1 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5 (hash_tree): Do not hash DECL_FCONTEXT
6 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7 Do not stream DECL_FCONTEXT.
8 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
11 2018-07-12 Richard Biener <rguenther@suse.de>
14 * dwarf2out.c (gen_block_die): Only output blocks when they have
15 at least one !DECL_IGNORED_P variable.
17 2018-07-12 Richard Biener <rguenther@suse.de>
20 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
21 Remove -mieee-fp handling.
23 2018-07-12 Richard Biener <rguenther@suse.de>
25 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
26 left-over from last patch.
28 2018-07-12 Jakub Jelinek <jakub@redhat.com>
30 PR tree-optimization/86492
31 * gimple-ssa-store-merging.c
32 (imm_store_chain_info::coalesce_immediate_stores): Call
33 check_no_overlap even for the merge_overlapping case. Formatting fix.
35 2018-07-12 Richard Biener <rguenther@suse.de>
38 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
39 move possibly trapping operations into the conditional.
41 2018-07-12 Richard Biener <rguenther@suse.de>
43 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
44 (vn_lookup_simplify_result): Remove recursion limit applied
46 (vn_nary_build_or_lookup_1): Adjust.
47 (try_to_simplify): Likewise.
48 * gimple-match-head.c (gimple_resimplify1): Instead apply one
50 (gimple_resimplify2): Likewise.
51 (gimple_resimplify3): Likewise.
52 (gimple_resimplify4): Likewise.
54 2018-07-11 Jakub Jelinek <jakub@redhat.com>
56 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
57 Use __mmask64 type instead of __mmask8 for __M argument.
58 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
59 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
61 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
62 __mmask16 for __M argument.
63 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
64 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
65 to __mmask16 instead of __mmask8.
66 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
67 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
68 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
69 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
70 instead of __mmask16 for __U argument.
71 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
72 __mmask16 instead of __mmask8 for __U argument.
73 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
75 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
77 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
79 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
81 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
83 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
85 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
87 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
89 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
91 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
93 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
94 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
95 return type as well as __M argument type and all casts from __mmask8
97 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
98 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
99 return type as well as __M argument type and all casts from __mmask8
101 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
102 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
103 return type as well as __M argument type and all casts from __mmask8
105 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
106 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
107 return type as well as __M argument type and all casts from __mmask8
109 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
110 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
113 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
115 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
116 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
119 2018-07-11 Paul Koning <ni1d@arrl.net>
121 * doc/md.texi (define_subst): Document how multiple occurrences of
122 the same argument in the replacement pattern are handled.
124 2018-07-11 Paul Koning <ni1d@arrl.net>
126 * doc/extend.texi (Common Variable Attributes): Move "mode" into
128 (Common Type Attributes): Add "mode" attribute.
130 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
132 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
133 stream DECL_ORIGINAL_TYPE.
134 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
135 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
136 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
137 Do not walk original type.
138 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
140 (write_ts_decl_non_common_tree_pointers): Do not stream
142 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
143 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
145 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
147 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
148 thread twice from the same starting edge.
150 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
152 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
153 * gimple.c (gimple_call_nonnull_result_p): ...here...
154 (gimple_call_nonnull_arg): ...and here.
155 * gimple.h (gimple_call_nonnull_result_p): New.
156 (gimple_call_nonnull_arg): New.
158 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
160 * config/arm/arm-cpus.in: Move information from fpu field of each
161 cpu definition to the isa field.
162 * config/arm/parsecpu.awk (fpu): Delete match rule.
163 (gen_comm_data): Don't add bits from the CPU's FPU entry.
165 2018-07-11 Richard Biener <rguenther@suse.de>
168 * dwarf2out.c (init_sections_and_labels): Use
169 output_asm_line_debug_info consistently.
170 (dwarf2out_early_finish): Likewise.
171 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
174 2018-07-11 Richard Biener <rguenther@suse.de>
176 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
177 Rework father_bb setting in a way to avoid propagating constants
178 multiple times on a loop body.
180 2018-07-10 Mark Wielaard <mark@klomp.org>
183 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
185 2018-07-10 Richard Biener <rguenther@suse.de>
187 * hash-map.h (hash_map::iterator::operator*): Return
188 references to key and value.
190 2018-07-10 Jakub Jelinek <jakub@redhat.com>
193 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
194 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
195 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
196 (gimplify_omp_for): For composite loops, move outer
197 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
198 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
199 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
200 TREE_LIST for both the original class iterator and the "last" helper
201 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
202 loop, remember has_decl_expr from outer composite loops for the
203 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
205 2018-07-09 Martin Sebor <msebor@redhat.com>
209 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
210 accessing implicitly initialized array elements.
211 * expr.c (string_constant): Handle string initializers of
212 character arrays within aggregates.
213 * gimple-fold.c (fold_array_ctor_reference): Add argument.
214 Store element offset. As a special case, handle zero size.
215 (fold_nonarray_ctor_reference): Same.
216 (fold_ctor_reference): Add argument. Store subobject offset.
217 * gimple-fold.h (fold_ctor_reference): Add argument.
219 2018-07-09 Paul Koning <ni1d@arrl.net>
221 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
222 (pdp11_insn_cost): New function.
223 (pdp11_md_asm_adjust): New function.
224 (TARGET_INVALID_WITHIN_DOLOOP): Define.
225 (pdp11_rtx_costs): Update to match machine better.
226 (output_addr_const_pdp11): Correct format mismatch warnings.
227 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
228 * config/pdp11/pdp11.md: General change to add base_cost and/or
229 length attributes for use by new pdp11_insn_cost function.
230 (MIN_BRANCH): Correct definition.
232 (doloop_end): Use standard pattern name for looping pattern.
233 (doloop_end_nocc): New.
234 (movsf): Add another constraint alternative.
235 (zero_extendqihi2): Add constraint alternatives for not in place
237 (zero_extendhisi2): Remove.
238 (shift patterns): Add CC handling variants.
242 (define_peephole2): New peephole to recognize mov that sets CC for
245 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
247 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
248 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
251 2018-07-09 Jakub Jelinek <jakub@redhat.com>
254 * real.c (real_nextafter): Return true if result is denormal.
256 2018-07-09 Martin Liska <mliska@suse.cz>
258 * common.opt: Add back wrongly removed attribute.
260 2018-07-09 Richard Biener <rguenther@suse.de>
263 * dwarf2out.c (gen_block_die): For an early generated DIE
264 always output high/low PC attributes.
266 2018-07-09 Tom de Vries <tdevries@suse.de>
268 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
269 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
270 onto VAR_DECL with abstract origin.
272 2018-07-07 Jim Wilson <jimw@sifive.com>
274 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
276 2018-07-07 Tom de Vries <tdevries@suse.de>
278 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
280 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
282 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
284 (combine_bound): Use wide-int overflow calculation instead of
286 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
288 * fold-const.c (int_const_binop_2): Same.
289 (extract_muldiv_1): Same.
290 (fold_div_compare): Same.
291 (fold_abs_const): Same.
293 * poly-int.h (add): Same.
297 * predict.c (predict_iv_comparison): Same.
298 * profile-count.c (slow_safe_scale_64bit): Same.
299 * simplify-rtx.c (simplify_const_binary_operation): Same.
300 * tree-chrec.c (tree_fold_binomial): Same.
301 * tree-data-ref.c (split_constant_offset_1): Same.
302 * tree-if-conv.c (idx_within_array_bound): Same.
303 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
304 * tree-ssa-phiopt.c (minmax_replacement): Same.
305 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
306 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
307 * vr-values.c (vr_values::adjust_range_with_scev): Same.
308 * wide-int.cc (wi::add_large): Same.
309 (wi::mul_internal): Same.
310 (wi::sub_large): Same.
311 (wi::divmod_internal): Same.
312 * wide-int.h: Change overflow type to overflow_type for neg, add,
313 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
314 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
315 mul_internal, divmod_internal.
316 (overflow_type): New enum.
317 (accumulate_overflow): New.
319 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
321 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
322 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
324 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
326 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
327 argument is checked for zero before entering loop, avoid checking again.
329 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
331 * gimplify.h (generic_expr_could_trap_p): Set as global function.
332 * gimplify.h (generic_expr_could_trap_p): Likwise.
333 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
335 2018-07-06 Jakub Jelinek <jakub@redhat.com>
337 PR tree-optimization/86401
338 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
339 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
340 (fold_bit_and_mask): ... here. New helper function for match.pd.
341 * fold-const.h (fold_bit_and_mask): Declare.
342 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
344 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
347 * target.def (translate_mode_attribute): New hook.
348 * targhooks.h (default_translate_mode_attribute): Declare.
349 * targhooks.c (default_translate_mode_attribute): New function.
350 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
351 * doc/tm.texi: Regenerate.
352 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
353 (rs6000_translate_mode_attribute): New function.
355 2018-07-06 Paul Koning <ni1d@arrl.net>
357 * doc/md.texi (define_split): Document DONE and FAIL.
358 (define_peephole2): Ditto.
360 2018-07-05 Jeff Law <law@redhat.com>
362 PR tree-optimization/86010
363 * tree-ssa-dse.c (compute_trims): More aggressively trim at
364 both the head and tail of mem* and str* calls.
366 2018-07-05 Jim Wilson <jimw@sifive.com>
368 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
370 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
372 * config/aarch64/aarch64-simd.md: correct flags text for
373 MIN_EXPR replacement.
375 2018-07-05 James Clarke <jrtc27@jrtc27.com>
377 * configure: Regenerated.
379 2018-07-05 Carl Love <cel@us.ibm.com>
381 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
382 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
383 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
384 VSX_BUILTIN_DOUBLEL_V4SF.
386 2018-07-05 Martin Sebor <msebor@redhat.com>
389 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
390 than its domain to compute its the upper bound of a char array.
392 2018-07-05 Nathan Sidwell <nathan@acm.org>
394 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
395 * doc/cpp.texi: Update comment.
396 * doc/tm.texi: Rebuilt.
397 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
398 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
399 * doc/extend.texi (Backwards Compatibility): Clarify it is system
400 headers affected by extern "C".
401 * system.h: Poison NO_IMPLICIT_EXTERN_C.
402 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
403 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
404 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
405 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
406 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
407 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
408 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
409 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
410 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
411 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
412 config/sparc/sp64-elf.h, config/spu/spu.h,
413 config/stormy16/stormy16.h, config/v850/v850.h,
414 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
415 define NO_IMPLICIT_EXTERN_C.
416 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
418 2018-07-05 Tamar Christina <tamar.christina@arm.com>
421 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
422 instead of GET_MODE_SIZE when comparing Units.
424 2018-07-05 Tamar Christina <tamar.christina@arm.com>
427 * rtlanal.c (set_noop_p): Constrain on mode change,
428 include hard-reg-set.h
430 2018-07-05 Tamar Christina <tamar.christina@arm.com>
432 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
434 2018-07-05 Jakub Jelinek <jakub@redhat.com>
437 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
440 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
442 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
444 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
447 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
449 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
451 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
454 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
456 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
458 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
461 2018-07-04 Martin Liska <mliska@suse.cz>
463 * tree-switch-conversion.c: Define
464 max_ratio_for_speed and max_ratio_for_size constants.
466 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
467 Martin Liska <mliska@suse.cz>
472 * common.opt: Rename align options with 'str_' prefix.
473 * common/config/i386/i386-common.c (set_malign_value): New
475 (ix86_handle_option): Use it to set -falign-* options/
476 * config/aarch64/aarch64-protos.h (struct tune_params): Change
477 type from int to string.
478 * config/aarch64/aarch64.c: Update default values from int
480 * config/alpha/alpha.c (alpha_override_options_after_change):
482 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
483 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
484 max skip conditionally.
485 * config/i386/freebsd.h (SUBALIGN_LOG): New.
486 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
487 max skip conditionally.
488 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
489 max skip conditionally.
490 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
491 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
492 max skip conditionally.
493 * config/i386/i386.c (struct ptt): Change type from int to
495 (ix86_default_align): Set default values.
496 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
497 max skip conditionally.
498 * config/i386/iamcu.h (SUBALIGN_LOG): New.
499 (ASM_OUTPUT_MAX_SKIP_ALIGN):
500 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
501 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
502 max skip conditionally.
503 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
504 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
505 * config/i386/x86-64.h (SUBALIGN_LOG): New.
506 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
507 max skip conditionally.
508 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
509 * config/ia64/ia64.c (ia64_option_override): Set default values
510 for alignment options.
511 * config/m68k/m68k.c: Handle new str_align_* options.
512 * config/mips/mips.c (mips_set_compression_mode): Change
514 (mips_option_override): Set default values for options.
515 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
517 * config/rs6000/rs6000.c (rs6000_option_override_internal):
519 * config/rx/rx.c (rx_option_override): Likewise.
520 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
521 (LABEL_ALIGN): Use align_labels_log.
522 (LOOP_ALIGN): Use align_loops_align.
523 * config/s390/s390.c (s390_asm_output_function_label): Use new
525 * config/sh/sh.c (sh_override_options_after_change):
526 Change type of constants.
527 * config/spu/spu.c (spu_sched_init): Likewise.
528 * config/sparc/sparc.c (sparc_option_override): Set default
530 * config/visium/visium.c (visium_option_override): Likewise.
531 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
532 emit p2align format with last argument if it's not needed.
533 * doc/invoke.texi: Document extended format of -falign-*.
534 * final.c: Use align_labels alignment.
535 * flags.h (struct target_flag_state): Change type to use
537 (struct align_flags_tuple): New.
538 (struct align_flags): Likewise.
539 (align_loops_log): Redefine macro to use new types.
540 (align_loops_max_skip): Redefine macro to use new types.
541 (align_jumps_log): Redefine macro to use new types.
542 (align_jumps_max_skip): Redefine macro to use new types.
543 (align_labels_log): Redefine macro to use new types.
544 (align_labels_max_skip): Redefine macro to use new types.
545 (align_functions_log): Redefine macro to use new types.
546 (align_loops): Redefine macro to use new types.
547 (align_jumps): Redefine macro to use new types.
548 (align_labels): Redefine macro to use new types.
549 (align_functions): Redefine macro to use new types.
550 (align_functions_max_skip): Redefine macro to use new types.
551 (align_loops_value): New macro.
552 (align_jumps_value): New macro.
553 (align_labels_value): New macro.
554 (align_functions_value): New macro.
555 * function.c (invoke_set_current_function_hook): Propagate
556 alignment values from flags to global variables default in
558 * ipa-icf.c (sem_function::equals_wpa): Use
559 cl_optimization_option_eq instead of memcmp.
560 * lto-streamer.h (cl_optimization_stream_out): Support streaming
562 (cl_optimization_stream_in): Likewise.
563 * optc-save-gen.awk: Support strings in cl_optimization.
564 * opth-gen.awk: Likewise.
565 * opts.c (finish_options): Remove error checking of invalid
567 (MAX_CODE_ALIGN): Remove.
568 (MAX_CODE_ALIGN_VALUE): Likewise.
569 (parse_and_check_align_values): New function.
570 (check_alignment_argument): Likewise.
571 (common_handle_option): Use check_alignment_argument.
572 * opts.h (parse_and_check_align_values): Declare.
573 * toplev.c (init_alignments): Remove.
574 (read_log_maxskip): New.
575 (parse_N_M): Likewise.
576 (parse_alignment_opts): Likewise.
577 (backend_init_target): Remove usage of init_alignments.
578 * toplev.h (parse_alignment_opts): Declare.
579 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
581 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
582 * tree.c (cl_option_hasher::equal): New.
583 * varasm.c: Use new global macros.
585 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
587 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
588 Use a simpler align directive also if MAXSKIP = ALIGN-1.
589 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
590 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
591 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
592 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
593 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
594 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
595 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
596 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
597 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
598 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
599 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
601 2018-07-04 Martin Liska <mliska@suse.cz>
602 Jonathan Wakely <jwakely@redhat.com>
604 * coverage.c: Use correct type.
605 * doc/invoke.texi: Language correction.
607 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
610 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
611 ENDBRANCH for non-tail call which may return via indirect branch.
612 * doc/extend.texi: Document indirect_return attribute.
614 2018-07-03 Martin Sebor <msebor@redhat.com>
616 PR tree-optimization/86274
617 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
619 (format_floating): Correct handling of infinities and NaNs.
621 2018-07-03 Martin Sebor <msebor@redhat.com>
623 * print-tree.c (print_real_cst): New function.
624 (print_node_brief): Call it.
627 2018-07-03 Jeff Law <law@redhat.com>
629 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
630 into a single pattern.
632 * config/h8300/h8300.md (ors code_iterator): New.
633 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
634 a single pattern and single splitter.
635 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
636 (iorqi3_1, xorqi3_1): Likewise.
637 (iorqi3, xorqi3 expanders): Similarly.
639 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
640 (movmd_internal) into a single pattern using the P mode iterator.
641 (movmd splitters): Similarly.
642 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
643 (movsd splitters): Similarly.
645 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
646 ADDB, ADDW and ADDL into a single ADD attribute which selects the
647 right table based on the size of the operand.
648 * config/h8300/h8300.md (length_table): Corresponding changes. All
649 references to "addb", "addw" and "addl" changed to "add".
650 (btst patterns): Merge two variants into a single pattern.
651 (tstqi, tsthi): Likewise.
652 (addhi3_incdec, addsi3_incdec): Likewise.
653 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
654 (mulhi3, mulsi3): Likewise.
655 (udivhi3, udivsi3): Likewise.
656 (divhi3, divsi3): Likewise.
657 (andorqi3, andorhi3, andorsi3): Likewise.
659 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
662 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
663 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
665 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
667 PR tree-optimization/85694
668 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
669 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
671 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
672 (<u>avg<mode>3_ceil): New patterns.
674 2018-07-03 David Malcolm <dmalcolm@redhat.com>
676 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
677 scan-tree-dump directive.
678 * gcc.dg/vect/slp-perm-2.c: Likewise.
679 * gcc.dg/vect/slp-perm-3.c: Likewise.
680 * gcc.dg/vect/slp-perm-5.c: Likewise.
681 * gcc.dg/vect/slp-perm-6.c: Likewise.
682 * gcc.dg/vect/slp-perm-7.c: Likewise.
683 * gcc.dg/vect/slp-perm-8.c: Likewise.
685 2018-07-03 Marek Polacek <polacek@redhat.com>
688 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
690 2018-07-03 Richard Biener <rguenther@suse.de>
693 * tree-ssa-structalias.c (find_func_clobbers): Properly
694 handle indirect calls.
696 2018-07-03 Jeff Law <law@redhat.com>
698 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
699 (shifts): New code iterator.
700 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
701 expander. Fix HImode handling on H8/SX.
702 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
703 (subqi3, subhi3, subsi3 expanders): Likewise.
704 (andqi3, andhi3, andsi3 expanders): Likewise.
705 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
706 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
707 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
708 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
709 (zero_extendqihi2, zero_extendqisi2): Likewise.
710 (extendqihi2, extendqisi2): Likewise.
711 (rotlqi3, rotlhi3, rotlsi3): Likewise.
712 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
713 (rotlqi3_1, rotlhi3_1): Likewise.
714 (logicalhi3_sn, logicalsi3_sn): Likewise.
715 (logicalhi3, logicalsi3): Likewise.
717 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
719 * tree-vect-patterns.c (vect_recog_rotate_pattern)
720 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
721 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
722 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
723 type to append_pattern_def_seq instead of creating a stmt_vec_info
725 (build_mask_conversion): Likewise. Remove vinfo argument.
726 (vect_add_conversion_to_patterm): Likewise, renaming to...
727 (vect_add_conversion_to_pattern): ...this.
728 (vect_recog_mask_conversion_pattern): Update call to
729 build_mask_conversion. Pass the vector type to
730 append_pattern_def_seq here too.
731 (vect_recog_gather_scatter_pattern): Update call to
732 vect_add_conversion_to_pattern.
734 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
736 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
737 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
738 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
739 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
740 STMT_VINFO_PATTERN_DEF_SEQ to null here.
741 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
742 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
743 append_pattern_def_seq instead of new_pattern_def_seq.
744 (vect_recog_divmod_pattern): Do both of the above.
745 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
748 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
750 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
751 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
752 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
753 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
754 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
755 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
756 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
757 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
758 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
759 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
760 parameter with a single stmt_vec_info.
761 (vect_recog_func_ptr): Likewise.
762 (vect_recog_gather_scatter_pattern): Likewise, folding in...
763 (vect_try_gather_scatter_pattern): ...this.
764 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
765 the stmt_vec_info of the statement to be matched. Don't clear
766 STMT_VINFO_RELATED_STMT.
767 (vect_pattern_recog): Update call accordingly.
769 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
771 PR tree-optimization/85694
772 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
773 (uavgM3_ceil): Document new optabs.
774 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
775 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
777 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
778 (savg_ceil_optab): New optabs.
779 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
780 (vect_vect_recog_func_ptrs): Add it.
781 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
782 constant directly from the associated lhs.
784 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
786 * tree-vect-patterns.c (vect_split_statement): New function.
787 (vect_convert_input): Use it to try to split an existing cast.
789 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
791 * poly-int.h (print_hex): New function.
792 * dumpfile.h (dump_dec, dump_hex): Declare.
793 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
794 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
795 min_input_precision, operation_precision and operation_sign.
796 * tree-vect-patterns.c (vect_get_range_info): New function.
797 (vect_same_loop_or_bb_p, vect_single_imm_use)
798 (vect_operation_fits_smaller_type): Delete.
799 (vect_look_through_possible_promotion): Add an optional
800 single_use_p parameter.
801 (vect_recog_over_widening_pattern): Rewrite to use new
802 stmt_vec_info infomration. Handle one operation at a time.
803 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
804 (vect_truncatable_operation_p, vect_set_operation_type)
805 (vect_set_min_input_precision): New functions.
806 (vect_determine_min_output_precision_1): Likewise.
807 (vect_determine_min_output_precision): Likewise.
808 (vect_determine_precisions_from_range): Likewise.
809 (vect_determine_precisions_from_users): Likewise.
810 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
811 (vect_vect_recog_func_ptrs): Put over_widening first.
813 (vect_pattern_recog): Call vect_determine_precisions.
815 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
817 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
818 statements that have been replaced by further pattern statements.
819 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
821 2018-07-03 Richard Biener <rguenther@suse.de>
823 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
824 always set *dt. Dump vectype in vectype overload.
825 * dumpfile.h (dump_gimple_expr): New function.
826 (dump_gimple_expr_loc): Likewise.
827 * dumpfile.c (dump_gimple_expr): New function.
828 (dump_gimple_expr_loc): Likewise.
830 2018-07-02 Jeff Law <law@redhat.com>
832 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
833 the H8/300, H8/300H and H8/S variants into a single pattern.
834 (movhi_h8300, movqi_h8300hs): Similarly.
835 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
836 (QHI mode iterator): New.
838 * config/h8300/h8300.md: Remove trailing whitespace.
840 2018-07-02 Jim Wilson <jimw@sifive.com>
842 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
843 instead of emit_insn for interrupt returns.
844 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
845 (riscv_sret, riscv_uret): Likewise.
847 2018-07-02 David Malcolm <dmalcolm@redhat.com>
849 * pretty-print.c (selftest::test_pp_format): Move save and restore
850 of quotes to class auto_fix_quotes, and add an instance.
851 * selftest.c: Include "intl.h".
852 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
853 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
854 * selftest.h (selftest::auto_fix_quotes): New class.
856 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
858 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
859 (aarch64_sve_prepare_conditional_op): Remove.
860 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
861 Allow aarch64_simd_reg_or_zero as select operand; remove
862 the aarch64_sve_prepare_conditional_op call.
863 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
864 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
865 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
866 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
867 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
868 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
869 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
870 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
871 and a splitters to match all of the *_any patterns.
872 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
874 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
875 (SVE_COND_FP_BINARY_REV): Remove.
876 (sve_int_op_rev, sve_fp_op_rev): New.
877 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
878 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
879 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
880 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
881 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
882 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
883 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
884 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
885 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
887 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
888 Remove match_dup 1 from the inner unspec.
889 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
891 * config/aarch64/aarch64.md (movprfx): New attr.
892 (length): Default movprfx to 8.
893 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
894 (*madd<SVE_I>, *msub<SVE_I): Likewise.
895 (*<su>mul<SVE_I>3_highpart): Likewise.
896 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
897 (*v<ASHIFT><SVE_I>3): Likewise.
898 (*<su><MAXMIN><SVE_I>3): Likewise.
899 (*<su><MAXMIN><SVE_F>3): Likewise.
900 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
901 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
902 (*div<SVE_F>4): Likewise.
904 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
906 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
909 2018-07-02 Richard Biener <rguenther@suse.de>
911 PR tree-optimization/86363
912 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
913 memset argument refers to a non-variable address.
915 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
917 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
918 {PLUS,MINUS}_EXPR code to...
919 (adjust_symbolic_bound): ...here,
920 (combine_bound): ...here,
921 (set_value_range_with_overflow): ...and here.
923 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
925 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
927 (extract_range_from_abs_expr): ...here.
929 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
931 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
932 -fno-omit-frame-pointer when not optimizing.
934 2018-07-02 Martin Liska <mliska@suse.cz>
937 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
938 (propagate_nothrow): Likewise.
940 2018-07-02 Martin Liska <mliska@suse.cz>
943 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
945 2018-07-02 David Malcolm <dmalcolm@redhat.com>
947 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
948 function in r262149, changing "loc" param from source_location to
949 const dump_location_t &.
950 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
951 declaration, as above.
953 2018-07-01 Paul Koning <ni1d@arrl.net>
955 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
956 -munit-asm, -mgnu-asm, -mdec-asm.
957 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
958 (pdp11_output_labelref): New.
959 (pdp11_output_def): New.
960 (pdp11_output_addr_vec_elt): New.
961 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
962 %# and %@ format codes.
963 (pdp11_option_override): New.
964 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
965 (pdp11_output_ident): New.
966 (pdp11_asm_named_section): New.
967 (pdp11_asm_init_sections): New.
968 (pdp11_file_start): New.
969 (pdp11_file_end): New.
970 (output_ascii): Use .ascii/.asciz for -mdec-asm.
971 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
972 %o, like %c but octal.
973 (pdp11_option_override): New.
974 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
976 (DATA_SECTION_ASM_OP): Ditto.
977 (READONLY_DATA_SECTION_ASM_OP): New.
978 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
979 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
980 (ASM_OUTPUT_LABELREF): Ditto.
981 (ASM_OUTPUT_DEF): Ditto.
982 (ASM_OUTPUT_EXTERNAL): New.
983 (ASM_OUTPUT_SOURCE_FILENAME): New.
984 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
985 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
986 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
987 %# and %@ format codes.
988 * config/pdp11/pdp11.opt (mgnu-asm): New.
989 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
990 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
991 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
993 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
995 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
996 dereferencing path[] beyond its length.
998 (debug_all_paths): New.
999 (rewire_first_differing_edge): New.
1000 (adjust_paths_after_duplication): New.
1001 (duplicate_thread_path): Call adjust_paths_after_duplication.
1003 (thread_through_all_blocks): Add new argument to
1004 duplicate_thread_path.
