec_perm_indices.
19286 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19287 instead of changing individual elements.
19288 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19289 the vector in d.perm.
19290 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19291 from auto_vec_perm_indices to vec_perm_indices.
19292 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19293 instead of changing individual elements.
19294 (arm_vectorize_vec_perm_const): Use new_vector to install
19295 the vector in d.perm.
19296 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19297 Update vec_perm_builder constructor.
19298 (rs6000_expand_interleave): Likewise.
19299 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19300 (rs6000_expand_interleave): Likewise.
19302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19304 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19305 to qimode could truncate the indices.
19306 * optabs.c (expand_vec_perm_var): Likewise.
19308 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19310 * Makefile.in (OBJS): Add vec-perm-indices.o.
19311 * vec-perm-indices.h: New file.
19312 * vec-perm-indices.c: Likewise.
19313 * target.h (vec_perm_indices): Replace with a forward class
19315 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19316 * optabs.h: Include vec-perm-indices.h.
19317 (expand_vec_perm): Delete.
19318 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19319 (expand_vec_perm_const): Declare.
19320 * target.def (vec_perm_const_ok): Replace with...
19321 (vec_perm_const): ...this new hook.
19322 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19323 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19324 * doc/tm.texi: Regenerate.
19325 * optabs.def (vec_perm_const): Delete.
19326 * doc/md.texi (vec_perm_const): Likewise.
19327 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19328 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19329 expand_vec_perm for constant permutation vectors. Assert that
19330 the mode of variable permutation vectors is the integer equivalent
19331 of the mode that is being permuted.
19332 * optabs-query.h (selector_fits_mode_p): Declare.
19333 * optabs-query.c: Include vec-perm-indices.h.
19334 (selector_fits_mode_p): New function.
19335 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19336 is defined, instead of checking whether the vec_perm_const_optab
19337 exists. Use targetm.vectorize.vec_perm_const instead of
19338 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19339 fit in the vector mode before using a variable permute.
19340 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19341 vec_perm_indices instead of an rtx.
19342 (expand_vec_perm): Replace with...
19343 (expand_vec_perm_const): ...this new function. Take the selector
19344 as a vec_perm_indices rather than an rtx. Also take the mode of
19345 the selector. Update call to shift_amt_for_vec_perm_mask.
19346 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19347 Use vec_perm_indices::new_expanded_vector to expand the original
19348 selector into bytes. Check whether the indices fit in the vector
19349 mode before using a variable permute.
19350 (expand_vec_perm_var): Make global.
19351 (expand_mult_highpart): Use expand_vec_perm_const.
19352 * fold-const.c: Includes vec-perm-indices.h.
19353 * tree-ssa-forwprop.c: Likewise.
19354 * tree-vect-data-refs.c: Likewise.
19355 * tree-vect-generic.c: Likewise.
19356 * tree-vect-loop.c: Likewise.
19357 * tree-vect-slp.c: Likewise.
19358 * tree-vect-stmts.c: Likewise.
19359 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19361 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19362 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19363 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19364 (aarch64_vectorize_vec_perm_const): ...this new function.
19365 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19366 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19367 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19368 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19369 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19370 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19371 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19373 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19374 check for NEON modes.
19375 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19376 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19377 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19378 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19380 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19381 the old VEC_PERM_CONST conditions.
19382 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19383 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19384 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19385 (ia64_vectorize_vec_perm_const_ok): Merge into...
19386 (ia64_vectorize_vec_perm_const): ...this new function.
19387 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19388 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19389 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19390 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19391 * config/mips/mips.c (mips_expand_vec_perm_const)
19392 (mips_vectorize_vec_perm_const_ok): Merge into...
19393 (mips_vectorize_vec_perm_const): ...this new function.
19394 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19395 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19396 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19397 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19398 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19399 (rs6000_expand_vec_perm_const): Delete.