1006 2018-06-30 Jim Wilson <jimw@sifive.com>
1008 * config/riscv/predicates.md (p2m1_shift_operand): New.
1009 (high_mask_shift_operand): New.
1010 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1011 pattern using p2m1_shift_operand.
1012 (lshsi3_zero_extend_3+2): New combiner pattern using
1013 high_mask_shift_operand.
1015 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1017 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1019 (vect_recog_rotate_pattern): ...here.
1020 (vect_convert_input): Try to insert casts of invariants in the
1022 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1023 preheader to be empty.
1025 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1027 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1028 vector type. If given, install it in the new statement's
1030 (vect_element_precision): New function.
1031 (vect_unpromoted_value): New struct.
1032 (vect_unpromoted_value::vect_unpromoted_value): New function.
1033 (vect_unpromoted_value::set_op): Likewise.
1034 (vect_look_through_possible_promotion): Likewise.
1035 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1036 (vect_widened_op_tree, vect_convert_input): Likewise.
1037 (vect_convert_inputs, vect_convert_output): Likewise.
1038 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1039 to handle the optional cast of the multiplication result and
1040 vect_widened_op_tree to detect the widened multiplication itself.
1041 Do not require the input and output of promotion casts to have
1042 the same sign, but base the signedness of the operation on the
1043 input rather than the result. If the pattern includes two
1044 promotions, check that those promotions have the same sign.
1045 Do not restrict the MULT_EXPR handling to a double-width result;
1046 handle quadruple-width results and wider. Use vect_convert_inputs
1047 to convert the inputs to the common type.
1048 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1049 to handle the optional cast of the ABS result. Also allow a sign
1050 change or a sign extension between the ABS and MINUS.
1051 Use vect_widened_op_tree to detect the widened subtraction and use
1052 vect_convert_inputs to convert the inputs to the common type.
1053 (vect_handle_widen_op_by_const): Delete.
1054 (vect_recog_widen_op_pattern): New function.
1055 (vect_recog_widen_mult_pattern): Use it.
1056 (vect_recog_widen_shift_pattern): Likewise.
1057 (vect_recog_widen_sum_pattern): Use
1058 vect_look_through_possible_promotion to handle the promoted
1061 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1063 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1064 the containing gimple_seq *.
1065 * gimple-iterator.h (gsi_for_stmt): Declare it.
1066 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1067 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1068 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1069 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1070 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1072 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1074 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1075 statement being replaced is part of an existing pattern
1076 definition sequence, inserting the new pattern statements before
1078 (vect_pattern_recog_1): Don't return a bool. If the statement
1079 is already part of a pattern, instead apply pattern matching
1080 to the pattern definition statements. Don't clear the
1081 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1082 (vect_pattern_recog): Don't break after the first match;
1083 continue processing the pattern definition statements instead.
1084 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1086 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1088 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1089 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1090 (vect_recog_widen_sum_pattern): Use it.
1092 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1094 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1095 phi is not a pattern statement and has not been replaced by
1096 a pattern statement.
1097 * tree-vect-patterns.c (type_conversion_p): Don't check
1098 STMT_VINFO_IN_PATTERN_P.
1099 (vect_recog_vector_vector_shift_pattern): Likewise.
1100 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1101 the pattern statement rather than the original statement; check
1102 directly for a WIDEN_MULT_EXPR here.
1103 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1104 vect_is_simple_use to return the pattern statement rather
1105 than the original statement; use is_pattern_stmt_p to check
1106 for such a pattern statement.
1107 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1108 to return the pattern statement rather than the original statement;
1109 don't do the same transformation here.
1110 (vect_is_simple_use): If the defining statement has been replaced
1111 by a pattern statement, return the pattern statement instead.
1112 Remove the corresponding (local) transformation from the vectype
1115 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1117 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1118 end and default to null.
1119 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1120 (vectorizable_reduction): Update calls accordingly, dropping the
1121 gimple ** argument if the passed-back statement isn't needed.
1122 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1123 (vect_recog_rotate_pattern): Likewise.
1124 (vect_recog_mask_conversion_pattern): Likewise.
1125 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1126 (vect_mask_constant_operand_p): Likewise.
1127 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1128 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1129 (get_group_load_store_type, get_load_store_type): Likewise.
1130 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1131 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1132 (vectorizable_conversion, vectorizable_assignment): Likewise.
1133 (vectorizable_shift, vectorizable_operation): Likewise.
1134 (vectorizable_store, vect_is_simple_cond): Likewise.
1135 (vectorizable_condition, vectorizable_comparison): Likewise.
1136 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1137 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1138 and move it to the end. Cope with null def_stmt_outs.
1140 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1142 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1144 2018-06-29 Jeff Law <law@redhat.com>
1146 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1147 displacements for TARGET_V850E2V3 and newer.
1148 (TARGET_LRA_P): Remove. Defaults to LRA now.
1149 * config/v850/v850.md (sign23byte_load): Remove.
1150 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1151 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1153 2018-06-29 Martin Liska <mliska@suse.cz>
1156 * coverage.c (coverage_init): Mangle full path name.
1157 * doc/invoke.texi: Document the change.
1158 * gcov-io.c (mangle_path): New.
1159 * gcov-io.h (mangle_path): Likewise.
1160 * gcov.c (mangle_name): Use mangle_path for path mangling.
1162 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1164 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1165 if starting source register is not even.
1167 2018-06-29 Martin Liska <mliska@suse.cz>
1169 PR tree-optimization/86263
1170 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1171 Make edge redirection.
1173 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1175 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1176 (dump_scope_depth): New variable.
1177 (get_dump_scope_depth): New function.
1178 (dump_begin_scope): New function.
1179 (dump_end_scope): New function.
1180 * dumpfile.h (get_dump_scope_depth): New declaration.
1181 (dump_begin_scope): New declaration.
1182 (dump_end_scope): New declaration.
1183 (class auto_dump_scope): New class.
1184 (AUTO_DUMP_SCOPE): New macro.
1185 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1188 2018-06-29 Richard Biener <rguenther@suse.de>
1190 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1191 compute_all_dependences succeeds.
1192 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1193 exceed --param loop-max-datarefs-for-datadeps.
1195 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1197 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1199 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1202 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1203 alternative 0 in preferred_for_speed attribute.
1205 2018-06-28 Paul Koning <ni1d@arrl.net>
1207 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1208 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1209 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1210 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1211 for shift insn patterns.
1213 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1215 * cgraph.c (cgraph_node::get_body): Replace assignments to
1216 "dump_file" with calls to set_dump_file.
1217 * dumpfile.c (alt_dump_file): Make static, and group with...
1218 (alt_flags): ...this definition.
1219 (dumps_are_enabled): New variable.
1220 (refresh_dumps_are_enabled): New function.
1221 (set_dump_file): New function.
1222 (set_alt_dump_file): New function.
1223 (gcc::dump_manager::dump_start): Replace assignments to
1224 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1226 (gcc::dump_manager::dump_finish): Likewise.
1227 * dumpfile.h (alt_dump_file): Delete decl.
1228 (dumps_are_enabled): New variable decl.
1229 (set_dump_file): New function decl.
1230 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1232 * tree-nested.c (lower_nested_functions): Replace assignments to
1233 "dump_file" with calls to set_dump_file.
1235 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1237 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1238 goto_locus of each outgoing edge of each basic block.
1240 2018-06-28 Richard Biener <rguenther@suse.de>
1242 * dwarf2out.c (decl_scope_table): Remove.
1243 (push_decl_scope): Likewise.
1244 (pop_decl_scope): Likewise.
1245 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1246 (gen_struct_or_union_type_die): Likewise.
1247 (gen_tagged_type_die): Likewise.
1248 (dwarf2out_init): Do not initialize decl_scope_table.
1249 (dwarf2out_c_finalize): Do not free it.
1251 2018-06-28 Richard Biener <rguenther@suse.de>
1253 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1254 deciding whether to not re-use a DIE.
1256 2018-06-28 Richard Biener <rguenther@suse.de>
1258 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1259 DW_AT_abstract_origin attribute.
1261 2018-06-28 Martin Liska <mliska@suse.cz>
1263 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1264 Use newly introduced constants.
1265 * tree-switch-conversion.h (struct jump_table_cluster):
1266 Define max_ratio_for_size and max_ratio_for_speed.
1268 2018-06-28 Martin Liska <mliska@suse.cz>
1270 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1271 Add new checking assert to catch invalid state.
1272 (jump_table_cluster::can_be_handled): Handle single case
1274 (jump_table_cluster::is_beneficial): Bail out for such case.
1275 (bit_test_cluster::find_bit_tests):
1276 Add new checking assert to catch invalid state.
1277 (bit_test_cluster::can_be_handled): Handle single case
1279 (bit_test_cluster::is_beneficial): Bail out for such case.
1280 (switch_decision_tree::analyze_switch_statement):
1283 2018-06-28 Martin Liska <mliska@suse.cz>
1285 * common.opt: Introduce -completion option.
1286 * gcc.c (driver_handle_option): Handle it.
1287 (driver::main): Print completions if completion
1289 * opt-suggestions.c (option_proposer::get_completions):
1291 (option_proposer::suggest_completion): Likewise.
1292 (option_proposer::find_param_completions): Likewise.
1293 (verify_autocompletions): Likewise.
1294 (test_completion_valid_options): Likewise.
1295 (test_completion_valid_params): Likewise.
1296 (in_completion_p): Likewise.
1297 (empty_completion_p): Likewise.
1298 (test_completion_partial_match): Likewise.
1299 (test_completion_garbage): Likewise.
1300 (opt_proposer_c_tests): Likewise.
1301 * opt-suggestions.h: Declare new functions.
1302 * opts.c (common_handle_option): Handle OPT__completion_.
1303 * selftest-run-tests.c (selftest::run_tests): Add
1304 opt_proposer_c_tests.
1305 * selftest.c (assert_str_startswith): New.
1306 * selftest.h (assert_str_startswith): Likewise.
1307 (opt_proposer_c_tests): New.
1308 (ASSERT_STR_STARTSWITH): Likewise.
1310 2018-06-28 Martin Liska <mliska@suse.cz>
1312 * Makefile.in: Add opt-suggestions.o.
1313 * gcc-main.c: Include opt-suggestions.h.
1314 * gcc.c (driver::driver): Likewise.
1315 (driver::~driver): Remove m_option_suggestions.
1316 (driver::build_option_suggestions): Moved to option_proposer.
1317 (driver::suggest_option): Likewise.
1318 (driver::handle_unrecognized_options): Use option_proposer.
1319 * gcc.h (class driver): Add new memver m_option_proposer.
1320 * opt-suggestions.c: New file.
1321 * opt-suggestions.h: New file.
1323 2018-06-28 Martin Liska <mliska@suse.cz>
1325 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1326 (auto_string_vec::~auto_string_vec): Likewise.
1328 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1330 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1331 prevent_decl_creation_for_types fields up and add reset_location field.
1332 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1333 statement if id->reset_location is true.
1334 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1335 id->reset_location is true.
1336 (copy_phis_for_bb): Force input_location on the arguments if
1337 id->reset_location is true.
1338 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1339 is set on the function to be inlined.
1341 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1343 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1345 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1347 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1348 registers for Pmode.
1349 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1350 hard registers for the clobbered pseudo.
1352 2018-06-27 Paul Koning <ni1d@arrl.net>
1354 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1355 mutually exclusive options.
1356 * config/pdp11/constraints.md (h): New constraint.
1357 (O): Update definition to match shift code generation.
1358 (D): New constraint.
1359 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1361 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1363 (output_jump): Change arguments.
1364 (pdp11_fixed_cc_regs): New function.
1365 (pdp11_cc_mode): Ditto.
1366 (pdp11_expand_shift): Ditto.
1367 (pdp11_assemble_shift): Ditto.
1368 (pdp11_small_shift): Ditto.
1369 (pdp11_branch_cost): Remove.
1370 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1372 (pdp11_register_move_cost): Update for CC registers.
1373 (pdp11_rtx_costs): Add case for LSHIFTRT.
1374 (pdp11_output_jump): Add CCNZ mode conditional branches.
1375 (notice_update_cc_on_set): Remove.
1376 (pdp11_cc_mode): New function.
1377 (simple_memory_operand): Correct pre/post decrement case.
1378 (no_side_effect_operand): New function.
1379 (pdp11_regno_reg_class): Add CC_REGS class.
1380 (pdp11_fixed_cc_regs): New function.
1381 (pdp11_small_shift): New function.
1382 (pdp11_expand_shift): New function to expand shift insns.
1383 (pdp11_assemble_shift): New function to output shifts.
1384 (pdp11_branch_cost): Remove.
1385 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1386 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1387 (WCHAR_TYPE): Ditto.
1388 (PTRDIFF_TYPE): Ditto.
1389 (ADJUST_INSN_LENGTH): New macro.
1390 (FIXED_REGISTERS): Add CC registers.
1391 (CALL_USED_REGISTERS): Ditto.
1393 (REG_CLASS_NAMES): Ditto.
1394 (REG_CLASS_CONTENTS): Ditto.
1395 (SELECT_CC_MODE): Use new function.
1396 (TARGET_FLAGS_REGNUM): New macro.
1397 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1398 (cc0_reg_rtx): Remove.
1399 (CC_STATUS_MDEP): Remove.
1400 (CC_STATUS_MDEFP_INIT): Remove.
1401 (CC_IN_FPU): Remove.
1402 (NOTICE_UPDATE_CC): Remove.
1403 (REGISTER_NAMES): Add CC registers.
1404 (BRANCH_COST): Change to constant 1.
1405 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1407 * config/pdp11/pdp11.opt (mbcopy): Remove.
1408 (mbcopy-builtin): Remove.
1409 (mbranch-cheap): Remove.
1410 (mbranch-expensive): Remove.
1411 * config/pdp11/predicates.md (expand_shift_operand): Update to
1412 match shift code generation.
1413 (ccnz_operator): New predicate.
1414 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1415 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1416 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1418 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1419 description of O constraint.
1421 2018-06-27 Jeff Law <law@redhat.com>
1422 Austin Law <austinklaw@gmail.com>
1424 * config/v850/v850.md (addsi3_set_flags): New pattern.
1425 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1426 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1427 (zero_extendhisi2_v850_set_flags): Likewise.
1428 (zero_extendqisi2_v850_set_flags): Likewise.
1429 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1430 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1431 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1433 * config/v850/v850-protos.h (notice_update_cc): Remove.
1434 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1435 (v850_print_operand): Handle 'D' and "d".
1436 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1437 Add handling of arithmetic/logical operations compared against zero.
1438 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1439 Do not look at v850_compare_op, instead get mode from last argument.
1440 (v850_gen_compare): Remove
1441 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1442 after reload for prologue insns.
1443 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1445 (construct_save_jarl): Likewise.
1446 (TARGET_FLAGS_REGNUM): Define.
1447 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1448 (NOTICE_UPDATE_CC): Remove.
1449 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1450 than cc0. Conditionalize on reload_completed.
1451 (cmpsi_insn, setfcc_insn): Likewise.
1452 (tst1 splitter): Turn into define_and_split which sets the flags
1454 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1455 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1456 (cstoresf4, cstoredf4): Clobber the flags.
1457 (cmpsi, cmpsf, cmpdf): Remove expanders.
1458 (setf_insn): Remove pattern.
1459 (addsi3): Turn into define_and_split which clobbers the flags after
1460 reload and a suitable pattern (addsi3_clobber_flags) for use after
1462 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1463 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1464 (ashrsi3, ashrsi3_v850e2): Likewise.
1465 (bins): Clobber the flags.
1466 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1467 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1468 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1469 (call_value_internal_short, call_value_internal_long): Likewise.
1470 (callt_save_interrupt, callt_return_interrupt): Likewise.
1471 (save_interrupt, return_interrupt): Likewise.
1472 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1473 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1474 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1475 (All FP comparisons): Only allow after reload has completed.
1477 (divh, divhu): Tweak output template.
1478 (branch_z_normal, branch_z_invert): Remove
1479 (branch_nz_normal, branch_nz_invert): Likewise.
1480 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1482 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1483 * config/v850/v850.c (notice_update_cc): Remove.
1484 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1485 (CC_NO_CARRY): Likewise.
1486 (NOTICE_UPDATE_CC): Define to nothing.
1487 * config/v850/v850.md: Remove block comment on cc0 handling
1488 Remove "cc" attribute from all patterns. Remove cc_status handling
1489 from all patterns. Minor formatting fixes.
1491 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1493 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1494 (cortex-a76.cortex-a55): Likewise.
1495 * config/aarch64/aarch64-tune.md: Regenerate.
1496 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1497 cortex-a76.cortex-a55.
1499 2018-06-27 Jeff Law <law@redhat.com>
1501 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1502 (MULTILIB_DIRNAMES): Similarly.
1504 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1506 * gimple.h (gimple_return_retbnd): Delete.
1507 (gimple_return_set_retbnd): Likewise.
1508 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1509 gimple_return_set_retbnd.
1510 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1511 gimple_return_retbnd and adjust.
1512 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1513 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1514 Explicitly return NULL in a couple more cases. Move assertion
1515 on debug statements and remove unreachable code.
1516 (reset_debug_binding): Do not test id->retbnd.
1517 (expand_call_inline): Do not set it.
1519 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1521 * configure.ac: Add --disable-gcov option.
1522 * configure: Regenerate.
1523 * Makefile.in: Honour @enable_gcov@.
1524 * doc/install.texi: Document --disable-gcov.
1526 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1528 * config/arm/arm-cpus.in (cortex-a76): New entry.
1529 (cortex-a76.cortex-a55): Likewise.
1530 * config/arm/arm-tables.opt: Regenerate.
1531 * config/arm/arm-tune.md: Likewise.
1532 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1533 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1534 cortex-a76.cortex-a55.
1536 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1539 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1541 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1543 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1545 (EPILOGUE_USES): Likewise.
1547 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1549 * tree-inline.c (remap_location): New function extracted from...
1550 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1551 (copy_phis_for_bb): ...here. Call remap_location.
1552 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1554 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1556 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1557 unaligned vsx for 16B memset.
1559 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1562 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1563 ieee128_float_type_node to long_double_type_node unless
1564 TARGET_LONG_DOUBLE_128 is set.
1566 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1568 * cfgloop.c (get_loop_location): Convert return type from
1569 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1570 by implicit construction from rtx_insn *, and using
1571 dump_user_location_t::from_function_decl for the fallback case.
1572 * cfgloop.h (get_loop_location): Convert return type from
1573 location_t to dump_user_location_t.
1574 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1575 dump_printf_loc to pass in a dump_location_t rather than a
1576 location_t, via the gimple stmt.
1577 * coverage.c (get_coverage_counts): Update calls to
1578 dump_printf_loc to pass in dump_location_t rather than a
1580 * doc/optinfo.texi (Dump types): Convert example of
1581 dump_printf_loc from taking "locus" to taking "insn". Update
1582 description of the "_loc" calls to cover dump_location_t.
1583 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1585 (dump_user_location_t::dump_user_location_t): New constructors,
1586 from gimple *stmt and rtx_insn *.
1587 (dump_user_location_t::from_function_decl): New function.
1588 (dump_loc): Make static.
1589 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1590 const dump_location_t &.
1591 (dump_generic_expr_loc): Delete.
1592 (dump_printf_loc): Convert param "loc" from location_t to
1593 const dump_location_t &.
1594 (selftest::test_impl_location): New function.
1595 (selftest::dumpfile_c_tests): New function.
1596 * dumpfile.h: Include "profile-count.h".
1597 (class dump_user_location_t): New class.
1598 (struct dump_impl_location_t): New struct.
1599 (class dump_location_t): New class.
1600 (dump_printf_loc): Convert 2nd param from source_location to
1601 const dump_location_t &.
1602 (dump_generic_expr_loc): Delete.
1603 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1604 const dump_location_t &.
1605 * gimple-fold.c (fold_gimple_assign): Update call to
1606 dump_printf_loc to pass in a dump_location_t rather than a
1607 location_t, via the gimple stmt.
1608 (gimple_fold_call): Likewise.
1609 * gimple-loop-interchange.cc
1610 (loop_cand::analyze_iloop_reduction_var): Update for change to
1611 check_reduction_path.
1612 (tree_loop_interchange::interchange): Update for change to
1614 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1615 change in return-type of find_loop_location.
1616 (graphite_regenerate_ast_isl): Likewise.
1617 * graphite-optimize-isl.c (optimize_isl): Likewise.
1618 * graphite.c (graphite_transform_loops): Likewise.
1619 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1620 pass in a dump_location_t rather than a location_t, via the
1622 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1623 * ipa.c (walk_polymorphic_call_targets): Likewise.
1624 * loop-unroll.c (report_unroll): Convert "locus" param from
1625 location_t to dump_location_t.
1626 (decide_unrolling): Update for change to get_loop_location's
1628 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1629 location_t to dump_user_location_t.
1630 (grid_find_single_omp_among_assignments_1): Updates calls to
1631 dump_printf_loc to pass in a dump_location_t rather than a
1632 location_t, via the gimple stmt.
1633 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1634 location_t to dump_location_t. Updates calls to dump_printf_loc
1635 to pass in a dump_location_t rather than a location_t, via the
1637 (grid_inner_loop_gridifiable_p): Likewise.
1638 (grid_dist_follows_simple_pattern): Likewise.
1639 (grid_gfor_follows_tiling_pattern): Likewise.
1640 (grid_target_follows_gridifiable_pattern): Likewise.
1641 (grid_attempt_target_gridification): Convert initialization
1642 of local "grid" from memset to zero-initialization; FIXME: does
1643 this require C++11? Update call to dump_printf_loc to pass in a
1644 optinfo_location rather than a location_t, via the gimple stmt.
1645 * profile.c (read_profile_edge_counts): Updates call to
1646 dump_printf_loc to pass in a dump_location_t rather than a
1648 (compute_branch_probabilities): Likewise.
1649 * selftest-run-tests.c (selftest::run_tests): Call
1651 * selftest.h (dumpfile_c_tests): New decl.
1652 * tree-loop-distribution.c (pass_loop_distribution::execute):
1653 Update for change in return type of find_loop_location.
1654 * tree-parloops.c (parallelize_loops): Likewise.
1655 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1656 "locus" from location_t to dump_user_location_t.
1657 (canonicalize_loop_induction_variables): Likewise.
1658 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1659 for change in return type of find_loop_location.
1660 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1661 to dump_printf_loc to pass in a dump_location_t rather than a
1662 location_t, via the stmt.
1663 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1665 * tree-vect-loop-manip.c (find_loop_location): Convert return
1666 type from source_location to dump_user_location_t.
1667 (vect_do_peeling): Update for above change.
1668 (vect_loop_versioning): Update for change in type of
1670 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1671 from location_t to dump_user_location_t.
1672 (vect_estimate_min_profitable_iters): Update for change in type
1674 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1675 location_t to dump_location_t.
1676 (vect_slp_bb): Update for change in type of vect_location.
1677 * tree-vectorizer.c (vect_location): Convert from source_location
1678 to dump_user_location_t.
1679 (try_vectorize_loop_1): Update for change in vect_location's type.
1680 (vectorize_loops): Likewise.
1681 (increase_alignment): Likewise.
1682 * tree-vectorizer.h (vect_location): Convert from source_location
1683 to dump_user_location_t.
1684 (find_loop_location): Convert return type from source_location to
1685 dump_user_location_t.
1686 (check_reduction_path): Convert 1st param from location_t to
1687 dump_user_location_t.
1688 * value-prof.c (check_counter): Update call to dump_printf_loc to
1689 pass in a dump_user_location_t rather than a location_t; update
1690 call to error_at for change in type of "locus".
1691 (check_ic_target): Update call to dump_printf_loc to
1692 pass in a dump_user_location_t rather than a location_t, via the
1695 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1697 * config/s390/s390.h (enum processor_flags): Do not use
1698 default tune parameter when -march was specified.
1700 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1703 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1704 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1707 2018-06-26 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/86287
1711 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1712 (vect_analyze_loop): Initialize n_stmts.
1714 2018-06-26 Richard Biener <rguenther@suse.de>
1717 * fold-const.c (fold_convertible_p): Pointer extension
1720 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1723 * dwarf2out.c (loc_list_has_views): Adjust comments.
1724 (dw_loc_list): Split single cross-partition range with
1727 2018-06-25 Jeff Law <law@redhat.com>
1729 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1730 on -mbig-switch by default.
1732 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1734 (const_float_0_operand): Remove unused predicate.
1735 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1736 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1737 (recipsf2): New expander. Original pattern now called
1739 (recipdf2, recipdf2_insn): Similarly.
1740 (rsqrtsf2, rsqrtsf2_insn): Similarly
1741 (rsqrtdf2, rsqrtdf2_insn): Similarly
1743 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1745 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1746 Simplify logic for FreeBSD (twice).
1748 2018-06-25 Martin Sebor <msebor@redhat.com>
1750 PR tree-optimization/86204
1751 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1752 a strnlen result if it's less than the length of the string.
1754 2018-06-25 Martin Sebor <msebor@redhat.com>
1756 PR tree-optimization/85700
1757 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1758 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1759 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1761 2018-06-25 Martin Sebor <msebor@redhat.com>
1763 * doc/extend.texi (Zero-length arrays): Update and clarify.
1765 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1767 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1768 added IEEE/IBM long double multilib support on PowerPC little
1769 endian Linux systems.
1770 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1771 (MULTILIB_DEFAULTS): Likewise.
1772 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1774 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1775 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1776 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1778 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1781 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1782 (REORDER_45): Likewise.
1784 2018-06-25 Jeff Law <law@redhat.com>
1786 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1787 dividend to 32 bits. Adjust length.
1788 (udivmodhi4): Cleanup output template. Fix length.
1790 2018-06-25 Carl Love <cel@us.ibm.com>
1792 * config/rs6000/vsx.md: Change word selector to prefered location.
1794 2018-06-25 Richard Biener <rguenther@suse.de>
1796 PR tree-optimization/86304
1797 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1798 epilogue-if-converted loops as well.
1800 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1802 * lto-section-out.c (lto_begin_section): Do not print section
1803 name for noaddr and unnumbered dumps.
1805 2018-06-25 Richard Biener <rguenther@suse.de>
1807 * tree-vectorizer.h (struct vec_info_shared): New structure
1808 with parts split out from struct vec_info and loop_nest from
1809 struct _loop_vec_info.
1810 (struct vec_info): Adjust accordingly.
1811 (struct _loop_vec_info): Likewise.
1812 (LOOP_VINFO_LOOP_NEST): Adjust.
1813 (LOOP_VINFO_DATAREFS): Likewise.
1814 (LOOP_VINFO_DDRS): Likewise.
1815 (struct _bb_vec_info): Likewise.
1816 (BB_VINFO_DATAREFS): Likewise.
1817 (BB_VINFO_DDRS): Likewise.
1818 (struct _stmt_vec_info): Add dr_aux member.
1819 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1820 (DR_MISALIGNMENT_UNINITIALIZED): New.
1821 (set_dr_misalignment): Adjust.
1822 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1823 (vect_analyze_loop): Adjust prototype.
1824 (vect_analyze_loop_form): Likewise.