19400 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19402 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19403 (altivec_expand_vec_perm_const_le): Take each operand individually.
19404 Operate on constant selectors rather than rtxes.
19405 (altivec_expand_vec_perm_const): Likewise. Update call to
19406 altivec_expand_vec_perm_const_le.
19407 (rs6000_expand_vec_perm_const): Delete.
19408 (rs6000_vectorize_vec_perm_const_ok): Delete.
19409 (rs6000_vectorize_vec_perm_const): New function.
19410 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19411 an element count and rtx array.
19412 (rs6000_expand_extract_even): Update call accordingly.
19413 (rs6000_expand_interleave): Likewise.
19414 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19415 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19416 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19417 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19418 (rs6000_expand_vec_perm_const): Delete.
19419 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19420 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19421 (altivec_expand_vec_perm_const_le): Take each operand individually.
19422 Operate on constant selectors rather than rtxes.
19423 (altivec_expand_vec_perm_const): Likewise. Update call to
19424 altivec_expand_vec_perm_const_le.
19425 (rs6000_expand_vec_perm_const): Delete.
19426 (rs6000_vectorize_vec_perm_const_ok): Delete.
19427 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19428 reference to the SPE evmerge intructions.
19429 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19430 an element count and rtx array.
19431 (rs6000_expand_extract_even): Update call accordingly.
19432 (rs6000_expand_interleave): Likewise.
19433 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19434 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19436 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19438 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19440 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19441 vector mode and that that mode matches the mode of the data
19443 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19444 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19445 directly using expand_vec_perm_1 when forcing selectors into
19447 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19449 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19451 * optabs-query.h (can_vec_perm_p): Delete.
19452 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19453 * optabs-query.c (can_vec_perm_p): Split into...
19454 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19455 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19456 particular selector is valid.
19457 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19458 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19459 (vect_grouped_load_supported): Likewise.
19460 (vect_shift_permute_load_chain): Likewise.
19461 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19462 (vect_transform_slp_perm_load): Likewise.
19463 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19464 (vectorizable_bswap): Likewise.
19465 (vect_gen_perm_mask_checked): Likewise.
19466 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19467 implementations of variable permutation vectors into account
19468 when deciding which selector to use.
19469 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19470 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19471 with a false third argument.
19472 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19473 to test whether the constant selector is valid and can_vec_perm_var_p
19474 to test whether a variable selector is valid.
19476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19478 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19479 * optabs-query.c (can_vec_perm_p): Likewise.
19480 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19481 instead of vec_perm_indices.
19482 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19483 (vect_gen_perm_mask_checked): Likewise,
19484 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19485 (vect_gen_perm_mask_checked): Likewise,
19487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19489 * optabs-query.h (qimode_for_vec_perm): Declare.
19490 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19491 (qimode_for_vec_perm): ...this new function.
19492 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19494 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19496 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19497 does not have a conditional at the top.
19499 2018-01-02 Richard Biener <rguenther@suse.de>
19501 * ipa-inline.c (big_speedup_p): Fix expression.
19503 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19506 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19509 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19513 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19514 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19515 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19516 cond_taken_branch_cost 3->4.
19518 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19520 PR tree-optimization/83581
19521 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19522 TODO_cleanup_cfg if any changes have been made.
19524 PR middle-end/83608
19525 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19526 convert_modes if target mode has the right side, but different mode
19529 PR middle-end/83609
19530 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19531 last argument when extracting from CONCAT. If either from_real or
19532 from_imag is NULL, use expansion through memory. If result is not
19533 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19534 the parts directly to inner mode, if even that fails, use expansion
19537 PR middle-end/83623
19538 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19539 check for bswap in mode rather than HImode and use that in expand_unop
19542 Copyright (C) 2018 Free Software Foundation, Inc.
19544 Copying and distribution of this file, with or without modification,
19545 are permitted in any medium without royalty provided the copyright
19546 notice and this notice are preserved.