1825 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1826 Compute dependences lazily.
1827 (vect_record_base_alignments): Use shared datarefs/ddrs.
1828 (vect_verify_datarefs_alignment): Likewise.
1829 (vect_analyze_data_refs_alignment): Likewise.
1830 (vect_analyze_data_ref_accesses): Likewise.
1831 (vect_analyze_data_refs): Likewise.
1832 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1833 constructor parameter for shared part.
1834 (vect_analyze_loop_form): Pass in shared part and adjust.
1835 (vect_analyze_loop_2): Pass in storage for the number of
1836 stmts. Move loop nest finding to the caller. Compute
1838 (vect_analyze_loop): Pass in shared part.
1839 (vect_transform_loop): Verify shared datarefs are unchanged.
1840 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1841 constructor parameter for shared part.
1842 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1843 (vect_slp_bb): Verify shared datarefs are unchanged before
1845 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1847 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1848 DR_MISALIGNMENT_UNINITIALIZED.
1849 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1850 parameter for shared part.
1851 (vec_info::~vec_info): Adjust.
1852 (vec_info_shared::vec_info_shared): New.
1853 (vec_info_shared::~vec_info_shared): Likewise.
1854 (vec_info_shared::save_datarefs): Likewise.
1855 (vec_info_shared::check_datarefs): Likewise.
1856 (try_vectorize_loop_1): Construct shared part live for analyses
1857 of a single loop for multiple vector sizes.
1858 * tree-parloops.c (gather_scalar_reductions): Adjust.
1860 2018-06-25 Richard Biener <rguenther@suse.de>
1862 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1863 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1864 (vect_analyze_data_refs): Remove similar code from here and
1865 simplify accordingly.
1867 2018-06-25 Richard Biener <rguenther@suse.de>
1869 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1870 for reverse storage order accesses rather than asserting
1871 they cannot happen here.
1873 2018-06-25 Tom de Vries <tdevries@suse.de>
1876 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1877 Use data16 instead of .byte for insn prefix.
1879 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1882 * parser.c (make_char_string_pack): Pass this literal chars
1883 through cpp_interpret_string.
1884 (cp_parser_userdef_numeric_literal): Check the result of
1885 make_char_string_pack.
1887 2018-06-24 Maya Rashish <coypu@sdf.org>
1889 * ginclude/stddef.h: Simplify conditions around avoiding
1890 re-definition of __size_t.
1892 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1894 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1895 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1897 2018-06-22 Maya Rashish <coypu@sdf.org>
1899 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1900 made to OpenBSD and NetBSD through the years.
1902 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1904 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1905 behavior of vec_pack (vector double, vector double) to match
1906 behavior of vec_float2 (vector double, vector double).
1908 2018-06-22 Olivier Hainque <hainque@adacore.com>
1910 * gimplify.c (gimplify_function_tree): Prevent creation
1911 of a trampoline for the address of the current function
1912 passed to entry/exit instrumentation hooks.
1914 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1917 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1920 2018-06-22 Martin Liska <mliska@suse.cz>
1922 PR tree-optimization/86263
1923 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1924 Bail out if is_enabled is false.
1925 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1927 (jump_table_cluster::is_enabled): New function.
1929 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1931 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1932 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1933 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1934 (lto_input_ts_binfo_tree_pointers): Likewise.
1935 * tree-streamer-out.c (streamer_write_tree_bitfields,
1936 write_ts_binfo_tree_pointers): Likewise.
1937 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1939 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1941 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1943 2018-06-22 Martin Liska <mliska@suse.cz>
1945 * symbol-summary.h (get): Make it pure and inline move
1946 functionality from ::get function.
1947 (get): Remove and inline into ::get and ::get_create.
1948 (get_create): Move code from ::get function.
1950 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1953 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1954 -x assembler-with-cpp.
1956 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1958 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1959 _FILE_OFFSET_BITS=64 for C++.
1961 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1963 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1964 conversion insn that shows up when pr85657-3.c is compiled using
1965 IEEE 128-bit long double.
1966 (neg<mode>2_internal): Use the correct mode to check whether the
1967 mode is IBM extended.
1968 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1969 multiply and divide external functions from being created more
1972 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1974 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1976 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1977 the edge can be forwarded.
1978 (cfg_layout_merge_blocks): Likewise.
1980 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1982 * except.c (finish_eh_generation): Commit edge insertions only after
1983 the EH edges have been redirected from post-landing to landing pads.
1985 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1987 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1988 create_tmp_var_for to create the FRAME decl.
1989 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1991 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1993 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1994 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1995 debug statement when resetting its value.
1996 (expand_call_inline): Copy the locus of the call onto the assignment
1997 of the return value, if any. Use local variable in more cases.
1999 2018-06-21 Martin Liska <mliska@suse.cz>
2001 * ipa-pure-const.c (propagate_nothrow): Use
2002 funct_state_summaries->get.
2003 (dump_malloc_lattice): Likewise.
2004 (propagate_malloc): Likewise.
2006 2018-06-21 Richard Biener <rguenther@suse.de>
2008 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2009 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2010 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2012 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2013 BLOCK_ABSTRACT_ORIGIN unconditionally.
2015 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2017 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2019 * ipa-reference.c (ipa_reference_c_finalize): Delete
2020 ipa_ref_opt_sum_summaries and set it to NULL.
2022 2018-06-21 Tom de Vries <tdevries@suse.de>
2024 PR tree-optimization/85859
2025 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2026 test with comment from bb_no_side_effects_p.
2028 2018-06-21 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/86232
2031 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2032 max for constant niter.
2034 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2036 * config/aarch64/aarch64-simd.md
2037 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2039 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2041 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2042 Make opernads of the unspec commutative.
2044 2018-06-21 Richard Biener <rguenther@suse.de>
2046 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2047 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2048 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2049 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2051 (vect_record_base_alignments): Do not record base alignment
2052 for gathers or scatters.
2053 (vect_compute_data_ref_alignment): Drop return value that is always
2054 true. Bail out early for gathers or scatters.
2055 (vect_enhance_data_refs_alignment): Bail out early for gathers
2057 (vect_find_same_alignment_drs): Likewise.
2058 (vect_analyze_data_refs_alignment): Remove dead code.
2059 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2060 (vect_analyze_data_refs): For possible gathers or scatters do
2061 not create an alternate DR, just check their possible validity
2062 and mark them. Adjust DECL_NONALIASED handling to not rely
2064 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2065 update inits of gathers or scatters.
2066 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2067 Also copy gather/scatter flag to pattern vinfo.
2069 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2071 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2072 behavior of vec_packsu (vector unsigned long long, vector unsigned
2073 long long) to match behavior of vec_packs with same signature.
2075 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2076 Thomas Schwinge <thomas@codesourcery.com>
2077 Cesar Philippidis <cesar@codesourcery.com>
2079 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2080 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2081 (gimplify_adjust_omp_clauses): Likewise.
2082 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2083 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2084 (gimplify_omp_target_update): Update handling of acc update and
2086 * omp-low.c (install_var_field): Remove unused parameter
2087 base_pointers_restrict.
2088 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2089 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2091 (omp_target_base_pointers_restrict_p): Delete.
2092 (scan_omp_target): Update call to scan_sharing_clauses.
2093 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2095 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2096 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2097 (convert_local_omp_clauses): Likewise.
2098 * tree-pretty-print.c (dump_omp_clause): Likewise.
2099 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2101 (omp_clause_code_name): Likewise.
2103 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2106 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2109 PR tree-optimization/86231
2110 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2111 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2113 2018-06-20 Tom de Vries <tdevries@suse.de>
2115 PR tree-optimization/86097
2116 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2117 iv type if signedness of iv type is not the same as that of *nit.
2119 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2121 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2122 EDGE_EH edges, verify they are all EDGE_EH.
2124 2018-06-20 Maya Rashish <coypu@sdf.org>
2126 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2128 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2130 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2131 * config/aarch64/aarch64.c (xgene1_tunings): Add
2132 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2133 (aarch64_mode_valid_for_sched_fusion_p):
2134 Allow 16-byte modes.
2135 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2136 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2138 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2140 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2141 * config/aarch64/iterators.md (VQ2): New mode iterator.
2143 2018-06-20 Martin Liska <mliska@suse.cz>
2145 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2146 Change default ratio from 10 to 8.
2148 2018-06-20 Martin Liska <mliska@suse.cz>
2150 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2152 (bit_test_cluster::find_bit_tests): Likewise.
2153 (switch_decision_tree::analyze_switch_statement): Find clusters.
2154 * tree-switch-conversion.h (struct jump_table_cluster): Document
2157 2018-06-20 Martin Liska <mliska@suse.cz>
2159 * tree-switch-conversion.c (switch_conversion::collect):
2160 Record m_uniq property.
2161 (switch_conversion::expand): Bail out for special conditions.
2162 (group_cluster::~group_cluster): New.
2163 (group_cluster::group_cluster): Likewise.
2164 (group_cluster::dump): Likewise.
2165 (jump_table_cluster::emit): New.
2166 (switch_decision_tree::fix_phi_operands_for_edges): New.
2167 (struct case_node): Remove struct.
2168 (jump_table_cluster::can_be_handled): New.
2169 (case_values_threshold): Moved to header.
2170 (reset_out_edges_aux): Likewise.
2171 (jump_table_cluster::is_beneficial): New.
2172 (bit_test_cluster::can_be_handled): Likewise.
2173 (add_case_node): Remove.
2174 (bit_test_cluster::is_beneficial): New.
2175 (case_bit_test::cmp): New.
2176 (bit_test_cluster::emit): New.
2177 (expand_switch_as_decision_tree_p): Remove.
2178 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2179 (fix_phi_operands_for_edge): Likewise.
2180 (switch_decision_tree::analyze_switch_statement): New.
2181 (compute_cases_per_edge): Move ...
2182 (switch_decision_tree::compute_cases_per_edge): ... here.
2183 (try_switch_expansion): Likewise.
2184 (switch_decision_tree::try_switch_expansion): Likewise.
2185 (record_phi_operand_mapping): Likewise.
2186 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2187 (emit_case_decision_tree): Likewise.
2188 (switch_decision_tree::emit): Likewise.
2189 (balance_case_nodes): Likewise.
2190 (switch_decision_tree::balance_case_nodes): Likewise.
2191 (dump_case_nodes): Likewise.
2192 (switch_decision_tree::dump_case_nodes): Likewise.
2193 (emit_jump): Likewise.
2194 (switch_decision_tree::emit_jump): Likewise.
2195 (emit_cmp_and_jump_insns): Likewise.
2196 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2197 (emit_case_nodes): Likewise.
2198 (switch_decision_tree::emit_case_nodes): Likewise.
2199 (conditional_probability): Remove.
2200 * tree-switch-conversion.h (enum cluster_type): New.
2202 (struct cluster): Likewise.
2203 (cluster::cluster): Likewise.
2204 (struct simple_cluster): Likewise.
2205 (simple_cluster::simple_cluster): Likewise.
2206 (struct group_cluster): Likewise.
2207 (struct jump_table_cluster): Likewise.
2208 (struct bit_test_cluster): Likewise.
2209 (struct min_cluster_item): Likewise.
2210 (struct case_tree_node): Likewise.
2211 (case_tree_node::case_tree_node): Likewise.
2212 (jump_table_cluster::case_values_threshold): Likewise.
2213 (struct case_bit_test): Likewise.
2214 (struct switch_decision_tree): Likewise.
2215 (struct switch_conversion): Likewise.
2216 (switch_decision_tree::reset_out_edges_aux): Likewise.
2218 2018-06-20 Martin Liska <mliska@suse.cz>
2220 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2221 (hoist_edge_and_branch_if_true): Likewise.
2222 (expand_switch_using_bit_tests_p): Likewise.
2223 (struct case_bit_test): Likewise.
2224 (case_bit_test_cmp): Likewise.
2225 (emit_case_bit_tests): Likewise.
2226 (switch_conversion::switch_conversion): New class.
2227 (struct switch_conv_info): Remove old struct.
2228 (collect_switch_conv_info): More to ...
2229 (switch_conversion::collect): ... this.
2230 (check_range): Likewise.
2231 (switch_conversion::check_range): Likewise.
2232 (check_all_empty_except_final): Likewise.
2233 (switch_conversion::check_all_empty_except_final): Likewise.
2234 (check_final_bb): Likewise.
2235 (switch_conversion::check_final_bb): Likewise.
2236 (create_temp_arrays): Likewise.
2237 (switch_conversion::create_temp_arrays): Likewise.
2238 (free_temp_arrays): Likewise.
2239 (gather_default_values): Likewise.
2240 (switch_conversion::gather_default_values): Likewise.
2241 (build_constructors): Likewise.
2242 (switch_conversion::build_constructors): Likewise.
2243 (constructor_contains_same_values_p): Likewise.
2244 (switch_conversion::contains_same_values_p): Likewise.
2245 (array_value_type): Likewise.
2246 (switch_conversion::array_value_type): Likewise.
2247 (build_one_array): Likewise.
2248 (switch_conversion::build_one_array): Likewise.
2249 (build_arrays): Likewise.
2250 (switch_conversion::build_arrays): Likewise.
2251 (gen_def_assigns): Likewise.
2252 (switch_conversion::gen_def_assigns): Likewise.
2253 (prune_bbs): Likewise.
2254 (switch_conversion::prune_bbs): Likewise.
2255 (fix_phi_nodes): Likewise.
2256 (switch_conversion::fix_phi_nodes): Likewise.
2257 (gen_inbound_check): Likewise.
2258 (switch_conversion::gen_inbound_check): Likewise.
2259 (process_switch): Use the newly created class.
2260 (switch_conversion::expand): New.
2261 (switch_conversion::~switch_conversion): New.
2262 * tree-switch-conversion.h: New file.
2264 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2266 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2267 tree-vect-patterns.c.
2268 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2269 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2270 (vect_recog_sad_pattern): Likewise.
2271 (vect_recog_widen_sum_pattern): Likewise.
2272 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2273 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2274 (vect_recog_rotate_pattern): Likewise.
2275 (vect_recog_mult_pattern): Likewise.
2276 (vect_recog_vector_vector_shift_pattern): Likewise.
2277 (vect_recog_divmod_pattern): Likewise.
2278 (vect_recog_mixed_size_cond_pattern): Likewise.
2279 (vect_recog_bool_pattern): Likewise.
2280 (vect_recog_mask_conversion_pattern): Likewise.
2281 (vect_try_gather_scatter_pattern): Likewise.
2282 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2283 (vect_recog_over_widening_pattern): Likewise.
2284 (vect_recog_gather_scatter_pattern): Likewise.
2285 (vect_recog_func_ptr): Move from tree-vectorizer.h
2286 (vect_vect_recog_func_ptrs): Move further down the file.
2287 (vect_recog_func): Likewise. Remove the third argument.
2288 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2289 (vect_pattern_recog_1): Expect the pattern function to do any
2290 necessary target tests. Also expect it to provide a vector type.
2291 Remove the type_in handling.
2293 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2295 * tree-vect-patterns.c (vect_pattern_detected): New function.
2296 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2297 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2298 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2299 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2300 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2301 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2302 (vect_recog_mask_conversion_pattern)
2303 (vect_try_gather_scatter_pattern): Likewise.
2305 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2307 * tree-vect-patterns.c (vect_get_internal_def): New function.
2308 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2309 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2310 (search_type_for_mask_1): Use it.
2312 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2314 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2315 redundant WIDEN_SUM_EXPR handling.
2316 (vect_recog_sad_pattern): Likewise.
2318 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2320 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2321 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2322 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2323 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2324 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2326 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2328 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2329 use the stmt_vec_info of the original bb statement for the
2330 new zero assignment, even if the call is part of a pattern.
2332 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2334 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2335 that the sequence is attached to the original statement rather
2336 than the pattern statement.
2337 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2338 PATTERN_DEF_SEQ from the original statement rather than
2339 the main pattern statement.
2340 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2341 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2342 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2344 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2346 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2347 definition statements before the early exit for statements that aren't
2349 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2351 (vect_transform_loop): ...here. Process pattern definition
2352 statements without first checking whether the main pattern
2353 statement is live or relevant.
2355 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2357 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2358 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2360 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2362 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2363 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2364 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2365 (expand_block_compare): Change select_block_compare_mode call.
2366 (expand_strncmp_align_check): Use new functions, fix comment.
2367 (emit_final_str_compare_gpr): New function.
2368 (expand_strn_compare): Refactor and clean up code.
2369 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2371 2018-06-19 Tony Reix <tony.reix@atos.com>
2372 Damien Bergamini <damien.bergamini@atos.com>
2373 David Edelsohn <dje.gcc@gmail.com>
2375 * collect2.c (static_obj): New variable.
2376 (static_libs): New variable.
2377 (is_in_list): Uncomment declaration.
2378 (main): Track AIX libraries linked statically.
2379 (is_in_list): Uncomment definition.
2380 (scan_prog_file): Don't add AIX shared libraries initializer
2381 to constructor list if linking statically.
2383 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2385 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2387 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2389 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2391 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2394 2018-06-19 Martin Liska <mliska@suse.cz>
2396 * config/i386/i386.c (ix86_can_inline_p): Do not use
2397 ipa_fn_summaries::get_create.
2398 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2400 (devirtualization_time_bonus): Likewise.
2401 (ipcp_propagate_stage): Likewise.
2402 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2403 (edge_set_predicate): Likewise.
2404 (evaluate_conditions_for_known_args): Likewise.
2405 (evaluate_properties_for_edge): Likewise.
2406 (ipa_call_summary::reset): Tranform to ...
2407 (ipa_call_summary::~ipa_call_summary): ... this.
2408 (ipa_fn_summary::reset): Transform to ...
2409 (ipa_fn_summary::~ipa_fn_summary): ... this.
2410 (ipa_fn_summary_t::remove): Rename to ...
2411 (ipa_fn_summary_t::remove_callees): ... this.
2412 (ipa_fn_summary_t::duplicate): Use placement new
2413 instead of memory copy.
2414 (ipa_call_summary_t::duplicate): Likewise.
2415 (ipa_call_summary_t::remove): Remove.
2416 (dump_ipa_call_summary): Change get_create to get.
2417 (ipa_dump_fn_summary): Dump only when summary exists.
2418 (analyze_function_body): Use symbol_summary::get instead
2420 (compute_fn_summary): Likewise.
2421 (estimate_edge_devirt_benefit): Likewise.
2422 (estimate_edge_size_and_time): Likewise.
2423 (inline_update_callee_summaries): Likewise.
2424 (remap_edge_change_prob): Likewise.
2425 (remap_edge_summaries): Likewise.
2426 (ipa_merge_fn_summary_after_inlining): Likewise.
2427 (write_ipa_call_summary): Likewise.
2428 (ipa_fn_summary_write): Likewise.
2429 (ipa_free_fn_summary): Likewise.
2430 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2431 (struct ipa_call_summary): Likewise.
2432 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2434 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2435 (estimate_size_after_inlining): Likewise.
2436 (estimate_growth): Likewise.
2437 (growth_likely_positive): Likewise.
2438 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2439 (inline_call): Likewise.
2440 * ipa-inline.c (caller_growth_limits): Likewise.
2441 (can_inline_edge_p): Likewise.
2442 (can_inline_edge_by_limits_p): Likewise.
2443 (compute_uninlined_call_time): Likewise.
2444 (compute_inlined_call_time): Likewise.
2445 (want_inline_small_function_p): Likewise.
2446 (edge_badness): Likewise.
2447 (update_caller_keys): Likewise.
2448 (update_callee_keys): Likewise.
2449 (inline_small_functions): Likewise.
2450 (inline_to_all_callers_1): Likewise.
2451 (dump_overall_stats): Likewise.
2452 (early_inline_small_functions): Likewise.
2453 (early_inliner): Likewise.
2454 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2455 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2456 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2457 * ipa-split.c (execute_split_functions): Likewise.
2458 * symbol-summary.h: Likewise.
2459 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2461 2018-06-19 Richard Biener <rguenther@suse.de>
2463 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2464 (vectorize_loops): ... here. Fix dbgcnt handling.
2465 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2467 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2470 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2471 ieee128 argument takes up only one (vector) register, not two (floating
2474 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2476 * gimplify.c (gimplify_init_constructor): Really never clear for an
2477 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2479 2018-06-19 Richard Biener <rguenther@suse.de>
2481 PR tree-optimization/86179
2482 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2483 after failed recognition.
2485 2018-06-18 Martin Sebor <msebor@redhat.com>
2488 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2489 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2490 Handle integer subtraction.
2491 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2492 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2494 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2496 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2497 param from rtx to rtx_insn *.
2498 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2500 (frv_ifcvt_modify_insn): Likwise.
2501 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2502 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2503 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2504 as_a <rtx_insn *> cast to local "unprotected_region" once
2505 it's been established that it's not NULL or pc_rtx.
2506 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2507 param "sethi" from rtx to rtx_insn *.
2508 (nds32_group_float_insns): Likewise for param "insn".
2509 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2511 (vax_output_int_subtract): Likewise.
2512 * config/vax/vax.c (vax_output_int_add): Likewise for param
2514 (vax_output_int_subtract): Likewise.
2515 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2516 (emit_pattern_after): Likewise for param "after".
2517 (emit_insn_after): Likewise.
2518 (emit_jump_insn_after): Likewise.
2519 (emit_call_insn_after): Likewise.
2520 (emit_debug_insn_after): Likewise.
2521 (emit_pattern_before): Likewise for param "before".
2522 (emit_insn_before): Likewise.
2523 (emit_jump_insn_before): Likewise.
2524 * final.c (get_insn_template): Likewise for param "insn", removing
2526 * output.h (get_insn_template): Likewise for 2nd param.
2527 * rtl.h (emit_insn_before): Likewise.
2528 (emit_jump_insn_before): Likewise.
2529 (emit_debug_insn_before_noloc): Likewise.
2530 (emit_insn_after): Likewise.
2531 (emit_jump_insn_after): Likewise.
2532 (emit_call_insn_after): Likewise.
2533 (emit_debug_insn_after): Likewise.
2534 (set_insn_deleted): Likewise for param.
2536 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2539 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2540 floating point modes, so that IFmode is numerically greater than
2541 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2542 to declare the ordering. This prevents IFmode from being
2543 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2544 machine. Include rs6000-modes.h to share the fractional values
2545 between genmodes* and the rest of the compiler.
2549 * config/rs6000/rs6000-modes.h: New file.
2550 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2551 meaning of rs6000_long_double_size so that 126..128 selects an
2552 appropriate 128-bit floating point type.
2553 (rs6000_option_override_internal): Likewise.
2554 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2555 (TARGET_LONG_DOUBLE_128): Change the meaning of
2556 rs6000_long_double_size so that 126..128 selects an appropriate
2557 128-bit floating point type.
2558 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2559 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2560 source and destination to match the standard usage.
2561 (truncifkf2): Likewise.
2562 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2563 ISA 2.07 to use an explicit clobber, instead of passing in a
2565 (copysign<mode>3_soft): Likewise.
2567 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2569 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2570 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2571 (vect_slp_analyze_instance_dependence): Likewise.
2572 (vect_enhance_data_refs_alignment): Likewise.
2573 (vect_analyze_data_refs_alignment): Likewise.
2574 (vect_slp_analyze_and_verify_instance_alignment
2575 (vect_analyze_data_ref_accesses): Likewise.
2576 (vect_prune_runtime_alias_test_list): Likewise.
2577 (vect_analyze_data_refs): Likewise.
2578 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2579 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2580 (vect_analyze_scalar_cycles_1): Likewise.
2581 (vect_get_loop_niters): Likewise.
2582 (vect_analyze_loop_form_1): Likewise.
2583 (vect_update_vf_for_slp): Likewise.
2584 (vect_analyze_loop_operations): Likewise.
2585 (vect_analyze_loop): Likewise.
2586 (vectorizable_induction): Likewise.
2587 (vect_transform_loop): Likewise.
2588 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2589 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2590 (vect_make_slp_decision): Likewise.
2591 (vect_detect_hybrid_slp): Likewise.
2592 (vect_slp_analyze_operations): Likewise.
2593 (vect_slp_bb): Likewise.
2594 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2595 (vectorizable_bswap): Likewise.
2596 (vectorizable_call): Likewise.
2597 (vectorizable_simd_clone_call): Likewise.
2598 (vectorizable_conversion): Likewise.
2599 (vectorizable_assignment): Likewise.
2600 (vectorizable_shift): Likewise.
2601 (vectorizable_operation): Likewise.
2602 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2604 2018-06-18 Martin Sebor <msebor@redhat.com>
2606 PR tree-optimization/81384
2607 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2608 * builtins.c (expand_builtin_strnlen): New function.
2609 (expand_builtin): Call it.
2610 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2611 * builtins.def (BUILT_IN_STRNLEN): New.
2612 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2613 Warn for bounds in excess of maximum object size.
2614 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2615 single-value ranges. Handle strnlen.
2616 (handle_builtin_strlen): Handle strnlen.
2617 (strlen_check_and_optimize_stmt): Same.
2618 * doc/extend.texi (Other Builtins): Document strnlen.
2620 2018-06-18 Maya Rashish <coypu@sdf.org>
2622 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2623 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2624 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2626 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2628 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2630 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2632 * tree.c (escaped_string::escape): Replace cast to char * by
2633 const_cast<char *> (unescaped).
2635 2018-06-18 Nick Clifton <nickc@redhat.com>
2638 * tree.c (escaped_string): New class. Converts an unescaped
2639 string into its escaped equivalent.
2640 (warn_deprecated_use): Use the new class to convert the
2641 deprecation message, if present.
2642 (test_escaped_strings): New self test.
2643 (test_c_tests): Add test_escaped_strings.
2644 * doc/extend.texi (deprecated): Add a note that the
2645 deprecation message is affected by the -fmessage-length
2646 option, and that control characters will be escaped.
2647 (#pragma GCC error): Document this pragma.
2648 (#pragma GCC warning): Likewise.
2649 * doc/invoke.texi (-fmessage-length): Document this option's
2650 effect on the #warning and #error preprocessor directives and
2651 the deprecated attribute.
2653 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2655 * tree.c (decl_value_expr_lookup): Revert latest change.
2656 (decl_value_expr_insert): Likewise.
2658 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2660 * gimplify.c (nonlocal_vlas): Delete.
2661 (nonlocal_vla_vars): Likewise.
2662 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2664 (gimplify_body): Do not create and destroy nonlocal_vlas.
2665 * tree-nested.c: Include diagnostic.h.
2666 (use_pointer_in_frame): Tweak.
2667 (lookup_field_for_decl): Add assertion and declare the transformation.
2668 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2669 internal error when the reference is in a wrong context. Do not
2670 create a debug decl by default.
2671 (note_nonlocal_block_vlas): Delete.
2672 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2673 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2674 create a debug decl by default.
2675 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2676 call is in a wrong context.
2677 (fixup_vla_decls): New function.
2678 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2679 debug variables were created.
2680 * tree.c (decl_value_expr_lookup): Add checking assertion.
2681 (decl_value_expr_insert): Likewise.
2683 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2686 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2687 * tree-scalar-evolution.c (interpret_expr): Likewise.
2688 (expression_expensive_p): Likewise.
2689 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2690 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2691 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2692 (ssa_defined_by_minus_one_stmt_p): New.
2694 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2697 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2698 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2699 * dojump.c (do_jump): Likewise.
2700 * expr.c (expand_expr_real_2): Check operand type's sign.
2701 * fold-const.c (const_unop): Handle ABSU_EXPR.
2702 (fold_abs_const): Likewise.
2703 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2704 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2705 (strip_sign_op_1): Likesise.
2706 * match.pd: Add new pattern to generate ABSU_EXPR.
2707 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2708 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2709 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2710 * tree-inline.c (estimate_operator_cost): Likewise.
2711 * tree-pretty-print.c (dump_generic_node): Likewise.
2712 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2713 * tree.def (ABSU_EXPR): New.
2715 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2718 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2719 documented as preserved for backward compatibility only.
2720 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2722 PR rtl-optimization/86108
2723 * bb-reorder.c (create_forwarder_block): Renamed to ...
2724 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2725 jump from new landing pad to the second part.
2726 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2729 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2732 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2733 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2734 Only call store_expr for halves if the mode is the same.
2737 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2738 Fix up comment formatting.
2740 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2742 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2743 (typed_splay_tree::closure,
2744 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2745 (typed_splay_tree::typed_splay_tree,
2746 typed_splay_tree::operator =): Declared private.
2747 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2748 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2749 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2750 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2751 typed_splay_tree::splay_tree_splay,
2752 typed_splay_tree::splay_tree_foreach_helper,
2753 typed_splay_tree::splay_tree_insert,
2754 typed_splay_tree::splay_tree_remove,
2755 typed_splay_tree::splay_tree_lookup,
2756 typed_splay_tree::splay_tree_predecessor,
2757 typed_splay_tree::splay_tree_successor,
2758 typed_splay_tree::splay_tree_min,
2759 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2760 (typed_splay_tree::root, typed_splay_tree::comp,
2761 typed_splay_tree::delete_key,
2762 typed_splay_tree::delete_value): New data members.
2763 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2764 typed_splay_tree::remove.
2766 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2768 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2769 -mginv and -mno-ginv to the assembler.
2770 * config/mips/mips.opt (-mcrc): New option.
2772 * doc/invoke.text (-mcrc): Document.
2775 2018-06-15 Nick Clifton <nickc@redhat.com>
2778 * tree.c (escaped_string): New class. Converts an unescaped
2779 string into its escaped equivalent.
2780 (warn_deprecated_use): Use the new class to convert the
2781 deprecation message, if present.
2782 (test_escaped_strings): New self test.
2783 (test_c_tests): Add test_escaped_strings.
2784 * doc/extend.texi (deprecated): Add a note that the
2785 deprecation message is affected by the -fmessage-length
2786 option, and that control characters will be escaped.
2787 (#pragma GCC error): Document this pragma.
2788 (#pragma GCC warning): Likewise.
2789 * doc/invoke.texi (-fmessage-length): Document this option's
2790 effect on the #warning and #error preprocessor directives and
2791 the deprecated attribute.
2793 2018-06-15 Richard Biener <rguenther@suse.de>
2795 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2796 here, also noting vector size used.
2797 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2798 size used in MSG_OPTIMIZED_LOCATIONS dump.
2799 (pass_slp_vectorize::execute): Adjust.
2801 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2804 * config/arc/arc.c (arc_return_address_register): Fix
2807 2018-06-15 Richard Biener <rguenther@suse.de>
2810 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2811 leave useless conversion stripping to force_gimple_operand_gsi.
2812 (gimplify_build2): Likewise.
2813 (gimplify_build1): Likewise.
2815 2018-06-15 Richard Biener <rguenther@suse.de>
2818 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2819 before adjusting their block.
2821 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2823 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2824 multilibs for *-*-rtems*.
2825 * config/riscv/t-rtems: New file.
2827 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2830 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2831 unsigned_type_for returns NULL.
2834 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2835 subregs of multi-word pseudos unless the float mode has word size.
2837 2018-06-14 Richard Biener <rguenther@suse.de>
2840 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2841 duplicate and harmful type_hash_canon.
2842 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2844 2018-06-14 Richard Biener <rguenther@suse.de>
2847 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2850 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2852 * config/rtems.h (STDINT_LONG32): Define.
2854 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2855 Prachi Godbole <prachi.godbole@imgtec.com>
2857 * config/mips/mips-cpus.def: Define P6600.
2858 * config/mips/mips-tables.opt: Regenerate.
2859 * config/mips/mips.c (mips_ucbranch_type): New enum.
2860 (mips_rtx_cost_data): Add support for P6600.
2861 (mips_issue_rate): Likewise.
2862 (mips_multipass_dfa_lookahead): Likewise.
2863 (mips_avoid_hazard): Likewise.
2864 (mips_reorg_process_insns): Likewise.
2865 (mips_classify_branch_p6600): New function.
2866 * config/mips/mips.h (TUNE_P6600): New define.
2867 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2868 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2869 * config/mips/mips.md: Include p6600.md.
2870 (processor): Add p6600.
2871 * config/mips/p6600.md: New file.
2872 * doc/invoke.texi: Add p6600 to supported architectures.
2874 2018-06-13 Martin Sebor <msebor@redhat.com>
2876 PR tree-optimization/86114
2877 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2879 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2881 2018-06-13 Richard Biener <rguenther@suse.de>
2883 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2884 Properly set vector type of the intermediate stmt.
2885 * tree-vect-stmts.c (vectorizable_operation): The destination
2886 var always has vectype_out type.
2888 2018-06-13 Jeff Law <law@redhat.com>
2890 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2891 integer 0 for argument to print_rtl_with_bb.
2892 (rl78_reorg): Likewise.
2894 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2896 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2897 from rtx to rtx_insn *.
2898 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2900 (add_sched_insns_for_speculation): Likewise for local "target",
2901 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2902 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2903 from rtx_insn ** to rtx_code_label **.
2904 (reorg_emit_nops): Likewise.
2905 (c6x_reorg): Likewise for local "call_labels".
2906 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2908 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2909 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2910 the loops over LABEL_REFS.
2911 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2913 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2914 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2915 (split_branches): Strengthen local "olabel" from rtx to
2916 rtx_insn *, adding a safe_as_a cast.
2917 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2919 (add_insn_after): Likewise for first two params.
2920 (add_insn_before): Likewise.
2921 (remove_insn): Likewise for param.
2922 (emit_pattern_before_noloc): Likewise for second and third params.
2923 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2924 (emit_call_insn_before_noloc): Likewise.
2925 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2927 (emit_barrier_before): Likewise.
2928 (emit_label_before): Strengthen "label" param from "rtx" to
2929 "rtx_code_label *". Strengthen "before" param from "rtx" to
2931 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2933 (emit_pattern_after_noloc): Likewise.
2934 (emit_insn_after_noloc): Likewise.
2935 (emit_jump_insn_after_noloc): Likewise.
2936 (emit_call_insn_after_noloc): Likewise.
2937 (emit_debug_insn_after_noloc): Likewise.
2938 (emit_barrier_after): Likewise.
2939 (emit_label_after): Likewise for both params.
2940 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2941 "loc" param from "int" to "location_t".
2942 (emit_insn_after_setloc): Likewise.
2943 (emit_jump_insn_after_setloc): Likewise.
2944 (emit_call_insn_after_setloc): Likewise.
2945 (emit_debug_insn_after_setloc): Likewise.
2946 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2947 "loc" param from "int" to "location_t".
2948 (emit_pattern_before): Convert NULL_RTX to NULL.
2949 (emit_insn_before_setloc): Convert "loc" param from "int" to
2951 (emit_jump_insn_before_setloc): Likewise.
2952 (emit_call_insn_before_setloc): Likewise.
2953 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2954 rtx_insn *. Convert "loc" param from "int" to "location_t".
2955 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2956 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2957 Convert 3rd param from "int" to "location_t".
2958 (emit_barrier_before, emit_barrier_after, next_real_insn):
2959 Strengthen param from rtx to rtx_insn *.
2960 (emit_label_before): Strengthen 1st param from "rtx" to
2961 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2963 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2964 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2965 Strengthen 2nd param from "rtx" to "rtx_insn *".
2966 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2967 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2968 Likewise. Convert 3rd param from "int" to "location_t".
2969 (emit_label_after): Strengthen 1st param from "rtx" to
2971 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2973 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2974 from "rtx" to "rtx_insn *".
2976 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2978 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2979 bodies streamed in with -Q.
2980 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2981 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2982 * gimple-streamer-out.c: Include gimple-pretty-print.h
2983 (output_bb): Dump stmts streamed.
2984 * lto-section-out.c: Include print-tree.h
2985 (lto_begin_section): Dump sections created.
2986 (lto_output_decl_index): Dump decl encoded.
2987 * lto-streamer-out.c: Include print-tree.h
2988 (create_output_block): Dump output block created.
2989 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2990 (output_function): Dump function output.
2991 (output_constructor): Dump constructor streamed.
2992 (write_global_stream): Output indexes encoded.
2993 (produce_asm_for_decls): Dump streams encoded.
2994 * lto-streamer.c (streamer_dump_file): New global var.
2995 * lto-streamer.h (streamer_dump_file): Declare.
2996 * passes.c (ipa_write_summaries): Initialize streamer dump.
2997 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3000 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3003 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3004 offsets for register save directives. Emit a second batch of save
3005 directives, if need be, when the function accesses prior frames.
3007 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3009 * config/arc/fpu.md (fmasf4): Force operand to register.
3010 (fnmasf4): Likewise.
3012 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3014 * config/arc/arc-protos.h (arc_pad_return): Remove.
3015 * config/arc/arc.c (machine_function): Remove force_short_suffix
3017 (arc_print_operand): Adjust printing of '&'.
3018 (arc_verify_short): Remove conditional printing of short suffix.
3019 (arc_final_prescan_insn): Remove reference to size_reason.
3020 (pad_return): New function.
3021 (arc_reorg): Call pad_return.
3022 (arc_pad_return): Remove.
3023 (arc_init_machine_status): Remove reference to force_short_suffix.
3024 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3025 (attr length): When attribute iscompact is true force to 2
3026 regardless; in the case of maybe check if we want to force the
3027 instruction to have 4 bytes length.
3028 (nopv): Change it to generate 4 byte long nop as well.
3029 (blockage): New pattern.
3030 (simple_return): Remove call to arc_pad_return.
3031 (p_return_i): Likewise.
3033 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3035 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3037 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3039 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3042 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3044 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3045 for ARC700 and ARCv2.
3047 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3050 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3051 operands[2] instead of operands[1].
3054 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3056 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3057 case, check whether the outer register overlaps an unallocatable
3058 register, not just whether it fits the required class.
3060 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3062 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3064 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3065 (memrefs_conflict_p): Likewise.
3066 (init_alias_analysis): Likewise.
3067 * cfgexpand.c (expand_debug_expr): Likewise.
3068 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3069 * cse.c (fold_rtx): Likewise.
3070 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3071 * expr.c (emit_block_move_hints): Likewise.
3072 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3073 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3074 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3075 (emit_group_store): Likewise.
3076 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3077 to read the PRE/POST_MODIFY increment.
3078 * calls.c (store_one_arg): Use strip_offset.
3079 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3081 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3083 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3084 (simplify_binary_operation_1): Extend CONST_INT handling to
3086 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3087 than a HOST_WIDE_INT.
3088 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3090 (adjust_mems, add_stores): Update accodingly.
3091 (vt_canonicalize_addr): Track polynomial offsets.
3092 (emit_note_insn_var_location): Likewise.
3093 (vt_add_function_parameter): Likewise.
3094 (vt_initialize): Likewise.
3096 2018-06-12 Jeff Law <law@redhat.com>
3098 * config.gcc (alpha*-*-freebsd*): Remove.
3099 * config/alpha/freebsd.h: Remove.
3101 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3104 * spellcheck-tree.c (levenshtein_distance): Rename to...
3105 (get_edit_distance): ...this, and update for underlying renaming.
3106 * spellcheck-tree.h (levenshtein_distance): Rename to...
3107 (get_edit_distance): ...this.
3108 * spellcheck.c (levenshtein_distance): Rename to...
3109 (get_edit_distance): ...this. Convert from Levenshtein distance
3110 to Damerau-Levenshtein distance by supporting transpositions of
3111 adjacent characters. Rename "v1" to "v_next" and "v0" to
3113 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3114 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3115 update for underlying renaming.
3116 (selftest::levenshtein_distance_unit_test): Rename to...
3117 (selftest::test_get_edit_distance_unit): ...this, and update for
3118 underlying renaming.
3119 (selftest::test_find_closest_string): Add example from PR 69968
3120 where transposition helps
3121 (selftest::test_metric_conditions): Update for renaming.
3122 (selftest::test_metric_conditions): Likewise.
3123 (selftest::spellcheck_c_tests): Likewise.
3124 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3125 (get_edit_distance): ...this.
3126 (best_match::consider): Update for renaming.
3128 2018-06-12 Martin Sebor <msebor@redhat.com>
3130 PR tree-optimization/85259
3131 * builtins.c (compute_objsize): Handle constant offsets.
3132 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3133 true iff a warning has been issued.
3134 * gimple.h (gimple_nonartificial_location): New function.
3135 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3136 gimple_nonartificial_location and handle -Wno-system-headers.
3137 (handle_builtin_stxncpy): Same.
3139 2018-06-12 Martin Sebor <msebor@redhat.com>
3142 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3144 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3146 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3147 BUILTIN_VEC_XST entries for pointer to double and long long.
3149 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3152 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3154 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3157 2018-06-12 Martin Liska <mliska@suse.cz>
3159 * doc/options.texi: Document IntegerRange.
3161 2018-06-12 Martin Liska <mliska@suse.cz>
3163 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3164 * opt-functions.awk: Handle Deprecated flag.
3165 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3167 (read_cmdline_option): Report warning for a deprecated option.
3168 * opts.h (struct cl_option): Add new field cl_deprecated.
3169 (CL_ERR_DEPRECATED): New.
3171 2018-06-12 Martin Liska <mliska@suse.cz>
3173 * doc/options.texi: Document Deprecated option flag.
3175 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3177 * config/arc/arc-arch.h (arc_extras): New enum.
3178 (arc_cpu_t):Add field extra.
3179 (arc_cpu_types): Consider the extras.
3180 * config/arc/arc-cpus.def: Add extras info.
3181 * config/arc/arc-opts.h (processor_type): Consider extra field.
3182 * config/arc/arc.c (arc_override_options): Handle extra field.
3184 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3186 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3187 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3189 * config/arc/arc.c (arc_init): Update pic warning.
3190 (irq_range): Update irq range parsing warnings.
3191 (arc_override_options): Update various warning messages.
3192 (arc_handle_aux_attribute): Likewise.
3194 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3196 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3198 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3200 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3201 compiler messages directives.
3203 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3205 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3206 * config/mips/mips-tables.opt: Regenerate.
3207 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3209 * doc/invoke.texi: Document -march=i6500.
3211 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3213 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3214 (i6400_gpmul): Add cpu_unit.
3215 (i6400_gpdiv): Likewise.
3216 (i6400_msa_add_d): Update reservations.
3217 (i6400_msa_int_add) Likewise.
3218 (i6400_msa_short_logic3) Likewise.
3219 (i6400_msa_short_logic2) Likewise.
3220 (i6400_msa_short_logic) Likewise.
3221 (i6400_msa_move) Likewise.
3222 (i6400_msa_cmp) Likewise.
3223 (i6400_msa_short_float2) Likewise.
3224 (i6400_msa_div_d) Likewise.
3225 (i6400_msa_long_logic1) Likewise.
3226 (i6400_msa_long_logic2) Likewise.
3227 (i6400_msa_mult) Likewise.
3228 (i6400_msa_long_float2) Likewise.
3229 (i6400_msa_long_float4) Likewise.
3230 (i6400_msa_long_float5) Likewise.
3231 (i6400_msa_long_float8) Likewise.
3232 (i6400_fpu_fadd): Include frint type.
3233 (i6400_fpu_store): New define_insn_reservation.
3234 (i6400_fpu_load): Likewise.
3235 (i6400_fpu_move): Likewise.
3236 (i6400_fpu_fcmp): Likewise.
3237 (i6400_fpu_fmadd): Likewise.
3238 (i6400_int_mult): Include imul3nc type and update reservation.
3239 (i6400_int_div): Include idiv3 type and update reservation.
3240 (i6400_int_load): Update to check type not move_type.
3241 (i6400_int_store): Likewise.
3242 (i6400_int_prefetch): Set zero latency.
3244 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3246 * gcc.c: Document new %@{...} sequence.
3247 (LINK_COMMAND_SPEC): Use it for the -L switches.
3248 (cpp_unique_options): Use it for the -I switches.
3249 (at_file_argbuf): New global variable.
3250 (in_at_file): Likewise.
3251 (alloc_args): Create at_file_argbuf.
3252 (clear_args): Truncate at_file_argbuf.
3253 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3254 (open_at_file): New function.
3255 (close_at_file): Likewise.
3256 (create_at_file): Delete.
3257 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3260 (validate_switches_from_spec): Deal with %@{...} sequence.
3261 (validate_switches): Likewise.
3262 (driver::finalize): Call clear_args.
3264 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3266 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3268 2018-06-11 Martin Sebor <msebor@redhat.com>
3270 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3271 (Wno-class-memaccess): Add @opindex.
3272 (Wno-templates, Wno-multiple-inheritance): Same.
3273 (Wno-virtual-inheritance, Wno-namespaces): Same.
3274 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3275 (Wno-format-overflow, Wno-format-truncation): Same.
3276 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3277 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3278 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3279 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3280 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3281 (Wno-misspelled-isr): Same.
3283 2018-06-11 Martin Sebor <msebor@redhat.com>
3285 * PR tree-optimization/86083
3286 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3288 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3290 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3292 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3295 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3296 on the correct operand.
3297 (*movdi_internal64): Ditto.
3299 2018-06-11 Martin Liska <mliska@suse.cz>
3301 PR tree-optimization/86089
3302 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3304 2018-06-11 Julia Koval <julia.koval@intel.com>
3306 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3307 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3308 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3310 2018-06-11 Olivier Hainque <hainque@adacore.com>
3312 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3313 for Ada with strict dwarf2.
3315 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3318 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3321 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3323 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3325 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3327 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3330 2018-06-08 Martin Liska <mliska@suse.cz>
3332 * tree-cfg.h (debug_function): Fix argument type to match
3335 2018-06-08 Martin Liska <mliska@suse.cz>
3337 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3338 Remove usage of MPX-related (and removed) fields.
3339 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3341 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3343 * cfg.c (debug): Use TDF_NONE rather than 0.
3344 * cfghooks.c (debug): Likewise.
3345 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3346 (struct dump_option_value_info): Convert to...
3347 (struct kv_pair): ...this template type.
3348 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3350 (optinfo_verbosity_options): Likewise.
3351 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3353 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3354 than int for "optgroup_flags" param.
3355 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3357 (dump_dec): Likewise.
3358 (dump_finish): Use TDF_NONE rather than 0.
3359 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3360 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3361 than 0. Update for change to option_ptr.
3362 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3363 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3364 0. Update for changes to optinfo_verbosity_options and
3366 (opt_info_switch_p): Convert optgroup_flags from int to
3368 (dump_basic_block): Use dump_flags_t rather than int
3369 for "dump_kind" param.
3370 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3371 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3372 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3373 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3374 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3375 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3376 TDF_NONE): Convert from macros to...
3377 (enum dump_flag): ...this new enum.
3378 (dump_flags_t): Update to use enum.
3379 (operator|, operator&, operator~, operator|=, operator&=):
3380 Implement for dump_flags_t.
3381 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3382 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3383 Convert from macros to...
3384 (enum optgroup_flag): ...this new enum.
3385 (optgroup_flags_t): New typedef.
3386 (operator|, operator|=): Implement for optgroup_flags_t.
3387 (struct dump_file_info): Convert field "alt_flags" to
3388 dump_flags_t. Convert field "optgroup_flags" to
3390 (dump_basic_block): Use dump_flags_t rather than int for param.
3391 (dump_generic_expr_loc): Likewise.
3392 (dump_dec): Likewise.
3393 (dump_register): Convert param "optgroup_flags" to
3395 (opt_info_enable_passes): Likewise.
3396 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3398 * gimple-pretty-print.c (debug): Likewise.
3399 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3400 (merged_store_group::apply_stores): Likewise.
3401 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3402 * gimple.c (verify_gimple_pp): Likewise.
3403 * graphite-poly.c (print_pbb_body): Likewise.
3404 * passes.c (pass_manager::register_one_dump_file): Convert
3405 local "optgroup_flags" to optgroup_flags_t.
3406 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3408 (debug_body): Likewise.
3409 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3410 to optgroup_flags_t.
3411 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3413 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3414 (convert_mult_to_fma): Likewise.
3415 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3416 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3417 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3418 "dump_kind" to dump_flags_t.
3420 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3422 * config/rs6000/rs6000.c (min, max): Delete.
3424 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3426 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3427 -mabi=spe and -mabi=no-spe.
3429 2018-06-08 Martin Liska <mliska@suse.cz>
3431 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3432 where we expect an existing summary.
3434 2018-06-08 Martin Liska <mliska@suse.cz>
3436 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3437 * ipa-inline.h (estimate_edge_growth): Likewise.
3439 2018-06-08 Martin Liska <mliska@suse.cz>
3441 * cgraph.c (function_version_hasher::hash): Use
3442 cgraph_node::get_uid ().
3443 (function_version_hasher::equal):
3444 * cgraph.h (cgraph_node::get_uid): New method.
3445 * ipa-inline.c (update_caller_keys): Use
3446 cgraph_node::get_uid ().
3447 (update_callee_keys): Likewise.
3448 * ipa-utils.c (searchc): Likewise.
3449 (ipa_reduced_postorder): Likewise.
3450 * lto-cgraph.c (input_node): Likewise.
3451 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3452 * symbol-summary.h (symtab_insertion): Likewise.
3453 (symtab_removal): Likewise.
3454 (symtab_duplication): Likewise.
3455 * tree-pretty-print.c (dump_function_header): Likewise.
3456 * tree-sra.c (convert_callers_for_node): Likewise.
3458 2018-06-08 Martin Liska <mliska@suse.cz>
3460 * cgraph.c (symbol_table::create_edge): Always assign a new
3462 (symbol_table::free_edge): Do not recycle numbers.
3463 * cgraph.h (cgraph_edge::get): New method.
3464 * symbol-summary.h (symtab_removal): Use it.
3465 (symtab_duplication): Likewise.
3466 (call_summary::hashable_uid): Remove.
3468 2018-06-08 Martin Liska <mliska@suse.cz>
3470 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3471 (initialize_growth_caches): Remove.
3472 (free_growth_caches): Likewise.
3473 (do_estimate_edge_time): Use edge_growth_cache.
3474 (do_estimate_edge_size): Likewise.
3475 (do_estimate_edge_hints): Likewise.
3476 * ipa-inline.c (reset_edge_caches): Likewise.
3477 (recursive_inlining): Likewise.
3478 (inline_small_functions): Likewise.
3479 * ipa-inline.h (initialize_growth_caches): Remove.
3480 (estimate_edge_size): Likewise.
3481 (estimate_edge_time): Likewise.
3482 (estimate_edge_hints): Likewise.
3483 (reset_edge_growth_cache): Likewise.
3484 * symbol-summary.h (call_summary::remove): New method.
3486 2018-06-08 Martin Liska <mliska@suse.cz>
3488 * ipa-cp.c (class edge_clone_summary): New summary.
3489 (grow_edge_clone_vectors): Remove.
3490 (ipcp_edge_duplication_hook): Remove.
3491 (class edge_clone_summary_t): New call_summary class.
3492 (ipcp_edge_removal_hook): Remove.
3493 (edge_clone_summary_t::duplicate): New function.
3494 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3495 (create_specialized_node): Likewise.
3496 (ipcp_driver): Initialize edge_clone_summaries and do not
3499 2018-06-08 Martin Liska <mliska@suse.cz>
3501 * symbol-summary.h (get): New function.
3502 (call_summary::m_initialize_when_cloning): New class member.
3504 2018-06-08 Martin Liska <mliska@suse.cz>
3506 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3507 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3508 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3509 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3510 (remove_cgraph_node_from_order): Use the removed_nodes set.
3511 (do_per_function_toporder): Likwise.
3512 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3513 instead of summary_uid.
3514 (symtab_removal): Likewise.
3515 (symtab_duplication): Likewise.
3517 2018-06-08 Martin Liska <mliska@suse.cz>
3519 * ipa-cp.c (ipcp_store_bits_results): Use
3520 ipcp_transformation_sum.
3521 (ipcp_store_vr_results): Likewise.
3522 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3524 (ipcp_transformation_initialize): ... this.
3525 (ipa_set_node_agg_value_chain):
3526 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3527 (write_ipcp_transformation_info): Likewise.
3528 (read_ipcp_transformation_info): Likewise.
3529 (ipcp_update_bits): Likewise.
3530 (ipcp_update_vr): Likewise.
3531 (ipcp_transform_function): Likewise.
3532 * ipa-prop.h: Rename ipcp_transformation_summary to
3533 ipcp_transformation.
3534 (class ipcp_transformation_t): New function summary.
3535 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3536 (ipa_get_agg_replacements_for_node): Likewise.
3538 2018-06-08 Martin Liska <mliska@suse.cz>
3540 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3542 (class funct_state_summary_t): New function_summary class.
3543 (has_function_state): Remove.
3544 (get_function_state): Likewise.
3545 (set_function_state): Likewise.
3546 (add_new_function): Likewise.
3547 (funct_state_summary_t::insert): New function.
3548 (duplicate_node_data): Remove.
3549 (remove_node_data): Remove.
3550 (funct_state_summary_t::duplicate): New function.
3551 (register_hooks): Create new funct_state_summaries.
3552 (pure_const_generate_summary): Use it.
3553 (pure_const_write_summary): Likewise.
3554 (pure_const_read_summary): Likewise.
3555 (propagate_pure_const): Likewise.
3556 (propagate_nothrow): Likewise.
3557 (dump_malloc_lattice): Likewise.
3558 (propagate_malloc): Likewise.
3559 (execute): Do not register hooks, just remove summary
3561 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3564 2018-06-08 Martin Liska <mliska@suse.cz>
3566 * ipa-reference.c (remove_node_data): Remove.
3567 (duplicate_node_data): Likewise.
3568 (class ipa_ref_var_info_summary_t): New class.
3569 (class ipa_ref_opt_summary_t): Likewise.
3570 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3571 (get_reference_optimization_summary): Use
3572 ipa_ref_opt_sum_summaries.
3573 (set_reference_vars_info): Remove.
3574 (set_reference_optimization_summary): Likewise.
3575 (ipa_init): Create summaries.
3576 (init_function_info): Use function summary.
3577 (ipa_ref_opt_summary_t::duplicate): New function.
3578 (ipa_ref_opt_summary_t::remove): New function.
3579 (get_read_write_all_from_node): Fix GNU coding style.
3580 (propagate): Use function summary.
3581 (write_node_summary_p): Fix GNU coding style.
3582 (stream_out_bitmap): Likewise.
3583 (ipa_reference_read_optimization_summary): Use function summary.
3584 (ipa_reference_c_finalize): Do not release hooks.
3586 2018-06-08 Martin Liska <mliska@suse.cz>
3588 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3589 (analyze_function_body): Extract multiple calls of get_create.
3590 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3591 * ipa-inline.c (recursive_inlining): Use ::get method.
3592 * ipa-inline.h (estimate_edge_growth): Likewise.
3594 2018-06-08 Martin Liska <mliska@suse.cz>
3596 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3598 (hsa_function_summary::hsa_function_summary): Use the new enum
3600 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3601 * hsa-gen.c (hsa_get_host_function): Likewise.
3602 (get_brig_function_name): Likewise.
3603 * ipa-hsa.c (process_hsa_functions): Likewise.
3604 (ipa_hsa_write_summary): Likewise.
3605 * symbol-summary.h (symtab_duplication): Use ::get function/
3606 (get): New function.
3608 2018-06-08 Martin Liska <mliska@suse.cz>
3610 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3612 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3613 (hsa_register_kernel): Likewise.
3614 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3615 * hsa-gen.c (hsa_get_host_function): Likewise.
3616 (get_brig_function_name): Likewise.
3617 (generate_hsa): Likewise.
3618 (pass_gen_hsail::execute): Likewise.
3619 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3620 (devirtualization_time_bonus): Likewise.
3621 (ipcp_propagate_stage): Likewise.
3622 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3623 (edge_set_predicate): Likewise.
3624 (evaluate_conditions_for_known_args): Likewise.
3625 (evaluate_properties_for_edge): Likewise.
3626 (ipa_fn_summary::reset): Likewise.
3627 (ipa_fn_summary_t::duplicate): Likewise.
3628 (dump_ipa_call_summary): Likewise.
3629 (ipa_dump_fn_summary): Likewise.
3630 (analyze_function_body): Likewise.
3631 (compute_fn_summary): Likewise.
3632 (estimate_edge_devirt_benefit): Likewise.
3633 (estimate_edge_size_and_time): Likewise.
3634 (estimate_calls_size_and_time): Likewise.
3635 (estimate_node_size_and_time): Likewise.
3636 (inline_update_callee_summaries): Likewise.
3637 (remap_edge_change_prob): Likewise.
3638 (remap_edge_summaries): Likewise.
3639 (ipa_merge_fn_summary_after_inlining): Likewise.
3640 (ipa_update_overall_fn_summary): Likewise.
3641 (read_ipa_call_summary): Likewise.
3642 (inline_read_section): Likewise.
3643 (write_ipa_call_summary): Likewise.
3644 (ipa_fn_summary_write): Likewise.
3645 (ipa_free_fn_summary): Likewise.
3646 * ipa-hsa.c (process_hsa_functions): Likewise.
3647 (ipa_hsa_write_summary): Likewise.
3648 (ipa_hsa_read_section): Likewise.
3649 * ipa-icf.c (sem_function::merge): Likewise.
3650 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3651 (do_estimate_edge_time): Likewise.
3652 (estimate_size_after_inlining): Likewise.
3653 (estimate_growth): Likewise.
3654 (growth_likely_positive): Likewise.
3655 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3656 (inline_call): Likewise.
3657 * ipa-inline.c (caller_growth_limits): Likewise.
3658 (can_inline_edge_p): Likewise.
3659 (can_inline_edge_by_limits_p): Likewise.
3660 (compute_uninlined_call_time): Likewise.
3661 (compute_inlined_call_time): Likewise.
3662 (want_inline_small_function_p): Likewise.
3663 (edge_badness): Likewise.
3664 (update_caller_keys): Likewise.
3665 (update_callee_keys): Likewise.
3666 (recursive_inlining): Likewise.
3667 (inline_small_functions): Likewise.
3668 (inline_to_all_callers_1): Likewise.
3669 (dump_overall_stats): Likewise.
3670 (early_inline_small_functions): Likewise.
3671 (early_inliner): Likewise.
3672 * ipa-inline.h (estimate_edge_growth): Likewise.
3673 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3674 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3675 * ipa-prop.h (IPA_NODE_REF): Likewise.
3676 (IPA_EDGE_REF): Likewise.
3677 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3678 (propagate_malloc): Likewise.
3679 * ipa-split.c (execute_split_functions): Likewise.
3680 * symbol-summary.h: Rename get to get_create.
3682 (get_create): Likewise.
3683 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3685 2018-06-08 Martin Liska <mliska@suse.cz>
3687 * symbol-summary.h (release): Move definition out of class
3689 (symtab_removal): Likewise.
3690 (symtab_duplication): Likewise.
3692 2018-06-08 Martin Liska <mliska@suse.cz>
3694 * symbol-summary.h (function_summary): Move constructor
3695 implementation out of class declaration.
3696 (release): Likewise.
3697 (symtab_insertion): Likewise.
3698 (symtab_removal): Likewise.
3699 (symtab_duplication): Likewise.
3702 2018-06-08 Martin Liska <mliska@suse.cz>
3704 * Makefile.in: Remove support for MPX (macros, related functions,
3705 fields in cgraph_node, ...).
3706 * builtin-types.def (BT_BND): Likewise.
3707 (BT_FN_BND_CONST_PTR): Likewise.
3708 (BT_FN_CONST_PTR_BND): Likewise.
3709 (BT_FN_VOID_PTR_BND): Likewise.
3710 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3711 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3712 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3713 (expand_builtin_mempcpy_with_bounds): Likewise.
3714 (expand_builtin_memset_with_bounds): Likewise.
3715 (expand_builtin_memset_args): Likewise.
3716 (std_expand_builtin_va_start): Likewise.
3717 (expand_builtin): Likewise.
3718 (expand_builtin_with_bounds): Likewise.
3719 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3720 (DEF_LIB_BUILTIN_CHKP): Likewise.
3721 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3722 (DEF_CHKP_BUILTIN): Likewise.
3723 (BUILT_IN_MEMCPY): Likewise.
3724 (BUILT_IN_MEMMOVE): Likewise.
3725 (BUILT_IN_MEMPCPY): Likewise.
3726 (BUILT_IN_MEMSET): Likewise.
3727 (BUILT_IN_STPCPY): Likewise.
3728 (BUILT_IN_STRCAT): Likewise.
3729 (BUILT_IN_STRCHR): Likewise.
3730 (BUILT_IN_STRCPY): Likewise.
3731 (BUILT_IN_STRLEN): Likewise.
3732 (BUILT_IN_MEMCPY_CHK): Likewise.
3733 (BUILT_IN_MEMMOVE_CHK): Likewise.
3734 (BUILT_IN_MEMPCPY_CHK): Likewise.
3735 (BUILT_IN_MEMSET_CHK): Likewise.
3736 (BUILT_IN_STPCPY_CHK): Likewise.
3737 (BUILT_IN_STRCAT_CHK): Likewise.
3738 (BUILT_IN_STRCPY_CHK): Likewise.
3739 * calls.c (store_bounds): Likewise.
3740 (emit_call_1): Likewise.
3741 (special_function_p): Likewise.
3742 (maybe_warn_nonstring_arg): Likewise.
3743 (initialize_argument_information): Likewise.
3744 (finalize_must_preallocate): Likewise.
3745 (compute_argument_addresses): Likewise.
3746 (expand_call): Likewise.
3747 * cfgexpand.c (expand_call_stmt): Likewise.
3748 (expand_return): Likewise.
3749 (expand_gimple_stmt_1): Likewise.
3750 (pass_expand::execute): Likewise.
3751 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3752 (cgraph_node::remove): Likewise.
3753 (cgraph_node::dump): Likewise.
3754 (cgraph_node::verify_node): Likewise.
3755 * cgraph.h (chkp_function_instrumented_p): Likewise.
3756 (symtab_node::get_alias_target): Likewise.
3757 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3758 (cgraph_local_p): Likewise.
3759 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3760 (cgraph_edge::rebuild_references): Likewise.
3761 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3762 (walk_polymorphic_call_targets): Likewise.
3763 (cgraph_node::expand_thunk): Likewise.
3764 (symbol_table::output_weakrefs): Likewise.
3765 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3766 (ix86_handle_option): Likewise.
3767 * config/i386/constraints.md: Likewise.
3768 * config/i386/i386-builtin-types.def (BND): Likewise.
3772 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3773 (BDESC_FIRST): Likewise.
3775 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3776 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3777 * config/i386/i386.c (enum reg_class): Likewise.
3778 (ix86_target_string): Likewise.
3779 (ix86_option_override_internal): Likewise.
3780 (ix86_conditional_register_usage): Likewise.
3781 (ix86_valid_target_attribute_inner_p): Likewise.
3782 (ix86_set_indirect_branch_type): Likewise.
3783 (ix86_set_current_function): Likewise.
3784 (ix86_function_arg_regno_p): Likewise.
3785 (init_cumulative_args): Likewise.
3786 (ix86_function_arg_advance): Likewise.
3787 (ix86_function_arg): Likewise.
3788 (ix86_pass_by_reference): Likewise.
3789 (ix86_function_value_regno_p): Likewise.
3790 (ix86_function_value_1): Likewise.
3791 (ix86_function_value_bounds): Likewise.
3792 (ix86_return_in_memory): Likewise.
3793 (ix86_setup_incoming_vararg_bounds): Likewise.
3794 (ix86_va_start): Likewise.
3795 (indirect_thunk_need_prefix): Likewise.
3796 (print_reg): Likewise.
3797 (ix86_print_operand): Likewise.
3798 (ix86_expand_call): Likewise.
3799 (ix86_output_function_return): Likewise.
3800 (reg_encoded_number): Likewise.
3801 (BDESC_VERIFYS): Likewise.
3802 (ix86_init_mpx_builtins): Likewise.
3803 (ix86_init_builtins): Likewise.
3804 (ix86_emit_cmove): Likewise.
3805 (ix86_emit_move_max): Likewise.
3806 (ix86_expand_builtin): Likewise.
3807 (ix86_builtin_mpx_function): Likewise.
3808 (ix86_get_arg_address_for_bt): Likewise.
3809 (ix86_load_bounds): Likewise.
3810 (ix86_store_bounds): Likewise.
3811 (ix86_load_returned_bounds): Likewise.
3812 (ix86_store_returned_bounds): Likewise.
3813 (ix86_class_likely_spilled_p): Likewise.
3814 (ix86_hard_regno_mode_ok): Likewise.
3815 (x86_order_regs_for_local_alloc): Likewise.
3816 (ix86_mitigate_rop): Likewise.
3817 (ix86_bnd_prefixed_insn_p): Likewise.
3818 (ix86_mpx_bound_mode): Likewise.
3819 (ix86_make_bounds_constant): Likewise.
3820 (ix86_initialize_bounds): Likewise.
3821 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3822 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3823 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3824 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3825 (TARGET_CHKP_BOUND_MODE): Likewise.
3826 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3827 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3828 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3829 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3830 * config/i386/i386.h (TARGET_MPX): Likewise.
3831 (TARGET_MPX_P): Likewise.
3832 (VALID_BND_REG_MODE): Likewise.
3833 (FIRST_BND_REG): Likewise.
3834 (LAST_BND_REG): Likewise.
3835 (enum reg_class): Likewise.
3836 (BND_REG_P): Likewise.
3837 (BND_REGNO_P): Likewise.
3838 (BNDmode): Likewise.
3839 (ADJUST_INSN_LENGTH): Likewise.
3840 * config/i386/i386.md: Likewise.
3841 * config/i386/i386.opt: Likewise.
3842 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3843 (defined): Likewise.
3844 (LINK_MPX): Likewise.
3845 (MPX_SPEC): Likewise.
3846 (LIBMPX_SPEC): Likewise.
3847 (LIBMPXWRAPPERS_SPEC): Likewise.
3848 (CHKP_SPEC): Likewise.
3849 * config/i386/predicates.md: Likewise.
3850 * dbxout.c (dbxout_type): Likewise.
3851 * doc/extend.texi: Likewise.
3852 * doc/invoke.texi: Likewise.
3853 * doc/md.texi: Likewise.
3854 * doc/tm.texi: Likewise.
3855 * doc/tm.texi.in: Likewise.
3856 * dwarf2out.c (is_base_type): Likewise.
3857 (gen_formal_types_die): Likewise.
3858 (gen_subprogram_die): Likewise.
3859 (gen_type_die_with_usage): Likewise.
3860 (gen_decl_die): Likewise.
3861 (dwarf2out_late_global_decl): Likewise.
3862 * expr.c (expand_assignment): Likewise.
3863 (emit_storent_insn): Likewise.
3864 (store_expr_with_bounds): Likewise.
3865 (store_expr): Likewise.
3866 (expand_expr_real_1): Likewise.
3867 * expr.h (store_expr_with_bounds): Likewise.
3868 * function.c (use_register_for_decl): Likewise.
3869 (struct bounds_parm_data): Likewise.
3870 (assign_parms_augmented_arg_list): Likewise.
3871 (assign_parm_find_entry_rtl): Likewise.
3872 (assign_parm_is_stack_parm): Likewise.
3873 (assign_parm_load_bounds): Likewise.
3874 (assign_bounds): Likewise.
3875 (assign_parms): Likewise.
3876 (expand_function_start): Likewise.
3877 * gcc.c (CHKP_SPEC): Likewise.
3878 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3879 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3880 (wrestrict_dom_walker::check_call): Likewise.
3881 * gimple.c (gimple_build_call_from_tree): Likewise.
3882 * gimple.h (enum gf_mask): Likewise.
3883 (gimple_call_with_bounds_p): Likewise.
3884 (gimple_call_set_with_bounds): Likewise.
3885 * gimplify.c (gimplify_init_constructor): Likewise.
3886 * ipa-cp.c (initialize_node_lattices): Likewise.
3887 (propagate_constants_across_call): Likewise.
3888 (find_more_scalar_values_for_callers_subset): Likewise.
3889 * ipa-hsa.c (process_hsa_functions): Likewise.
3890 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3891 * ipa-icf.c (sem_function::merge): Likewise.
3892 * ipa-inline.c (early_inliner): Likewise.
3893 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3894 (warn_function_cold): Likewise.
3895 (propagate_pure_const): Likewise.
3896 * ipa-ref.h (enum GTY): Likewise.
3897 * ipa-split.c (find_retbnd): Likewise.
3898 (consider_split): Likewise.
3899 (split_function): Likewise.
3900 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3901 * ipa.c (walk_polymorphic_call_targets): Likewise.
3902 (symbol_table::remove_unreachable_nodes): Likewise.
3903 (process_references): Likewise.
3904 (cgraph_build_static_cdtor_1): Likewise.
3905 * lto-cgraph.c (lto_output_node): Likewise.
3906 (output_refs): Likewise.
3907 (compute_ltrans_boundary): Likewise.
3908 (input_overwrite_node): Likewise.
3909 (input_node): Likewise.
3910 (input_cgraph_1): Likewise.
3911 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3912 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3913 (class pass_chkp_instrumentation_passes): Likewise.
3914 (make_pass_chkp_instrumentation_passes): Likewise.
3915 * passes.def: Likewise.
3916 * rtl.h (struct GTY): Likewise.
3917 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3918 * stor-layout.c (layout_type): Likewise.
3919 * symtab.c: Likewise.
3920 * target.def: Likewise.
3921 * targhooks.c (default_chkp_bound_type): Likewise.
3922 (default_chkp_bound_mode): Likewise.
3923 (default_builtin_chkp_function): Likewise.
3924 (default_chkp_function_value_bounds): Likewise.
3925 (default_chkp_make_bounds_constant): Likewise.
3926 (default_chkp_initialize_bounds): Likewise.
3927 * targhooks.h (default_chkp_bound_type): Likewise.
3928 (default_chkp_bound_mode): Likewise.
3929 (default_builtin_chkp_function): Likewise.
3930 (default_chkp_function_value_bounds): Likewise.
3931 (default_chkp_make_bounds_constant): Likewise.
3932 (default_chkp_initialize_bounds): Likewise.
3933 * toplev.c (compile_file): Likewise.
3934 (process_options): Likewise.
3935 * tree-core.h (DEF_BUILTIN): Likewise.
3936 (DEF_BUILTIN_CHKP): Likewise.
3937 * tree-inline.c (declare_return_variable): Likewise.
3938 (remap_gimple_stmt): Likewise.
3939 (copy_bb): Likewise.
3940 (initialize_inlined_parameters): Likewise.
3941 (expand_call_inline): Likewise.
3942 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3943 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3944 (make_pass_ipa_chkp_produce_thunks): Likewise.
3945 (make_pass_chkp): Likewise.
3946 (make_pass_chkp_opt): Likewise.
3947 (make_pass_chkp_instrumentation_passes): Likewise.
3948 * tree-pretty-print.c (dump_generic_node): Likewise.
3949 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3950 * tree-ssa-dce.c (propagate_necessity): Likewise.
3951 (eliminate_unnecessary_stmts): Likewise.
3952 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3953 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3954 * tree-ssa-sccvn.h: Likewise.
3955 * tree-ssa-strlen.c (get_string_length): Likewise.
3956 (valid_builtin_call): Likewise.
3957 (adjust_last_stmt): Likewise.
3958 (handle_builtin_strchr): Likewise.
3959 (handle_builtin_strcpy): Likewise.
3960 (handle_builtin_stxncpy): Likewise.
3961 (handle_builtin_memcpy): Likewise.
3962 (handle_builtin_strcat): Likewise.
3963 (strlen_check_and_optimize_stmt): Likewise.
3964 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3965 * tree-streamer-in.c: Likewise.
3966 * tree-streamer.c (record_common_node): Likewise.
3967 * tree.c (tree_code_size): Likewise.
3968 (wide_int_to_tree_1): Likewise.
3969 (type_contains_placeholder_1): Likewise.
3970 (build_common_tree_nodes): Likewise.
3971 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3972 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3973 (POINTER_BOUNDS_P): Likewise.
3974 (BOUNDED_TYPE_P): Likewise.
3975 (BOUNDED_P): Likewise.
3976 (CALL_WITH_BOUNDS_P): Likewise.
3977 (pointer_bounds_type_node): Likewise.
3978 * value-prof.c (gimple_ic): Likewise.
3979 * var-tracking.c (vt_add_function_parameters): Likewise.
3980 * varasm.c (make_decl_rtl): Likewise.
3981 (assemble_start_function): Likewise.
3982 (output_constant): Likewise.
3983 (maybe_assemble_visibility): Likewise.
3984 * varpool.c (ctor_for_folding): Likewise.
3985 * chkp-builtins.def: Remove.
3986 * ipa-chkp.c: Remove.
3987 * ipa-chkp.h: Remove.
3988 * rtl-chkp.c: Remove.
3989 * rtl-chkp.h: Remove.
3990 * tree-chkp-opt.c: Remove.
3991 * tree-chkp.c: Remove.
3992 * tree-chkp.h: Remove.
3994 2018-06-07 Carl Love <cel@us.ibm.com>
3996 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3997 vextract_fp_from_shortl): Add BE support.
3999 2018-06-07 Paul Koning <ni1d@arrl.net>
4001 * compare-elim.c (try_merge_compare): Don't merge compare if
4002 address contains a side effect.
4003 (try_eliminate_compare): Likewise.
4005 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4007 * config.gcc: Support "tremont".
4008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4009 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4011 * config/i386/i386.c (m_TREMONT): Define.
4012 (processor_target_table): Add "tremont".
4013 (PTA_TREMONT): Define.
4014 (ix86_lea_outperforms): Add TARGET_TREMONT.
4015 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4016 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4017 and M_INTEL_GOLDMONT_PLUS.
4018 (fold_builtin_cpu): Add "tremont".
4019 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4020 (ix86_option_override_internal): Add "tremont".
4021 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4022 (processor_type): Add PROCESSOR_TREMONT.
4023 * config/i386/x86-tune.def: Add m_TREMONT.
4024 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4026 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4028 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4029 symbol defined for msp430i* devices to be lower case.
4031 gcc/testsuite/gcc.target/msp430/
4032 * msp430i-device-symbol.c: New test.
4033 * msp430f-device-symbol.c: New test.
4034 * msp430.h: New test header file.
4036 2018-06-07 Richard Biener <rguenther@suse.de>
4038 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4039 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4040 Properly wrap signed arithmetic if overflow wraps.
4042 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4044 PR tree-optimization/69615
4045 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4046 of a cast from a same precision integral SSA_NAME in a bb dominated
4047 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4048 cast to utype if rhs2 has already a compatible type.
4050 2018-06-07 Richard Biener <rguenther@suse.de>
4052 PR tree-optimization/85935
4053 * graphite-scop-detection.c (find_params_in_bb): Analyze
4054 condition operands with respect to the correct loop. Assert
4055 the analysis doesn't fail.
4057 2018-06-04 Carl Love <cel@us.ibm.com>
4059 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4060 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4061 as it is slightly cheaper.
4062 (first_match_or_eos_index_<mode>):
4063 Calculate index using natural element order.
4064 (first_match_index_<mode>):
4065 Calculate index using natural element order.
4066 (first_match_or_eos_index_<mode>):
4067 Calculate index using natural order.
4068 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4069 for BE and LE modes.
4070 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4071 P9V_BUILTIN_VCLZLSBB_V16QI.
4072 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4075 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4077 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4078 indentation and line wrap for many prototypes. Add missing
4079 @smallexample directives around block of prototypes for vec_xl and
4082 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4084 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4085 track if we pass or return IEEE 128-bit floating point.
4086 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4087 C++ mangling that is compatible with GCC 8.1.
4088 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4089 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4090 floating point types.
4091 (rs6000_function_arg_advance_1): Likewise.
4092 (rs6000_mangle_type): Optionally generate mangled names that match
4093 what GCC 8.1 generated for IEEE 128-bit floating point types.
4094 (rs6000_globalize_decl_name): If we have an external function that
4095 passes or returns IEEE 128-bit types, generate a weak reference
4096 from the mangled name used in GCC 8.1 to the current mangled
4098 (rs6000_init_builtins): Make __ibm128 use the long double type if
4099 long double is IBM extended double. Make __float128 use the long
4100 double type if long double is IEEE 128-bit.
4103 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4104 macro for __ibm128 built-in functions.
4105 (PACK_IF): Add __ibm128 pack/unpack functions.
4106 (UNPACK_IF): Likewise.
4107 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4108 enable long double built-in functions if long double is IEEE
4109 128-bit floating point.
4110 (rs6000_invalid_builtin): Update long double built-in function
4112 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4113 functions, adjust the built-in function to use the long double
4114 built-in function if __ibm128 and long double are the same type.
4115 * doc/extend.texi (PowerPC builtins): Update documention for
4116 __builtin_{,un}pack_longdouble. Add documentation for
4117 __builtin_{,un}pack_ibm128.
4119 2018-06-06 Jim Wilson <jimw@sifive.com>
4121 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4122 (struct machine_function): New field interrupt_mode.
4123 (riscv_handle_type_attribute): New function. Add forward declaration.
4124 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4125 (riscv_expand_epilogue): Check interrupt_mode field.
4126 (riscv_set_current_function): Check interrupt attribute args and
4127 set interrupt_mode field.
4128 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4129 (riscv_sret, riscv_uret): New.
4130 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4131 new arguments to interrupt attribute.
4133 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4136 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4137 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4139 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4141 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4142 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4143 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4144 several redundant entries.
4146 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4148 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4149 type from "rtx" to "rtx_insn *".
4150 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4151 for local "call_insn", removing cast.
4152 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4154 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4156 PR tree-optimization/86066
4157 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4158 for BIT_INSERT_EXPR stores.
4160 2018-06-06 Richard Biener <rguenther@suse.de>
4162 PR tree-optimization/86062
4163 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4164 component refs ontop
4165 of to be offsetted base.
4167 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4169 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4170 to be static and remove check on interrupt attribute name.
4172 gcc/testsuite/gcc.target/msp430/
4173 * function-attributes-4.c: New test.
4174 * static-interrupts.c: New test.
4176 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4178 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4179 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4181 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4184 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4186 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4187 Remove second argument, change how error is called.
4188 (aarch64_layout_arg): Remove second argument from
4189 aarch64_err_no_fpadvsimd call.
4190 (aarch64_init_cumulative_args): Ditto.
4191 (aarch64_gimplify_va_arg_expr): Ditto.
4192 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4194 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4196 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4197 (*simple_return_indirect_internal<mode>): Rename from
4198 simple_return_indirect_internal. Use W mode iterator.
4199 (rstorssp): New expander.
4200 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4201 (clrssbsy): New expander.
4202 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4204 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4206 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4208 (cmse_check_pointed_object): Likewise.
4210 2018-06-05 Martin Liska <mliska@suse.cz>
4212 PR gcov-profile/47618
4213 * doc/invoke.texi: Document how -fprofile-dir format
4216 2018-06-05 Richard Biener <rguenther@suse.de>
4218 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4219 removal pretend DOM info isn't available so we do not update
4220 it and only remove edges, not dominated blocks. Actually free
4221 DOM info in case we removed something. Remove unreachable blocks.
4222 (mfb_keep_latches): Work with either DOM info or marked backedges.
4223 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4224 first. Mark backedges if DOM info isn't available.
4225 (Re-)compute DOM info after cleanup_control_flow_pre.
4227 2018-06-05 Richard Biener <rguenther@suse.de>
4229 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4230 (locus_discrim_hasher::hash): Adjust.
4231 (locus_discrim_hasher::equal): Likewise.
4232 (next_discriminator_for_locus): Work on line directly.
4233 (same_line_p): Pass in expanded locus1 as well.
4234 (assign_discriminators): Avoid redundant location expansions.
4236 2018-06-05 Richard Biener <rguenther@suse.de>
4238 PR tree-optimization/86046
4239 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4240 if required after clearing TREE_ADDRESSABLE.
4242 2018-06-05 Richard Biener <rguenther@suse.de>
4244 PR tree-optimization/86047
4245 * tree-ssa-loop.c (for_each_index): Glob handling of all
4246 decls and constants and really handle all of them.
4248 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4251 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4252 qualifier_void_pointer and qualifier_const_void_pointer.
4253 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4254 (arm_init_builtins): Handle the above.
4255 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4256 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4257 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4260 2018-06-05 Martin Liska <mliska@suse.cz>
4262 * auto-profile.c (read_autofdo_file): Do not use
4263 gcov_ctr_summary struct.
4264 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4265 * coverage.c (struct counts_entry): Likewise.
4266 (read_counts_file): Read just single summary entry.
4267 (get_coverage_counts): Use gcov_summary struct.
4268 * coverage.h (get_coverage_counts): Likewise.
4269 * gcov-dump.c (dump_working_sets): Likewise.
4270 (tag_summary): Dump just single summary.
4271 * gcov-io.c (gcov_write_summary): Write just histogram
4273 (gcov_read_summary): Read just single summary.
4274 (compute_working_sets): Use gcov_summary struct.
4275 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4276 of GCOV_COUNTERS_SUMMABLE.
4277 (GCOV_COUNTERS_SUMMABLE): Remove.
4278 (GCOV_FIRST_VALUE_COUNTER): Replace with
4279 GCOV_COUNTER_V_INTERVAL.
4280 (struct gcov_ctr_summary): Remove.
4281 (struct gcov_summary): Directly use fields of former
4283 (compute_working_sets): Use gcov_summary struct.
4284 * gcov.c (read_count_file): Do not use ctrs fields.
4285 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4287 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4289 * profile.c: Likewise.
4290 * profile.h: Likewise.
4292 2018-06-05 Martin Liska <mliska@suse.cz>
4294 PR gcov-profile/84846
4295 * gcov.c (output_lines): Print working directory only
4296 in intermediate format.
4298 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4300 * config/s390/s390-builtin-types.def: Add void function type.
4301 * config/s390/s390-builtins.def: Use the function type for the
4304 2018-06-04 Jim Wilson <jimw@sifive.com>
4306 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4308 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4309 and maybe_eh_return. Change regno to unsigned int. Use new args to
4310 handle EH_RETURN_DATA_REGNO registers properly.
4311 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4312 (riscv_expand_epilogue): Update comment. Change argument name and
4313 type. Update code to use new name and type. Pass new args to
4314 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4316 * config/riscv/riscv.md (NORMAL_RETURN): New.
4317 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4318 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4319 (eh_return): Call gen_eh_return_internal and emit barrier.
4320 (eh_return_internal): Call riscv_expand_epilogue.
4322 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4324 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4325 bit_insertion field and declare can_be_merged_into method.
4326 (merged_store_group::can_be_merged_into): New method.
4327 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4328 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4329 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4331 2018-06-04 Richard Biener <rguenther@suse.de>
4333 PR tree-optimization/85955
4334 * builtins.c (fold_builtin_sincos): Convert pointers to
4335 destination to appropriate type before dereferencing.
4337 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4339 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4341 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4343 * expr.c (expand_expr_real_1): Force the operand into memory if
4344 its TYPE_MODE is BLKmode and if there is no integer mode for
4345 the number of bits being extracted.
4347 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4351 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4352 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4354 2018-06-04 Richard Biener <rguenther@suse.de>
4356 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4357 (cleanup_tree_cfg_noloop): ... single caller. Do
4358 start_recording_case_labels later.
4360 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4362 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4363 to _IMMINTRIN_H_INCLUDED.
4364 * config/i386/pconfigintrin.h: Ditto.
4365 * config/i386/waitpkgintrin.h: Ditto.
4366 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4367 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4368 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4369 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4370 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4371 waitpkgintrin.h and cldemoteintrin.h.
4373 2018-06-04 Richard Biener <rguenther@suse.de>
4375 PR tree-optimization/86038
4376 * tracer.c (find_best_successor): Check probability for
4377 being initialized, bail out if not.
4379 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4382 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4383 of bits to ignore when comparing architectures.
4385 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4387 PR tree-optimization/69615
4388 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4389 maximum or minimum of the type, try to merge it also as if
4390 range1 is + [-, x - 1] or + [x + 1, -].
4393 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4395 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4397 PR tree-optimization/86034
4398 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4399 the unsigned bitfield type in a bit insertion sequence if it does not
4400 have a larger precision than the bitfield size.
4401 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4403 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4405 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4407 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4409 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4410 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4411 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4412 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4414 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4416 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4417 Disable -fdelete-null-pointer-checks for ELF toolchain.
4419 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4420 Kito Cheng <kito.cheng@gmail.com>
4422 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4423 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4424 (nds32*-*-*): Add glibc and uclibc conditions.
4425 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4426 (TARGET_EXCEPT_UNWIND_INFO): Define.
4427 * config/nds32/elf.h: New file.
4428 * config/nds32/linux.h: New file.
4429 * config/nds32/nds32-elf.opt: New file.
4430 * config/nds32/nds32-linux.opt: New file.
4431 * config/nds32/nds32-fp-as-gp.c
4432 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4433 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4435 (nds32_asm_file_end): Ditto.
4436 (nds32_print_operand): Ditto.
4437 (nds32_insert_attributes): Ditto.
4438 (nds32_init_libfuncs): New function.
4439 (TARGET_HAVE_TLS): Define.
4440 (TARGET_INIT_LIBFUNCS): Define.
4441 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4443 (TARGET_ELF): Apply different mcmodel setting.
4444 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4445 been migrated into elf.h and linux.h files.
4446 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4447 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4448 (mcmodel): The content has been migrated into nds32-elf.opt and
4449 nds32-linux.opt files.
4450 * config/nds32/t-elf: New file.
4451 * config/nds32/t-linux: New file.
4453 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4454 Shiva Chen <shiva0217@gmail.com>
4456 * config/nds32/constants.md (unspec_volatile_element): Add
4457 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4458 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4460 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4461 (make_pass_nds32_fp_as_gp): Declare.
4462 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4464 (nds32_asm_function_end_prologue): Remove unused asm output.
4465 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4466 (nds32_asm_file_start): Output necessary fp_as_gp information.
4467 (nds32_option_override): Adjust register usage.
4468 (nds32_expand_prologue): Consider fp_as_gp situation.
4469 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4470 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4473 (simple_return): Ditto.
4474 (omit_fp_begin): Output special directive for fp_as_gp.
4475 (omit_fp_end): Output special directive for fp_as_gp.
4476 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4477 mforbid-fp-as-gp): New options.
4479 2018-06-01 Mark Wielaard <mark@klomp.org>
4481 * dwarf2out.c (dwarf2out_finish): Remove generation of
4482 DW_AT_loclists_base.
4484 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4486 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4487 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4488 (struct merged_store_group): Add bit_insertion field.
4489 (dump_char_array): Use standard hexadecimal format.
4490 (merged_store_group::merged_store_group): Set bit_insertion to false.
4491 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4492 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4493 also print the mask in the dump file.
4494 (pass_store_merging::gate): Minor tweak.
4495 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4496 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4497 stores with INTEGER_CST stores.
4498 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4499 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4500 and use it throughout. Generate bit insertion sequences if need be.
4501 (pass_store_merging::process_store): Remove redundant condition.
4502 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4504 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4506 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4507 the 128-bit floating point types. Fix function comment.
4509 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4511 * config/aarch64/aarch64-simd.md
4512 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4514 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4517 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4519 PR tree-optimization/85989
4520 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4522 (backprop::intersect_uses): Check it when deciding whether this
4523 is a backedge reference.
4524 (backprop::process_block): Add each phi to m_visited_phis
4525 after visiting it, then clear it at the end.
4527 2018-06-01 Richard Biener <rguenther@suse.de>
4529 * tree-vectorizer.h (vect_dr_stmt): New function.
4530 (vect_get_load_cost): Adjust.
4531 (vect_get_store_cost): Likewise.
4532 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4533 Use vect_dr_stmt instead of DR_SMTT.
4534 (vect_record_base_alignments): Likewise.
4535 (vect_calculate_target_alignment): Likewise.
4536 (vect_compute_data_ref_alignment): Likewise and make static.
4537 (vect_update_misalignment_for_peel): Likewise.
4538 (vect_verify_datarefs_alignment): Likewise.
4539 (vector_alignment_reachable_p): Likewise.
4540 (vect_get_data_access_cost): Likewise. Pass down
4541 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4542 (vect_get_peeling_costs_all_drs): Likewise.
4543 (vect_peeling_hash_get_lowest_cost): Likewise.
4544 (vect_enhance_data_refs_alignment): Likewise.
4545 (vect_find_same_alignment_drs): Likewise.
4546 (vect_analyze_data_refs_alignment): Likewise.
4547 (vect_analyze_group_access_1): Likewise.
4548 (vect_analyze_group_access): Likewise.
4549 (vect_analyze_data_ref_access): Likewise.
4550 (vect_analyze_data_ref_accesses): Likewise.
4551 (vect_vfa_segment_size): Likewise.
4552 (vect_small_gap_p): Likewise.
4553 (vectorizable_with_step_bound_p): Likewise.
4554 (vect_prune_runtime_alias_test_list): Likewise.
4555 (vect_analyze_data_refs): Likewise.
4556 (vect_supportable_dr_alignment): Likewise.
4557 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4558 (vect_gen_prolog_loop_niters): Likewise.
4559 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4560 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4562 (vect_recog_mask_conversion_pattern): Likewise.
4563 (vect_try_gather_scatter_pattern): Likewise.
4564 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4565 to vect_get_store_cost.
4566 (vect_get_store_cost): Get stmt_info instead of DR.
4567 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4568 (vect_get_load_cost): Get stmt_info instead of DR.
4570 2018-06-01 Richard Biener <rguenther@suse.de>
4573 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4574 vars wrapped in MEM_REFs.
4576 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4578 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4579 Fix subreg tests so that we only return a choice between
4580 GENERAL_REGS and FP_REGS if the original classes included both.
4582 2018-06-01 Richard Biener <rguenther@suse.de>
4585 * tree-ssa-structalias.c (get_function_part_constraint):
4586 Handle NULL fi->decl.
4587 (find_func_aliases_for_call): Properly handle indirect
4588 fi from direct call.
4589 (find_func_clobbers): Likewise.
4590 (ipa_pta_execute): Likewise.
4591 (create_variable_info_for): For functions that are ifunc_resolver
4592 resolve to a varinfo that contains the result of the resolver call.
4593 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4596 2018-05-31 Michael Collison <michael.collison@arm.com>
4598 * config/aarch64/aarch64.md:
4599 (*fix_to_zero_extenddfdi2): New pattern.
4600 * gcc.target/aarch64/fix_extend1.c: New testcase.
4602 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4606 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4607 and BUILT_IN_STRNCMP_EQ.
4608 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4609 BUILT_IN_STRNCMP_EQ.
4610 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4611 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4612 (gimple_fold_builtin): Likewise.
4613 * tree-ssa-strlen.c (compute_string_length): New function.
4614 (determine_min_obsize): New function.
4615 (handle_builtin_string_cmp): New function to handle calls to
4616 string compare functions.
4617 (strlen_optimize_stmt): Add handling to builtin string compare
4619 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4620 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4621 * tree.c (build_common_builtin_nodes): Add new defines of
4622 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4624 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4627 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4628 functions with naked attribute.
4630 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4632 * config/i386/sse.md (avx_vec_concat<mode>):
4633 Substitute concat_tg_mode mode attribute with xtg_mode.
4634 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4635 (concat_tg_mode): Remove mode attribute.
4637 2018-05-31 Martin Sebor <msebor@redhat.com>
4640 * calls.c (alloc_max_size): Correct a logic error/typo.
4641 Treat excessive arguments as infinite. Warn for invalid arguments.
4642 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4644 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4647 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4648 and movx for Haswell.
4650 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4651 Cesar Philippidis <cesar@codesourcery.com>
4654 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4655 when emitting error on private/firstprivate reductions.
4656 * omp-low.c (lower_omp_target): Avoid reference-type processing
4657 on pointers for firstprivate clause.
4659 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4661 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4664 * config/aarch64/aarch64-simd.md
4665 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4666 (aarch64_ld1_x3_<mode>): Likewise
4667 (aarch64_st1x2<VALLDIF:mode>): Likewise
4668 (aarch64_st1_x2_<mode>): Likewise
4669 (aarch64_st1x3<VALLDIF:mode>): Likewise
4670 (aarch64_st1_x3_<mode>): Likewise
4671 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4672 (vld1_s8_x3): Likewise.
4673 (vld1_u16_x3): Likewise.
4674 (vld1_s16_x3): Likewise.
4675 (vld1_u32_x3): Likewise.
4676 (vld1_s32_x3): Likewise.
4677 (vld1_u64_x3): Likewise.
4678 (vld1_s64_x3): Likewise.
4679 (vld1_f16_x3): Likewise.
4680 (vld1_f32_x3): Likewise.
4681 (vld1_f64_x3): Likewise.
4682 (vld1_p8_x3): Likewise.
4683 (vld1_p16_x3): Likewise.
4684 (vld1_p64_x3): Likewise.
4685 (vld1q_u8_x3): Likewise.
4686 (vld1q_s8_x3): Likewise.
4687 (vld1q_u16_x3): Likewise.
4688 (vld1q_s16_x3): Likewise.
4689 (vld1q_u32_x3): Likewise.
4690 (vld1q_s32_x3): Likewise.
4691 (vld1q_u64_x3): Likewise.
4692 (vld1q_s64_x3): Likewise.
4693 (vld1q_f16_x3): Likewise.
4694 (vld1q_f32_x3): Likewise.
4695 (vld1q_f64_x3): Likewise.
4696 (vld1q_p8_x3): Likewise.
4697 (vld1q_p16_x3): Likewise.
4698 (vld1q_p64_x3): Likewise.
4699 (vst1_s64_x2): Likewise.
4700 (vst1_u64_x2): Likewise.
4701 (vst1_f64_x2): Likewise.
4702 (vst1_s8_x2): Likewise.
4703 (vst1_p8_x2): Likewise.
4704 (vst1_s16_x2): Likewise.
4705 (vst1_p16_x2): Likewise.
4706 (vst1_s32_x2): Likewise.
4707 (vst1_u8_x2): Likewise.
4708 (vst1_u16_x2): Likewise.
4709 (vst1_u32_x2): Likewise.
4710 (vst1_f16_x2): Likewise.
4711 (vst1_f32_x2): Likewise.
4712 (vst1_p64_x2): Likewise.
4713 (vst1q_s8_x2): Likewise.
4714 (vst1q_p8_x2): Likewise.
4715 (vst1q_s16_x2): Likewise.
4716 (vst1q_p16_x2): Likewise.
4717 (vst1q_s32_x2): Likewise.
4718 (vst1q_s64_x2): Likewise.
4719 (vst1q_u8_x2): Likewise.
4720 (vst1q_u16_x2): Likewise.
4721 (vst1q_u32_x2): Likewise.
4722 (vst1q_u64_x2): Likewise.
4723 (vst1q_f16_x2): Likewise.
4724 (vst1q_f32_x2): Likewise.
4725 (vst1q_f64_x2): Likewise.
4726 (vst1q_p64_x2): Likewise.
4727 (vst1_s64_x3): Likewise.
4728 (vst1_u64_x3): Likewise.
4729 (vst1_f64_x3): Likewise.
4730 (vst1_s8_x3): Likewise.
4731 (vst1_p8_x3): Likewise.
4732 (vst1_s16_x3): Likewise.
4733 (vst1_p16_x3): Likewise.
4734 (vst1_s32_x3): Likewise.
4735 (vst1_u8_x3): Likewise.
4736 (vst1_u16_x3): Likewise.
4737 (vst1_u32_x3): Likewise.
4738 (vst1_f16_x3): Likewise.
4739 (vst1_f32_x3): Likewise.
4740 (vst1_p64_x3): Likewise.
4741 (vst1q_s8_x3): Likewise.
4742 (vst1q_p8_x3): Likewise.
4743 (vst1q_s16_x3): Likewise.
4744 (vst1q_p16_x3): Likewise.
4745 (vst1q_s32_x3): Likewise.
4746 (vst1q_s64_x3): Likewise.
4747 (vst1q_u8_x3): Likewise.
4748 (vst1q_u16_x3): Likewise.
4749 (vst1q_u32_x3): Likewise.
4750 (vst1q_u64_x3): Likewise.
4751 (vst1q_f16_x3): Likewise.
4752 (vst1q_f32_x3): Likewise.
4753 (vst1q_f64_x3): Likewise.
4754 (vst1q_p64_x3): Likewise.
4756 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4758 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4759 user_label_prefix to name.
4761 * tree-core.h: Update comment about the format of NAME string
4762 passed to handler in attribute_spec.
4764 * config/msp430/msp430.md: Remove erroneous subreg expression from
4765 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4766 zero_extend{q,h}isi2.
4768 2018-05-30 Borislav Petkov <bp@suse.de>
4770 * doc/extend.texi: Document some architecture specific
4771 constraints and sort entries.
4773 2018-05-30 Martin Sebor <msebor@redhat.com>
4776 * builtins.c (expand_builtin_stpcpy_1): New function.
4777 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4778 only if the former succeeds.
4780 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4782 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4785 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4787 * doc/invoke.texi (-flinker-output): Document
4789 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4791 * passes.c (ipa_write_summaries): Only modify statements if body
4793 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4794 incrementally linking.
4795 (ipa_passes): Likewise.
4796 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4797 pass down resolution info.
4798 * common.opt (flag_incremental_link): Update info.
4799 * gcc.c (plugin specs): Turn flinker-output=* to
4800 -plugin-opt=-linker-output-known
4801 * toplev.c (compile_file): Also cut compilation when doing incremental
4803 * flag-types. (enum lto_partition_model): Add
4804 LTO_LINKER_OUTPUT_NOLTOREL.
4805 (invoke.texi): Add -flinker-output docs.
4806 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4807 link same way as WPA; do not stream in dead initializers.
4809 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4810 darf2out_register_external_decl): Support incremental link.
4812 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4814 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4816 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4818 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4819 it down to simple_object_copy_lto_debug_sections.
4820 (run_gcc): Determine incremental LTO link time and configure
4821 lto1 into non-wpa mode, disable renaming of debug sections.
4823 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4825 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4826 descriptions of various incorrectly documented functions.
4828 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4831 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4832 address check not strict.
4834 2018-05-30 Richard Biener <rguenther@suse.de>
4836 PR tree-optimization/85964
4837 * tracer.c (better_p): Drop initialized count check, we only
4838 call the function with initialized counts now.
4839 (find_best_successor): Do find a best edge if one
4840 has uninitialized count.
4841 (find_best_predecessor): Likewise. Do BB frequency check only
4842 if count is initialized.
4844 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4846 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4847 (aarch64_ldrstr_offset_compare): New.
4848 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4849 load/store orderings.
4850 (aarch64_gen_adjusted_ldpstp): Likewise.
4852 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4854 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4855 Check for subset of GENERAL_REGS and FP_REGS.
4856 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4859 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4861 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4862 and wi::to_poly_offset. Add the current offset and then check
4863 whether the sum fits, rather than using an unchecked addition of
4864 a checked term. Check for a shwi rather than a uhwi.
4865 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4866 (store_constructor): Use poly_int_tree_p.
4867 (expand_expr_real_1): Likewise.
4868 * function.c (assign_temp): Likewise.
4869 * fold-const.c (const_binop): Use poly_int_tree_p and
4871 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4873 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4874 to_poly_offset for MEM offsets.
4875 * ipa-icf.c (sem_variable::equals): Likewise.
4876 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4877 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4878 wi::to_poly_offset for BIT_FIELD_REF offsets.
4879 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4881 * var-tracking.c (emit_note_insn_var_location): Use
4882 tree_to_poly_uint64.
4884 2018-05-29 Jim Wilson <jimw@sifive.com>
4886 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4888 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4891 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4892 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4894 (sse4_1_round<mode>2): Use nonimmediate_operand
4895 for operand 1 predicate.
4897 2018-05-29 Martin Sebor <msebor@redhat.com>
4898 Richard Biener <rguenther@suse.de>
4901 * calls.c (get_size_range): Call determine_value_range instead
4902 of get_value_range..
4903 * tree-vrp.h (determine_value_range): Declared new function.
4904 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4906 2018-05-29 Richard Biener <rguenther@suse.de>
4908 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4909 sure to use non-pattern stmts for get_earlier_stmt arguments.
4910 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4911 called on pattern stmts.
4912 (get_later_stmt): Likewise.
4914 2018-05-29 Martin Liska <mliska@suse.cz>
4916 PR gcov-profile/85759
4917 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4920 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4922 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4924 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4928 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4929 VEC_PACK_FLOAT_EXPR): New tree codes.
4930 * tree-pretty-print.c (op_code_prio): Handle
4931 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4932 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4933 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4934 * tree-inline.c (estimate_operator_cost): Likewise.
4935 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4936 * fold-const.c (const_binop): Likewise.
4937 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4938 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4939 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4940 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4941 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4942 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4943 * expr.c (expand_expr_real_2): Likewise.
4944 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4945 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4946 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4948 * optabs.c (expand_widen_pattern_expr): For
4949 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4950 sign from result type rather than operand's type.
4951 (expand_binop_directly): For vec_packu_float_optab and
4952 vec_packs_float_optab allow result type to be different from operand's
4954 * optabs-tree.c (optab_for_tree_code): Handle
4955 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4956 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4957 * tree-vect-generic.c (expand_vector_operations_1): Handle
4958 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4959 VEC_PACK_FLOAT_EXPR.
4960 * tree-vect-stmts.c (supportable_widening_operation): Handle
4962 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4963 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4964 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4965 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4966 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4968 (vec_pack<floatprefix>_float_<mode>): New expander.
4969 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4971 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4972 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4973 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4974 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4975 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4977 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4978 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4979 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4980 VEC_PACK_FLOAT_EXPR): Document.
4982 2018-05-29 Richard Biener <rguenther@suse.de>
4984 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4986 (stmt_vec_info_vec): Make pointer.
4987 (init_stmt_vec_info_vec): Remove.
4988 (free_stmt_vec_info_vec): Likewise.
4989 (set_stmt_vec_info_vec): New function.
4990 (free_stmt_vec_infos): Likewise.
4991 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4992 (set_vinfo_for_stmt): Likewise.
4993 (get_earlier_stmt): Likewise.
4994 (get_later_stmt): Likewise.
4995 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4996 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4997 (vec_info::~vec_info): Free stmt_vec_infos.
4998 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4999 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5000 (pass_slp_vectorize::execute): Likewise.
5001 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5002 (free_stmt_vec_info_vec): Likewise.
5003 (set_stmt_vec_info_vec): New function.
5004 (free_stmt_vec_infos): Likewise.
5005 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5006 the global stmt_vec_info_vec.
5007 * tree-parloops.c (gather_scalar_reductions): Use
5008 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5011 2018-05-29 Richard Biener <rguenther@suse.de>
5013 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5015 2018-05-29 Martin Liska <mliska@suse.cz>
5016 David Malcolm <dmalcolm@redhat.com>
5018 * vec.c (test_reverse): New.
5019 (vec_c_tests): Add new test.
5020 * vec.h (vl_ptr>::reverse): New function.
5022 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5024 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5026 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5029 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5031 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5033 2018-05-28 Richard Biener <rguenther@suse.de>
5035 PR tree-optimization/85933
5036 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5037 look at stmts marked as vectorizable.
5039 2018-05-28 Richard Biener <rguenther@suse.de>
5041 PR tree-optimization/85934
5042 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5043 vector boolean check before scalar optimization.
5045 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5047 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5050 2018-05-28 Mark Wielaard <mark@klomp.org>
5052 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5053 if it is an expression containing a minus sign.
5055 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5057 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5059 2018-05-27 Paul Koning <ni1d@arrl.net>
5061 * config/pdp11/pdp11.md (truncsihi2): Remove.
5063 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5064 Chung-Ju Wu <jasonwucj@gmail.com>
5066 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5068 (unaligned_store_dw): Ditto.
5069 * config/nds32/nds32-memory-manipulation.c
5070 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5071 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5072 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5073 (emit_setmem_word_loop): Rename to ...
5074 (emit_setmem_doubleword_loop): ... this.
5075 (nds32_gen_dup_4_byte_to_word_value): New function.
5076 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5077 (nds32_expand_setmem_loop): Refine implementation.
5078 (nds32_expand_setmem_loop_v3m): Ditto.
5079 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5082 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5084 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5086 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5088 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5089 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5090 machine->attr_no_prologue_p.
5091 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5092 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5093 (nds32_expand_epilogue): Consider attr_naked_p.
5094 (nds32_expand_epilogue_v3pop): Likewise.
5095 (nds32_can_use_return_insn): Likewise.
5096 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5097 attr_no_prologue_p fields.
5098 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5100 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5103 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5105 * config/i386/sse.md
5106 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5108 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5110 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5112 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5114 (*<floatsuffix>floatv2div2sf2): Rename to ...
5115 (*float<floatunssuffix>v2div2sf2): ... this.
5116 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5117 (float<floatunssuffix>v2div2sf2_mask): ... this.
5118 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5119 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5120 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5122 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5124 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5126 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5128 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5130 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5132 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5133 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5134 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5135 gen_ufix_truncv8dfv8si2.
5136 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5137 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5138 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5139 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5140 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5141 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5142 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5143 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5145 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5149 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5151 2018-05-25 Jim Wilson <jimw@sifive.com>
5153 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5154 * config/riscv/riscv.c (struct machine_function): Add
5155 interrupt_handler_p and attribute_checked_p fields.
5156 (riscv_attribute_table): Add interrupt.
5157 (riscv_interrupt_type_p): New.
5158 (riscv_save_reg_p): Save extra regs for interrupt handler.
5159 (riscv_use_save_libcall): Return false for interrupt handler.
5160 (riscv_first_stack_step): Add forward declaration.
5161 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5162 for interrupt handler with large frame. Use it for saved reg list.
5163 (riscv_expand_prologue): Move flag_stack_usage_info support to
5164 eliminate duplication.
5165 (riscv_expand_epilogue): Generate mret for interrupt handler.
5166 (riscv_epilogue_uses): New.
5167 (riscv_can_use_return_insn): Return false for interrupt handler.
5168 (riscv_function_ok_for_sibcall): Likewise.
5169 (riscv_set_current_function): Add interrupt handler support.
5170 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5171 * config/riscv/riscv.md (UNSPECV_MRET): New.
5173 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5175 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5177 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5179 PR tree-optimization/85712
5180 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5181 this candidate has already been replaced in-situ by a copy.
5183 2018-05-25 Jason Merrill <jason@redhat.com>
5185 PR c++/80485 - inline function non-zero address.
5186 * symtab.c (nonzero_address): Check DECL_COMDAT.
5188 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5191 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5192 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5193 extension of SImode operation. Use const123_operand predicate.
5194 (*saddsi_1): Remove.
5195 (*saddl_se_1): Ditto.
5197 (*ssubl_se_1): Ditto.
5198 * config/alpha/predicates.md (const123_operand): New predicate.
5199 * config/alpha/constraints.md (P): Use IN_RANGE.
5201 2018-05-25 Richard Biener <rguenther@suse.de>
5203 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5205 (ref_maybe_used_by_stmt_p): Likewise.
5206 (stmt_may_clobber_ref_p): Likewise.
5207 (stmt_may_clobber_ref_p_1): Likewise.
5208 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5210 (ref_maybe_used_by_stmt_p): Likewise.
5211 (stmt_may_clobber_ref_p): Likewise.
5212 (stmt_may_clobber_ref_p_1): Likewise.
5213 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5214 the alias oracle to disambiguate DRs with stmts DR analysis
5216 (vect_analyze_data_refs): Do not give up on not analyzable
5217 DRs for BB vectorization. Remove code truncating the dataref
5220 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5223 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5224 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5225 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5227 2018-05-25 Richard Biener <rguenther@suse.de>
5229 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5230 function, combining stmt data ref gathering and fatal analysis
5232 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5233 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5234 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5235 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5236 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5238 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5240 PR tree-optimization/85720
5241 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5242 SCC if all partitions are builtins.
5243 (version_loop_by_alias_check): New parameter. Generate cancelable
5244 runtime alias check if all partitions are builtins.
5245 (distribute_loop): Update call to above function.
5247 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5249 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5250 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5251 (parm_default_def_partition_arg): Ditto.
5252 (set_parm_default_def_partition): Ditto.
5253 (get_parm_default_def_partitions): Ditto and make it static.
5254 (get_undefined_value_partitions): Ditto and make it static.
5255 (remove_ssa_form): Refactor call to init_var_map here.
5256 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5257 computation for loop region.
5258 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5259 (register_default_def): Delete.
5260 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5261 (parm_default_def_partition_arg): Ditto.
5262 (set_parm_default_def_partition): Ditto.
5263 (get_parm_default_def_partitions): Ditto and make it static.
5264 (get_undefined_value_partitions): Ditto and make it static.
5265 (coalesce_with_default, coalesce_with_default): Update comment.
5266 (create_coalesce_list_for_region): New func factored out from
5267 create_outofssa_var_map.
5268 (populate_coalesce_list_for_outofssa): New func factored out from
5269 create_outofssa_var_map and coalesce_ssa_name.
5270 (create_outofssa_var_map): Delete.
5271 (coalesce_ssa_name): Refactor to support live range computation.
5272 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5273 (get_parm_default_def_partitions): Delete.
5274 (get_undefined_value_partitions): Ditto.
5275 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5276 computation for loop region.
5277 (new_tree_live_info, loe_visit_block): Ditto.
5278 (live_worklist, set_var_live_on_entry): Ditto.
5279 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5280 * tree-ssa-live.h (struct _var_map): New fields.
5281 (init_var_map): Change decl.
5282 (region_contains_p): New.
5284 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5286 * tree-ssa-live.h (live_merge_and_clear): Delete.
5288 2018-05-25 Richard Biener <rguenther@suse.de>
5291 * tree-dump.c (dequeue_and_dump): Remove access to removed
5292 operand 2 of a SWITCH_EXPR.
5294 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5296 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5297 multiplication and division.
5298 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5299 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5300 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5301 (cond_udiv_optab, cond_umod_optab): New optabs.
5302 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5303 (IFN_COND_RDIV): New internal functions.
5304 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5305 TRUNC_MOD_EXPR and RDIV_EXPR.
5306 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5307 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5309 (SVE_INT_BINARY): Include mult.
5310 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5311 (optab, sve_int_op): Handle mult.
5312 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5314 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5315 for SVE_INT_BINARY_SD.
5317 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5319 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5320 (optab, sve_int_op): Handle div and udiv.
5321 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5322 for SVE_INT_BINARY_SD.
5323 (*<optab><mode>3): New insn for the same.
5325 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5327 * tree-vect-patterns.c: Include predict.h.
5328 (vect_recog_divmod_pattern): Restrict check for division support
5329 to when optimizing for size.
5331 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5333 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5334 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5335 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5336 (gimple_match_op::set_op): Likewise.
5337 (gimple_resimplify4): Declare.
5338 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5339 (expr::gen_transform): Likewise.
5340 (decision_tree::gen): Generate a simplification routine for 4 operands.
5341 * gimple-match-head.c (gimple_simplify): Add an overload for
5342 4 operands. In the top-level function, handle up to 4 call
5343 arguments and call gimple_resimplify4.
5344 (gimple_resimplify4): New function.
5345 (build_call_internal): Pass a fourth operand.
5346 (maybe_push_to_seq): Likewise.
5347 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5348 Fold VEC_COND_EXPRs of an operation and a default value into
5349 an IFN_COND_* function if possible.
5350 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5352 (SVE_COND_FP_BINARY): Include them.
5353 (optab, sve_fp_op): Handle them.
5354 (SVE_INT_BINARY_REV): New code iterator.
5355 (SVE_COND_FP_BINARY_REV): New int iterator.
5356 (commutative): New int attribute.
5357 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5359 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5361 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5362 (*cond_<optab><mode>): New patterns for reversed operands.
5364 2018-05-25 Richard Biener <rguenther@suse.de>
5366 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5367 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5368 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5369 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5370 * tree-vect-data-refs.c (everywhere): Adjust users.
5371 * tree-vect-loop.c (everywhere): Likewise.
5372 * tree-vect-slp.c (everywhere): Likewise.
5373 * tree-vect-stmts.c (everywhere): Likewise.
5374 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5376 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5378 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5380 (gcc_cv_as_section_exclude): ... this.
5381 Try Solaris as #exclude syntax.
5382 * configure: Regenerate.
5383 * config.in: Regenerate.
5384 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5386 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5387 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5389 * varasm.c (default_elf_asm_named_section): Don't check if
5390 HAVE_GAS_SECTION_EXCLUDE is defined.
5392 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5394 * doc/md.texi: Update the documentation of the cond_* optabs
5395 to mention the new final operand. Fix GET_MODE_NUNITS call.
5396 Describe the scalar case too.
5397 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5398 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5400 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5401 (get_conditional_internal_fn): Update comment.
5402 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5403 accumulator value as a final argument to conditional functions.
5404 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5405 a define_expand and add an "else" operand. Assert for now that
5406 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5407 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5408 (*cond_<optab><mode>): New patterns.
5409 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5410 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5411 (UNSPEC_COND_EOR): Delete.
5412 (optab): Remove associated mappings.
5413 (SVE_INT_BINARY): New code iterator.
5414 (sve_int_op): Remove int attribute and add "minus" to the code
5416 (SVE_COND_INT_OP): Delete.
5417 (SVE_COND_FP_OP): Rename to...
5418 (SVE_COND_FP_BINARY): ...this.
5420 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5422 * optabs.c (can_reuse_operands_p): New function.
5423 (maybe_legitimize_operands): Try to reuse the results for
5426 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5428 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5429 Add {q} suffix to insn mnemonic.
5431 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5433 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5434 (msp430_warn_func_return): New.
5436 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5438 * fold-const.c (tree_nonzero_bits): New function.
5439 * fold-const.h (tree_nonzero_bits): Likewise.
5440 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5441 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5443 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5447 * varasm.c (assemble_alias): Check ifunc_resolver only on
5450 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5453 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5454 when memory input operand is handled.
5456 2018-05-24 Luis Machado <luis.machado@linaro.org>
5458 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5460 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5462 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5464 * match.pd: Delay FMA folds until after vectorization.
5466 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5469 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5470 address check not strict.
5472 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5474 * gimple-match.h (gimple_match_op): New class.
5475 (mprts_hook): Replace parameters with a gimple_match_op *.
5476 (maybe_build_generic_op): Likewise.
5477 (gimple_simplified_result_is_gimple_val): Replace parameters with
5478 a const gimple_match_op *.
5479 (gimple_simplify): Replace code_helper * and tree * parameters with
5480 a gimple_match_op * parameter.
5481 (gimple_resimplify1): Replace code_helper *, tree and tree *
5482 parameters with a gimple_match_op * parameter.
5483 (gimple_resimplify2): Likewise.
5484 (gimple_resimplify3): Likewise.
5485 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5486 parameters with a gimple_match_op * parameter.
5487 * gimple-match-head.c (gimple_simplify): Change prototypes of
5488 auto-generated functions to take a gimple_match_op * instead of
5489 separate code_helper * and tree * parameters. Make the same
5490 change in the top-level overload and update calls to the
5491 gimple_resimplify routines. Update calls to the auto-generated
5492 functions and to maybe_push_res_to_seq in the publicly-facing
5493 operation-specific gimple_simplify overloads.
5494 (gimple_match_op::MAX_NUM_OPS): Define.
5495 (gimple_resimplify1): Replace rcode and ops with a single res_op
5496 parameter. Update call to gimple_simplify.
5497 (gimple_resimplify2): Likewise.
5498 (gimple_resimplify3): Likewise.
5499 (mprts_hook): Replace parameters with a gimple_match_op *.
5500 (maybe_build_generic_op): Likewise.
5501 (build_call_internal): Replace type, nargs and ops with
5502 a gimple_match_op *.
5503 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5504 with a single gimple_match_op *. Update calls to mprts_hook,
5505 build_call_internal and gimple_simplified_result_is_gimple_val.
5506 Factor out code that is common to the tree_code and combined_fn cases.
5507 * genmatch.c (expr::gen_transform): Replace tem_code and
5508 tem_ops with a gimple_match_op called tem_op. Update calls
5509 to the gimple_resimplify functions and maybe_push_res_to_seq.
5510 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5511 res_ops. Update call to the gimple_resimplify functions.
5512 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5513 (decision_tree::gen): Make the functions take a gimple_match_op *
5514 called res_op instead of separate res_code and res_ops parameters.
5515 Update call accordingly.
5516 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5517 and ops with a single res_op parameter. Update calls to
5518 maybe_build_generic_op and maybe_push_res_to_seq.
5519 (fold_stmt_1): Update calls to gimple_simplify and
5520 replace_stmt_with_simplification.
5521 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5522 and gimple_simplified_result_is_gimple_val.
5523 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5525 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5526 with a gimple_match_op *.
5527 (vn_nary_build_or_lookup): Likewise. Update call to
5528 vn_nary_build_or_lookup_1.
5529 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5530 gimple_match_op *. Update calls to the gimple_resimplify routines
5531 and to gimple_simplified_result_is_gimple_val.
5532 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5533 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5534 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5535 (visit_nary_op): Likewise.
5536 (visit_reference_op_load): Likewise.
5538 2018-05-23 Luis Machado <luis.machado@linaro.org>
5540 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5541 modifier for printing the step amount.
5543 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5546 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5549 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5551 * doc/sourcebuild.texi (Endianness): New subsubsection.
5553 2018-05-23 Luis Machado <luis.machado@linaro.org>
5555 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5556 <prefetch_dynamic_strides>: New const bool field.
5557 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5558 prefetch_dynamic_strides.
5559 (exynosm1_prefetch_tune): Likewise.
5560 (thunderxt88_prefetch_tune): Likewise.
5561 (thunderx_prefetch_tune): Likewise.
5562 (thunderx2t99_prefetch_tune): Likewise.
5563 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5565 (aarch64_override_options_internal): Update to set
5566 PARAM_PREFETCH_DYNAMIC_STRIDES.
5567 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5568 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5569 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5570 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5571 prefetch-dynamic-strides setting.
5573 2018-05-23 Luis Machado <luis.machado@linaro.org>
5575 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5576 <minimum_stride>: New const int field.
5577 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5578 minimum_stride field defaulting to -1.
5579 (exynosm1_prefetch_tune): Likewise.
5580 (thunderxt88_prefetch_tune): Likewise.
5581 (thunderx_prefetch_tune): Likewise.
5582 (thunderx2t99_prefetch_tune): Likewise.
5583 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5584 <default_opt_level>: Set to 3.
5585 (aarch64_override_options_internal): Update to set
5586 PARAM_PREFETCH_MINIMUM_STRIDE.
5587 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5588 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5589 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5590 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5591 stride is constant and is below the minimum stride threshold.
5593 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5595 * config/arm/arm-cpus.in (mode26): Delete.
5596 (armv4): Delete mode26 reference.
5597 * config/arm/arm.c (arm_configure_build_target): Delete use of
5600 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5602 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5604 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5605 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5607 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5608 (floatunsdidf2): Ditto.
5610 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5612 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5613 (fixuns_trunc<mode>si2_avx512f): Ditto.
5614 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5615 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5616 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5618 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5620 PR rtl-optimization/79985
5621 * df-scan.c (df_insn_refs_collect): Remove special case for
5622 global registers and asm statements.
5624 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5626 * extend.texi (Global Register Variables): Rewrite the bullet list.
5627 Note that the register is available for allocation. Note that access
5628 via inline asm must use constraints. Add note about async-signal
5629 handlers. Remove paragraph about automagic register selection.
5631 2018-05-23 Richard Biener <rguenther@suse.de>
5633 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5634 of fixed offset from memset VN.
5636 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5638 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5640 (alloc_cand_and_find_basis): Initialize first_interp field.
5641 (slsr_process_mul): Modify first_interp field.
5642 (slsr_process_add): Likewise.
5643 (slsr_process_cast): Modify first_interp field for each new
5645 (slsr_process_copy): Likewise.
5646 (dump_candidate): Dump first_interp field.
5647 (replace_mult_candidate): Process all interpretations, not just
5649 (replace_rhs_if_not_dup): Likewise.
5650 (replace_one_candidate): Likewise.
5652 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5654 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5656 (aarch64_needs_frame_chain): New function.
5657 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5659 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5662 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5663 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5664 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5665 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5666 as true for strict-align.
5667 (aarch64_can_inline_p): Perform checks even when callee has no
5668 attributes to check for strict alignment.
5669 * doc/extend.texi (AArch64 Function Attributes): Document
5671 * doc/invoke.texi: (AArch64 Options): Likewise.
5673 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5675 PR tree-optimization/85853
5676 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5677 the handling of the root of the node to...
5678 (vect_slp_analyze_node_operations_1): ...this new function,
5679 and run the whole thing with the child nodes' def types
5680 set according to their SLP node's def type.
5682 2018-05-23 Richard Biener <rguenther@suse.de>
5685 * tree-data-ref.c (create_runtime_alias_checks): Defer
5686 and ignore overflow warnings.
5688 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5690 PR tree-optimization/85822
5691 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5694 2018-05-23 Richard Biener <rguenther@suse.de>
5696 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5697 memset constants via native_interpret_expr.
5699 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5702 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5704 (cgraph_node::create_alias): Likewise.
5705 (cgraph_node::get_availability): Check ifunc_resolver instead
5706 of looking up ifunc attribute.
5707 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5708 * varasm.c (do_assemble_alias): Likewise.
5709 (assemble_alias): Likewise.
5710 (default_binds_local_p_3): Likewise.
5711 * cgraph.h (cgraph_node): Add ifunc_resolver.
5712 (cgraph_node::only_called_directly_or_aliased_p): Return false
5714 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5716 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5717 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5718 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5719 instead of looking up ifunc attribute.
5721 2018-05-22 Luis Machado <luis.machado@linaro.org>
5723 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5725 2018-05-22 Martin Sebor <msebor@redhat.com>
5728 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5729 only when expasion succeeds.
5730 (expand_builtin_strcmp): Same.
5731 (expand_builtin_strncmp): Same.
5733 2018-05-22 Martin Sebor <msebor@redhat.com>
5735 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5737 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5738 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5740 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5741 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5743 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5744 Add check for legitimate_address.
5745 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5746 (aarch64_swap_ldrstr_operands): New.
5747 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5750 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5751 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5753 * config/aarch64/aarch64.md: New patterns to generate stp
5755 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5756 single words and double words.
5757 (load_pair_sw, load_pair_dw): Likewise.
5758 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5760 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5762 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5763 for different mode ldpstp and add peephole for merged zero stores.
5765 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5767 (aarch64_gen_store_pair): Rename calls to match new patterns.
5768 (aarch64_gen_load_pair): Rename calls to match new patterns.
5769 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5770 (load_pair<DREG:mode><DREG2:mode>): ... This.
5771 (store_pair<mode>): Rename to...
5772 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5773 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5775 (V_INT_EQUIV): Handle SImode.
5776 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5779 2018-05-22 Martin Sebor <msebor@redhat.com>
5782 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5783 or ajust the presumed bound on an operation to avoid unnecessary
5786 2018-05-22 Martin Sebor <msebor@redhat.com>
5788 PR tree-optimization/85826
5789 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5790 assuming that a DECL necesarily has a constant size.
5792 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5795 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5797 2018-05-22 Richard Biener <rguenther@suse.de>
5799 PR tree-optimization/85834
5800 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5801 non-constant and non-zero memset arguments.
5803 2018-05-22 Martin Liska <mliska@suse.cz>
5806 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5808 2018-05-22 Richard Biener <rguenther@suse.de>
5810 PR tree-optimization/85863
5811 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5812 comparisons when vectype is specified.
5813 (vectorizable_condition): Do not specify vectype for
5814 vect_is_simple_cond when SLP vectorizing.
5816 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5819 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5820 define __ibm128 as long double.
5821 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5822 as a distinct type when IEEE 128-bit support is enabled.
5823 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5824 128-bit types to use the correct functions.
5825 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5826 convert between 128-bit floating point types that have different
5827 modes but the same representation, instead of using gen_lowpart to
5829 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5831 (IFKF_reg): New attributes to give the register constraints for
5833 (extend<mode>tf2_internal): New insns to mark an explicit
5834 conversion between 128-bit floating point types that have a
5835 different mode but share the same representation.
5837 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5839 PR tree-optimization/85814
5840 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5841 a null return from get_strinfo when unsharing the next
5842 strinfo in the chain.
5844 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5847 * varasm.c (weak_finish): Clean up weak_decls.
5849 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5851 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5852 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5853 UNSPEC_UADALP values.
5854 * config/aarch64/iterators.md (ABAL): New int iterator.
5857 (sur): Add mappings for the above.
5858 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5860 (aarch64_<sur>abal<mode>_4): Likewise.
5861 (aarch64_<sur>adalp<mode>_3): Likewise.
5862 (<sur>sadv16qi): New define_expand.
5864 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5866 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5867 (*movdf_internal): Ditto.
5868 (*rcpsf2_sse): Ditto.
5869 (*rsqrtsf2_sse): Ditto.
5870 (*sqrt<mode>2_sse): Ditto.
5872 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5874 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5876 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5877 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5878 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5879 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5881 * config/aarch64/arm_neon.h: Likewise.
5882 * config/aarch64/iterators.md (VQ_I): New.
5884 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5886 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5888 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5891 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5893 * config/nds32/constraints.md (S): New constraint.
5894 * config/nds32/nds32.md (call_internal): Use constraint S.
5895 (call_value_internal): Likewise.
5896 (sibcall_internal): Likewise.
5897 (sibcall_value_internal): Likewise.
5899 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5900 Chung-Ju Wu <jasonwucj@gmail.com>
5902 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5905 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5906 Chung-Ju Wu <jasonwucj@gmail.com>
5908 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5909 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5910 (nds32_rtx_costs_impl): Simplify.
5911 (nds32_address_cost_impl): Simplify.
5912 (nds32_init_rtx_costs): New function.
5913 (nds32_rtx_costs_speed_prefer): Likewise.
5914 (nds32_rtx_costs_size_prefer): Likewise.
5915 (nds32_address_cost_speed_prefer): Likewise.
5916 (nds32_address_cost_speed_fwprop): Likewise.
5917 (nds32_address_cost_size_prefer): Likewise.
5918 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5919 * config/nds32/nds32.c (nds32_option_override): Use
5920 nds32_init_rtx_costs function.
5922 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5924 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5925 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5926 (TARGET_PIPELINE_N8): Likewise.
5927 (TARGET_PIPELINE_N10): Likewise.
5928 (TARGET_PIPELINE_N13): Likewise.
5929 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5931 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5933 * config/nds32/nds32-fpu.md: Update copyright year.
5935 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5937 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5939 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5941 * config/nds32/nds32.c
5942 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5943 * config/nds32/nds32.opt (minline-asm-r15): New option.
5945 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5947 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5949 * config/nds32/nds32.md (abssi2): New pattern.
5951 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5953 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5954 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5955 Merge insn pattern from sse_cvtsi2ss<round_name> and
5956 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5957 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5958 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5959 using SWI48 mode iterator.
5960 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5961 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5962 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5963 pattern from sse_cvttss2si<round_saeonly_name>
5964 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5965 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5966 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5967 using SWI48 mode iterator.
5968 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5969 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5970 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5971 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5972 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5973 using SWI48 mode iterator.
5974 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5975 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5976 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5977 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5978 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5979 SWI48 mode iterator.
5980 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5981 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5982 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5983 pattern from sse_cvttsd2si<round_saeonly_name>
5984 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5986 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5988 * config/nds32/nds32-md-auxiliary.c
5989 (nds32_valid_smw_lwm_base_p): Refine.
5990 (nds32_output_smw_single_word): Refine.
5991 (nds32_output_smw_double_word): New.
5992 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5994 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5996 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5997 (nds32_output_stack_pop): Refine.
5998 (nds32_expand_unaligned_load): Refine.
5999 (nds32_expand_unaligned_store): Refine.
6001 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6002 Chung-Ju Wu <jasonwucj@gmail.com>
6004 * config/nds32/constants.md: Add TP_REGNUM constant.
6005 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6006 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6008 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6009 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6010 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6011 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6012 and PIC code generation.
6013 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6015 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6017 * config/nds32/nds32.md: Support TLS and PIC.
6018 * config/nds32/nds32.c: Support TLS and PIC.
6019 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6020 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6023 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6025 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6026 mode with E_ prefix.
6028 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6029 Chung-Ju Wu <jasonwucj@gmail.com>
6031 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6032 * config/nds32/nds32-md-auxiliary.c
6033 (symbolic_reference_mentioned_p): New.
6034 (nds32_legitimize_ict_address): New.
6035 (nds32_expand_ict_move): New.
6036 (nds32_indirect_call_referenced_p): New.
6037 (nds32_symbol_binds_local_p): Delete.
6038 (nds32_long_call_p): Modify.
6039 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6040 * config/nds32/nds32-protos.h
6041 (symbolic_reference_mentioned_p): Declare.
6042 (nds32_legitimize_ict_address): Declare.
6043 (nds32_expand_ict_move): Declare.
6044 (nds32_indirect_call_referenced_p): Declare.
6045 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6046 (nds32_relax_group): Use nds32_ict_const_p as condition.
6047 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6048 (nds32_asm_file_start): Output ict_model directive in asm code.
6049 (nds32_legitimate_address_p): Consider indirect call.
6050 (nds32_print_operand): Consider indirect call.
6051 (nds32_print_operand_address): Consider indirect call.
6052 (nds32_insert_attributes): Handle "indirect_call" attribute.
6053 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6054 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6055 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6056 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6057 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6058 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6059 (TARGET_ICT_MODEL_SMALL): Define.
6060 (TARGET_ICT_MODEL_LARGE): Define.
6061 * config/nds32/nds32.md (movsi): Consider ict model.
6062 (call, call_value): Consider ict model.
6063 (sibcall, sibcall_value): Consider ict model.
6064 * config/nds32/nds32.opt (mict-model): New option.
6065 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6068 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6069 Monk Chiang <sh.chiang04@gmail.com>
6070 Jim Wilson <jimw@sifive.com>
6072 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6073 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6074 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6075 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6076 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6077 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6078 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6079 compute save_libcall_adjustment properly.
6080 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6081 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6082 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6083 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6084 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6085 (ABI_SPEC): Handle mabi=ilp32e.
6086 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6087 (RVE): Add RVE mask.
6088 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6089 <-march>: Add rv32e as an example.
6091 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6094 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6095 (intra_create_variable_infos): Handle C++ constructors.
6097 2018-05-18 Martin Liska <mliska@suse.cz>
6099 * passes.def: Remove a redundant pass.
6101 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6104 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6106 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6108 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6110 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6112 (armv2, armv2a, armv3, armv3m): Delete architectures.
6113 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6114 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6115 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6117 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6118 (*mulsidi3adddi): Likewise.
6119 (mulsidi3): Likewise.
6120 (*mulsidi3_nov6): Likewise.
6121 (umulsidi3): Likewise.
6122 (umulsidi3_nov6): Likewise.
6123 (umaddsidi4): Likewise.
6124 (*umulsidi3adddi): Likewise.
6125 (smulsi3_highpart): Likewise.
6126 (*smulsi3_highpart_nov6): Likewise.
6127 (umulsi3_highpart): Likewise.
6128 (*umulsi3_highpart_nov6): Likewise.
6129 * config/arm/arm.h (arm_arch3m): Delete.
6130 * config/arm/arm.c (arm_arch3m): Delete.
6131 (arm_option_override_internal): Update armv3-related comment.
6132 (arm_configure_build_target): Delete use of isa_bit_mode32.
6133 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6134 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6135 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6137 (mulusa3): Likewise.
6138 * config/arm/arm-protos.h (arm_arch3m): Delete.
6139 * config/arm/arm-tables.opt: Regenerate.
6140 * config/arm/arm-tune.md: Likewise.
6141 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6142 deleted architectures.
6144 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6146 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6147 (armv5t, armv5te): New features.
6148 (ARMv5, ARMv5e): Delete fgroups.
6149 (ARMv5t, ARMv5te): Adjust for above changes.
6151 (armv5, armv5e): Delete arches.
6152 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6154 (*call_reg_arm): Likewise.
6155 (*call_value_reg_armv5): Likewise.
6156 (*call_value_reg_arm): Likewise.
6157 (*call_symbol): Likewise.
6158 (*call_value_symbol): Likewise.
6159 (*sibcall_insn): Likewise.
6160 (*sibcall_value_insn): Likewise.
6162 (prefetch): Likewise.
6163 (define_split and define_peephole2 dependent on arm_arch5):
6165 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6167 (TARGET_ARM_QBIT): Likewise.
6168 (TARGET_DSP_MULTIPLY): Likewise.
6169 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6170 (arm_arch5, arm_arch5e): Delete.
6171 (arm_arch5t, arm_arch5te): Declare.
6172 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6173 (arm_arch5t): Declare.
6174 (arm_option_reconfigure_globals): Update for the above.
6175 (arm_options_perform_arch_sanity_checks): Update comment, replace
6176 use of arm_arch5 with arm_arch5t.
6177 (use_return_insn): Likewise.
6178 (arm_emit_call_insn): Likewise.
6179 (output_return_instruction): Likewise.
6180 (arm_final_prescan_insn): Likewise.
6181 (arm_coproc_builtin_available): Likewise.
6182 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6183 arm_arch5e with arm_arch5t and arm_arch5te.
6184 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6185 (arm_arch5t, arm_arch5te): Declare.
6186 * config/arm/arm-tables.opt: Regenerate.
6187 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6188 * config/arm/t-multilib: Likewise.
6189 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6190 instead of arm_arch5.
6191 (*call_reg_thumb1): Likewise.
6192 (*call_value_reg_thumb1_v5): Likewise.
6193 (*call_value_reg_thumb1): Likewise.
6194 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6196 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6198 2018-05-18 Martin Liska <mliska@suse.cz>
6200 PR gcov-profile/84846
6201 * doc/gcov.texi: Document -t option of gcov tool.
6203 2018-05-18 Martin Liska <mliska@suse.cz>
6205 PR gcov-profile/84846
6206 * gcov.c (print_usage): Add new -t option.
6207 (process_args): Handle the option.
6208 (generate_results): Use stdout as output when requested by
6211 2018-05-18 Martin Liska <mliska@suse.cz>
6213 PR gcov-profile/84846
6214 * coverage.c (coverage_init): Write PWD to .gcno file.
6215 * doc/gcov.texi: Document how working directory is printed.
6216 * gcov-dump.c (dump_gcov_file): Print PWD.
6217 * gcov.c (output_intermediate_file): Likewise.
6218 (read_graph_file): Read PWD string.
6219 (output_lines): Print PWD.
6221 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6224 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6225 for retval and return false if all args to phi are zero.
6227 2018-05-18 Richard Biener <rguenther@suse.de>
6229 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6231 (evrp_dom_walker::before_dom_children): Call it.
6233 2018-05-18 Richard Biener <rguenther@suse.de>
6235 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6236 results when processing array refs with variable index.
6238 2018-05-18 Toon Moene <toon@moene.org>
6240 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6241 directly after that of -floop-interchange. Indicate that both
6242 options are enabled by default when specifying -O3.
6244 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6246 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6247 iterator. Delete separate integer-mode vec_set<mode> expander.
6248 (aarch64_simd_vec_setv2di): Delete.
6249 (vec_setv2di): Delete.
6250 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6251 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6252 the "w, r" alternative.
6254 2018-05-18 Martin Liska <mliska@suse.cz>
6256 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6257 * tree-pass.h (make_pass_lower_switch_O0): New function.
6258 * tree-switch-conversion.c (node_has_low_bound): Remove.
6259 (node_has_high_bound): Likewise.
6260 (node_is_bounded): Likewise.
6261 (class pass_lower_switch): Make it a template type and create
6263 (pass_lower_switch::execute): Add template argument.
6264 (make_pass_lower_switch): New function.
6265 (make_pass_lower_switch_O0): New function.
6266 (do_jump_if_equal): Remove.
6267 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6268 all the hard work to tree optimization passes.
6270 2018-05-18 Martin Liska <mliska@suse.cz>
6272 * dbgcnt.c (limit_low): Renamed from limit.
6273 (limit_high): New variable.
6274 (dbg_cnt_is_enabled): Check for upper limit.
6275 (dbg_cnt): Adjust dumping.
6276 (dbg_cnt_set_limit_by_index): Add new argument for high
6278 (dbg_cnt_set_limit_by_name): Likewise.
6279 (dbg_cnt_process_single_pair): Parse new format.
6280 (dbg_cnt_process_opt): Use strtok.
6281 (dbg_cnt_list_all_counters): Remove 'value' and add
6283 * doc/invoke.texi: Document changes.
6285 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6287 * doc/sourcebuild.texi (scalar_all_fma): Document.
6288 * tree.def (FMA_EXPR): Delete.
6289 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6290 * internal-fn.c (ternary_direct): New macro.
6291 (expand_ternary_optab_fn): Likewise.
6292 (direct_ternary_optab_supported_p): Likewise.
6293 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6294 * builtins.c (fold_builtin_fma): Delete.
6295 (fold_builtin_3): Don't call it.
6296 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6297 * expr.c (expand_expr_real_2): Likewise.
6298 * fold-const.c (operand_equal_p): Likewise.
6299 (fold_ternary_loc): Likewise.
6300 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6301 * gimple.c (DEFTREECODE): Likewise.
6302 * gimplify.c (gimplify_expr): Likewise.
6303 * optabs-tree.c (optab_for_tree_code): Likewise.
6304 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6305 * tree-eh.c (operation_could_trap_p): Likewise.
6306 (stmt_could_throw_1_p): Likewise.
6307 * tree-inline.c (estimate_operator_cost): Likewise.
6308 * tree-pretty-print.c (dump_generic_node): Likewise.
6309 (op_code_prio): Likewise.
6310 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6311 * tree-ssa-operands.c (get_expr_operands): Likewise.
6312 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6313 * fold-const-call.h (fold_fma): Delete.
6314 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6315 CFN_FNMA and CFN_FNMS.
6317 * genmatch.c (combined_fn): New enum.
6318 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6319 (commutative_op): New function.
6320 (commutate): Use it. Handle more than 2 operands.
6321 (dt_operand::gen_gimple_expr): Use commutative_op.
6322 (parser::parse_expr): Allow :c to be used with non-binary
6323 operators if the commutative operand is known.
6324 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6325 CFN_FMS, CFN_FNMA and CFN_FNMS.
6326 (backprop::process_assign_use): Remove FMA_EXPR handling.
6327 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6328 (gen_hsa_fma): New function.
6329 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6330 IFN_FNMA and IFN_FNMS.
6331 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6332 * gimple-fold.h (follow_all_ssa_edges): Declare.
6333 * gimple-fold.c (follow_all_ssa_edges): New function.
6334 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6335 gimple_build interface and use follow_all_ssa_edges to fold the result.
6336 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6337 instead of checking for optabs directly.
6338 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6339 rather than FMA_EXPRs.
6340 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6341 call to IFN_FMA instead of an FMA_EXPR.
6343 2018-05-17 Jim Wilson <jimw@sifive.com>
6345 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6346 sign extended promoted subreg, and the range does not have the sign bit
6347 set, then do a sign extend.
6349 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6350 test, check for sign extended subreg and/or constant operands, and
6351 do a sign extend in that case.
6353 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6355 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6356 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6358 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6359 Change logics_shift_reg to logics_shift_imm.
6360 (thunderx2t99_fp_loadpair_basic): Delete.
6361 (thunderx2t99_fp_storepair_basic): Delete.
6362 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6363 (thunderx2t99_asimd_polynomial): Delete.
6364 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6365 and neon_fp_mul_d_scalar_q.
6366 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6367 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6368 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6369 (thunderx2t99_asimd_lut): Add missing tbl types.
6370 (thunderx2t99_asimd_ext): Delete.
6371 (thunderx2t99_asimd_load1_1_mult): Delete.
6372 (thunderx2t99_asimd_load1_2_mult): Delete.
6373 (thunderx2t99_asimd_load1_ldp): New.
6374 (thunderx2t99_asimd_load1): New.
6375 (thunderx2t99_asimd_load2): Add missing *load2* types.
6376 (thunderx2t99_asimd_load3): New.
6377 (thunderx2t99_asimd_load4): New.
6378 (thunderx2t99_asimd_store1_1_mult): Delete.
6379 (thunderx2t99_asimd_store1_2_mult): Delete.
6380 (thunderx2t99_asimd_store2_mult): Delete.
6381 (thunderx2t99_asimd_store2_onelane): Delete.
6382 (thunderx2t99_asimd_store_stp): New.
6383 (thunderx2t99_asimd_store1): New.
6384 (thunderx2t99_asimd_store2): New.
6385 (thunderx2t99_asimd_store3): New.
6386 (thunderx2t99_asimd_store4): New.
6388 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6390 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6391 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6393 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6394 Segher Boessenkool <segher@kernel.crashing.org>
6397 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6400 2018-05-17 Richard Biener <rguenther@suse.de>
6402 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6403 for pruning loop and prune defs feeding only already visited PHIs.
6405 2018-05-17 Richard Biener <rguenther@suse.de>
6407 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6409 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6410 Richard Biener <rguenther@suse.de>
6412 PR tree-optimization/85793
6413 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6414 for VMAT_ELEMENTWISE.
6416 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6418 * internal-fn.h (lookup_internal_fn): Declare
6419 * internal-fn.c (lookup_internal_fn): New function.
6420 * gimple.c (gimple_build_call_from_tree): Handle calls to
6422 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6423 internal function names.
6424 * tree-pretty-print.c (dump_generic_node): Likewise.
6425 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6427 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6429 * gimple-fold.h (gimple_build): Make the function forms take
6430 combined_fn rather than built_in_function.
6431 (gimple_simplify): Likewise.
6432 * gimple-match-head.c (gimple_simplify): Likewise.
6433 * gimple-fold.c (gimple_build): Likewise.
6434 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6435 rather than gimple_build_call_internal.
6436 (get_initial_defs_for_reduction): Likewise.
6437 (vect_create_epilog_for_reduction): Likewise.
6438 (vectorizable_live_operation): Likewise.
6440 2018-05-17 Martin Liska <mliska@suse.cz>
6442 * gimple-ssa-sprintf.c (format_directive): Do not use
6443 space in between 'G_' and '('.
6445 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6448 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6449 even if the mask is not all ones.
6452 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6454 (ix86_gimple_fold_builtin): Likewise.
6457 * config/i386/i386.c: Include tree-vector-builder.h.
6458 (ix86_vector_shift_count): New function.
6459 (ix86_fold_builtin): Fold shift builtins by scalar count.
6460 (ix86_gimple_fold_builtin): Likewise.
6462 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6463 _mm512_setzero): New intrinsics.
6465 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6466 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6468 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6469 code generation for cases where splatting a value is not useful.
6470 * simplify-rtx.c (simplify_ternary_operation): Simplify
6471 vec_merge across a vec_duplicate and a paradoxical subreg forming
6472 a vector mode to a vec_concat.
6474 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6476 * config.gcc: Support "goldmont-plus".
6477 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6479 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6480 PROCESSOR_GOLDMONT_PLUS.
6481 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6482 (processor_target_table): Add "goldmont-plus".
6483 (PTA_GOLDMONT_PLUS): Define.
6484 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6485 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6486 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6487 (fold_builtin_cpu): Add "goldmont-plus".
6488 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6489 (ix86_option_override_internal): Add "goldmont-plus".
6490 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6491 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6492 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6493 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6495 2018-05-17 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/85757
6498 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6499 remove defs that only feed that PHI from further processing.
6501 2018-05-16 Jim Wilson <jimw@sifive.com>
6503 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6505 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6507 2018-05-16 Mark Wielaard <mark@klomp.org>
6509 * dwarf2out.c (count_index_strings): New function.
6510 (output_indirect_strings): Call count_index_strings and generate
6511 header for dwarf_version >= 5.
6513 2018-05-16 Mark Wielaard <mark@klomp.org>
6515 * dwarf2out.c (dwarf_FORM): New function.
6516 (set_indirect_string): Use dwarf_FORM.
6517 (reset_indirect_string): Likewise.
6518 (size_of_die): Likewise.
6519 (value_format): Likewise.
6520 (output_die): Likewise.
6521 (add_skeleton_AT_string): Likewise.
6522 (output_macinfo_op): Likewise.
6523 (index_string): Likewise.
6524 (output_index_string_offset): Likewise.
6525 (output_index_string): Likewise.
6526 (count_index_strings): Likewise.
6528 2018-05-16 Carl Love <cel@us.ibm.com>
6530 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6531 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6533 2018-05-16 Martin Jambor <mjambor@suse.cz>
6535 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6536 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6538 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6540 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6541 (fnma<mode>4): Likewise.
6542 (fms<mode>4): Likewise.
6543 (fnms<mode>4): Likewise.
6544 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6545 (aarch64_fnma<mode>4): Likewise.
6546 (aarch64_fms<mode>4): Likewise.
6547 (aarch64_fnms<mode>4): Likewise.
6548 (aarch64_fnmadd<mode>4): Likewise.
6550 2018-05-16 Jason Merrill <jason@redhat.com>
6552 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6554 2018-05-16 Richard Biener <rguenther@suse.de>
6556 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6557 (dump_stmt_cost): Declare.
6558 (add_stmt_cost): Dump cost we add.
6559 (add_stmt_costs): New function.
6560 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6562 (vect_analyze_stmt): Adjust prototype.
6563 (vectorizable_condition): Likewise.
6564 (vectorizable_live_operation): Likewise.
6565 (vectorizable_reduction): Likewise.
6566 (vectorizable_induction): Likewise.
6567 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6568 cost vector to pass to vectorizable_ and record afterwards.
6569 (vect_model_reduction_cost): Take cost vector argument and adjust.
6570 (vect_model_induction_cost): Likewise.
6571 (vectorizable_reduction): Likewise.
6572 (vectorizable_induction): Likewise.
6573 (vectorizable_live_operation): Likewise.
6574 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6575 SLP_TREE_NUMBER_OF_VEC_STMTS.
6576 (vect_analyze_slp_cost_1): Remove.
6577 (vect_analyze_slp_cost): Likewise.
6578 (vect_slp_analyze_node_operations): Take visited args and
6579 a target cost vector. Avoid processing already visited stmt sets.
6580 (vect_slp_analyze_operations): Use a local cost vector to gather
6581 costs and register those of non-discarded instances.
6582 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6583 (vect_schedule_slp_instance): Remove copying of
6584 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6586 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6587 adding cost. Record cost entry location.
6588 (vect_prologue_cost_for_slp_op): Function to compute cost of
6589 a constant or invariant generated for SLP vect in the prologue,
6590 split out from vect_analyze_slp_cost_1.
6591 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6592 (vect_model_promotion_demotion_cost): Likewise.
6593 (vect_model_store_cost): Likewise, make static.
6594 (vect_model_load_cost): Likewise.
6595 (vectorizable_bswap): Add cost vector arg and adjust.
6596 (vectorizable_call): Likewise.
6597 (vectorizable_simd_clone_call): Likewise.
6598 (vectorizable_conversion): Likewise.
6599 (vectorizable_assignment): Likewise.
6600 (vectorizable_shift): Likewise.
6601 (vectorizable_operation): Likewise.
6602 (vectorizable_store): Likewise.
6603 (vectorizable_load): Likewise.
6604 (vectorizable_condition): Likewise.
6605 (vectorizable_comparison): Likewise.
6606 (can_vectorize_live_stmts): Likewise.
6607 (vect_analyze_stmt): Likewise.
6608 (vect_transform_stmt): Adjust calls to vectorizable_*.
6609 * tree-vectorizer.c: Include gimple-pretty-print.h.
6610 (dump_stmt_cost): New function.
6612 2018-05-16 Richard Biener <rguenther@suse.de>
6614 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6615 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6616 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6617 (check_name): New callback.
6618 (dse_classify_store): Track cycles via a visited bitmap of PHI
6619 defs and simplify handling of in-loop and across loop dead stores
6620 and properly fail for loop-variant refs. Handle byte-tracking with
6621 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6624 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6626 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6627 (vect_get_mask_type_for_stmt): Likewise.
6628 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6630 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6631 to determine the statement's vector type and the vector type that
6632 should be used for calculating nunits. Deal with cases in which
6633 the type has to be deferred.
6634 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6635 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6636 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6637 (vect_determine_vf_for_stmt): New functions, split out from...
6638 (vect_determine_vectorization_factor): ...here.
6639 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6640 (vect_get_mask_type_for_stmt): New functions, split out from
6641 vect_determine_vectorization_factor.
6643 2018-05-16 Richard Biener <rguenther@suse.de>
6645 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6646 verify the [VEC_]COND_EXPR embedded comparison.
6648 2018-05-15 Martin Sebor <msebor@redhat.com>
6650 PR tree-optimization/85753
6651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6652 RECORD_TYPE in addition to ARRAY_TYPE.
6654 2018-05-15 Martin Sebor <msebor@redhat.com>
6657 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6659 2018-05-15 Richard Biener <rguenther@suse.de>
6661 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6662 add by_clobber_p one. Change algorithm to collect all defs
6663 representing uses we need to walk and try reducing them to
6664 a single one before failing.
6665 (dse_dom_walker::dse_optimize_stmt): Adjust.
6667 2018-05-13 Mark Wielaard <mark@klomp.org>
6669 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6670 (size_of_loc_descr): Likewise.
6671 (output_loc_operands): Likewise.
6672 (output_loc_operands_raw): Likewise.
6673 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6674 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6675 (hash_loc_operands): Likewise.
6676 (compare_loc_operands): Likewise.
6678 2018-05-14 Mark Wielaard <mark@klomp.org>
6680 * dwarf2out.c (count_index_addrs): New function.
6681 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6683 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6685 PR tree-optimization/83648
6686 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6687 return value as malloc candidate.
6689 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6692 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6693 param as true in call to suggest_attribute.
6695 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6697 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6698 -mreadonly-in-sdata.
6700 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6702 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6704 (aarch64_crypto_aesd_fused): Likewise.
6706 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6708 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6709 (movsi_aarch64): Likewise.
6710 (load_pairsi): Likewise.
6711 (load_pairdi): Likewise.
6712 (store_pairsi): Likewise.
6713 (store_pairdi): Likewise.
6714 (load_pairsf): Likewise.
6715 (load_pairdf): Likewise.
6716 (store_pairsf): Likewise.
6717 (store_pairdf): Likewise.
6718 (zero_extend): Likewise.
6719 (trunc): Swap alternatives.
6720 (fcvt_target): Add '?' to prefer w over r.
6722 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6725 * config/i386/i386.md: Disallow non-commutative arithmetics in
6726 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6727 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6728 in the peephole2 before it.
6730 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6732 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6733 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6734 (ix86_handle_option): Handle -mcldemote.
6735 * config.gcc: New header.
6736 * config/i386/cldemoteintrin.h: New file.
6737 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6738 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6740 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6741 OPTION_MASK_ISA_CLDEMOTE.
6742 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6743 (ix86_valid_target_attribute_inner_p): Ditto.
6744 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6745 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6746 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6747 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6748 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6750 * config/i386/i386.opt: Add -mcldemote.
6751 * config/i386/x86intrin.h: New header.
6752 * doc/invoke.texi: Add -mcldemote.
6754 2018-05-14 Richard Biener <rguenther@suse.de>
6756 * doc/match-and-simplify.texi: Adjust :s documentation.
6758 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6760 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6761 intended memcpy size.
6762 (REORDER_45): Likewise.
6764 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6766 * sort.cc: New file.
6767 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6768 * vec.c (qsort_chk): Use gcc_qsort.
6769 * Makefile.in (OBJS-libcommon): Add sort.o.
6770 (build/sort.o): New target. Use it...
6771 (BUILD_RTL): ... here, and...
6772 (build/gencfn-macros): ... here, and...
6773 (build/genmatch): ... here.
6775 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6776 Chung-Ju Wu <jasonwucj@gmail.com>
6778 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6779 * config/nds32/nds32-graywolf.md: New file.
6780 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6783 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6784 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6785 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6786 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6787 * config/nds32/pipelines.md: Include n15 settings.
6789 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6790 Chung-Ju Wu <jasonwucj@gmail.com>
6792 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6793 * config/nds32/nds32-n13.md: New file.
6794 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6795 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6797 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6798 * config/nds32/nds32.md (pipeline_model): Add n13.
6799 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6800 * config/nds32/pipelines.md: Include n13 settings.
6802 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6803 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6806 * config/nds32/nds32-n10.md: New file.
6807 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6808 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6810 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6811 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6812 * config/nds32/nds32.md (pipeline_model): Add n10.
6813 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6814 * config/nds32/pipelines.md: Include n10 settings.
6816 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6817 Kito Cheng <kito.cheng@gmail.com>
6818 Chung-Ju Wu <jasonwucj@gmail.com>
6820 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6821 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6822 Add enum values for DSP extension instructions.
6823 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6825 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6826 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6828 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6829 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6830 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6831 * config/nds32/nds32-intrinsic.md: Likewise.
6832 * config/nds32/nds32_intrinsic.h: Likewise.
6833 * config/nds32/nds32-md-auxiliary.c: Likewise.
6834 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6835 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6836 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6837 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6838 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6839 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6840 TYPE_DMAC in switch statement.
6841 * config/nds32/nds32.c: New checking and implementation for DSP
6842 extension instructions.
6843 * config/nds32/nds32.h: Likewise.
6844 * config/nds32/nds32.md: Likewise.
6845 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6846 * config/nds32/predicates.md: Implement new predicates for DSP
6849 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6851 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6852 Reformat alternatives and attributes so it is easier to identify
6853 which constraints/attributes go with which instruction.
6854 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6855 (mov<mode>_softfloat32, FMOVE64): Likewise.
6856 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6857 (mov<mode>_softfloat64, FMOVE64): Likewise.
6859 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6861 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6863 (Basic PowerPC Built-in Functions): The new name of the
6864 subsection previously known as "PowerPC Built-in Functions".
6865 (Basic PowerPC Built-in Functions Available on all Configurations):
6867 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6868 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6869 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6870 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6872 2018-05-11 Martin Jambor <mjambor@suse.cz>
6875 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6878 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6881 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6883 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6885 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6886 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6887 (ix86_handle_option): Handle -mwaitpkg.
6888 * config.gcc: New header.
6889 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6891 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6893 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6894 OPTION_MASK_ISA_WAITPKG.
6895 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6896 (ix86_option_override_internal): Add PTA_WAITPKG.
6897 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6898 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6899 IX86_BUILTIN_TPAUSE.
6900 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6901 __builtin_ia32_umwait and __builtin_ia32_tpause.
6902 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6903 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6904 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6905 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6906 UNSPECV_TPAUSE): New.
6907 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6908 * config/i386/i386.opt: Add -mwaitpkg.
6909 * config/i386/waitpkgintrin.h: New file.
6910 * config/i386/x86intrin.h: New header.
6911 * doc/invoke.texi: Add -mwaitpkg.
6913 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6916 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6918 (cortex-m0): Use armv6s-m isa.
6919 (cortex-m0plus): Likewise.
6920 (cortex-m1): Likewise.
6921 (cortex-m0.small-multiply): Likewise.
6922 (cortex-m0plus.small-multiply): Likewise.
6923 (cortex-m1.small-multiply): Likewise.
6925 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6926 Jakub Jelinek <jakub@redhat.com>
6928 PR tree-optimization/85692
6929 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6930 source permute as well.
6932 2018-05-11 Martin Liska <mliska@suse.cz>
6935 * doc/extend.texi: Document LLVM style format for no_sanitize
6938 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6940 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6941 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6942 (mode_supports_vsx_dform_quad): Likewise.
6943 (mode_supports_vmx_dform): Move these functions to be next to the
6944 other mode_supports functions.
6945 (mode_supports_dq_form): Likewise.
6946 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6947 mode_supports_dq_form.
6948 (reg_offset_addressing_ok_p): Likewise.
6949 (offsettable_ok_by_alignment): Likewise.
6950 (rs6000_legitimate_offset_address_p): Likewise.
6951 (legitimate_lo_sum_address_p): Likewise.
6952 (rs6000_legitimize_address): Likewise.
6953 (rs6000_legitimize_reload_address): Likewise.
6954 (rs6000_secondary_reload_inner): Likewise.
6955 (rs6000_preferred_reload_class): Likewise.
6956 (rs6000_output_move_128bit): Likewise.
6958 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6960 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6961 Generate SImode target register for null target.
6962 <case IX86_BUILTIN_XGETBV>: Ditto.
6963 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6964 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6966 2018-05-10 Carl Love <cel@us.ibm.com>
6968 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6969 dcbtt and dcbtstt if operands[2] is 0.
6971 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6974 * config/i386/sse.md (usadv64qi): New expander.
6976 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6978 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6979 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6980 -maltivec=be support.
6981 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6982 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6983 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6984 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6985 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6986 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6987 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6988 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6989 altivec_vsumsws): Adjust.
6990 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6991 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6992 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6994 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6995 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6996 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6997 (altivec_lve<VI_char>x): Delete expand.
6998 (*altivec_lve<VI_char>x_internal): Rename to...
6999 (altivec_lve<VI_char>x): ... this.
7000 (altivec_lvxl_<mode>): Delete expand.
7001 (*altivec_lvxl_<mode>_internal): Rename to ...
7002 (altivec_lvxl_<mode>): ... this.
7003 (altivec_stvxl_<mode>): Delete expand.
7004 (*altivec_stvxl_<mode>_internal): Rename to ...
7005 (altivec_stvxl_<mode>): ... this.
7006 (altivec_stve<VI_char>x): Delete expand.
7007 (*altivec_stve<VI_char>x_internal): Rename to ...
7008 (altivec_stve<VI_char>x): ... this.
7009 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7010 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7011 reduc_plus_scal_<mode>): Adjust.
7012 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7014 (rs6000_cpu_cpp_builtins): Adjust.
7015 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7016 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7017 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7018 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7019 -maltivec=be support.
7020 (rs6000_split_vec_extract_var): Adjust.
7021 (rs6000_split_v4si_init): Adjust.
7022 (swap_selector_for_mode): Delete.
7023 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7024 altivec_expand_stvex_be): Delete.
7025 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7026 -maltivec=be support.
7027 (rs6000_gimple_fold_builtin): Ditto.
7028 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7030 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7031 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7032 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7033 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7034 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7035 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7036 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7037 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7038 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7039 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7040 anonymous split): Adjust.
7041 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7042 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7044 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7046 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7047 when --with-gxx-include-dir is also specified.
7048 * configure: Regenerate.
7050 2018-05-09 Jim Wilson <jimw@sifive.com>
7053 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7054 * config/riscv/t-withmultilib: New.
7055 * config/riscv/withmultilib.h: New.
7056 * doc/install.texi: Document RISC-V --with-multilib-list support.
7058 2018-05-09 Richard Biener <rguenther@suse.de>
7060 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7062 (vect_bb_vectorization_profitable_p): Adjust. Compute
7063 actual scalar cost using the cost vector and the add_stmt_cost
7066 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7068 PR rtl-optimization/85645
7069 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7070 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7072 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7074 PR rtl-optimization/85645
7075 * shrink-wrap.c (spread_components): Return a boolean saying if
7076 anything was changed.
7077 (try_shrink_wrapping_separate): Iterate spread_components until
7078 nothing changes anymore.
7080 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7082 PR rtl-optimization/85645
7083 * regrename.c (build_def_use): Also kill the chains that include the
7084 destination of a REG_CFA_REGISTER note.
7086 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7088 PR rtl-optimization/85645
7089 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7090 insn that has a REG_CFA_REGISTER note.
7092 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7094 * cfgexpand.c (expand_clobber): New function.
7095 (expand_gimple_stmt_1): Use it.
7096 * tree-vect-stmts.c (vect_clobber_variable): New function,
7098 (vectorizable_simd_clone_call): ...here.
7099 (vectorizable_store): Emit a clobber either side of an
7100 IFN_STORE_LANES sequence.
7101 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7103 2018-05-09 Tom de Vries <tom@codesourcery.com>
7106 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7107 (define_insn "trap_if_false"): Add exit after trap.
7109 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7111 PR rtl-optimization/85638
7112 * bb-reorder.c: Include common/common-target.h.
7113 (create_forwarder_block): New function extracted from...
7114 (fix_up_crossing_landing_pad): ...here. Rename into...
7115 (dw2_fix_up_crossing_landing_pad): ...this.
7116 (sjlj_fix_up_crossing_landing_pad): New function.
7117 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7118 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7119 from both partitions and exit the loop after one iteration.
7121 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7124 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7126 (Basic PowerPC Built-in Functions): The new name of the
7127 subsection previously known as "PowerPC Built-in Functions".
7128 (Basic PowerPC Built-in Functions Available on all Configurations):
7130 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7132 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7134 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7136 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7139 2018-05-08 Jim Wilson <jimw@sifive.com>
7141 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7142 (LD_EMUL_SUFFIX): New.
7143 (LINK_SPEC): Use it.
7145 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7147 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7149 (Basic PowerPC Built-in Functions): The new name of the
7150 subsection previously known as "PowerPC Built-in Functions".
7151 (Basic PowerPC Built-in Functions Available on all Configurations):
7153 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7155 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7157 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7159 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7162 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7165 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7166 after cmpelim optimization.
7168 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7170 * config.gcc: Support "goldmont".
7171 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7172 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7174 * config/i386/i386.c (m_GOLDMONT): Define.
7175 (processor_target_table): Add "goldmont".
7176 (PTA_GOLDMONT): Define.
7177 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7178 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7179 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7180 (fold_builtin_cpu): Add "goldmont".
7181 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7182 (ix86_option_override_internal): Add "goldmont".
7183 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7184 (processor_type): Add PROCESSOR_GOLDMONT.
7185 * config/i386/i386.md: Add CPU "glm".
7186 * config/i386/glm.md: New file.
7187 * config/i386/x86-tune.def: Add m_GOLDMONT.
7188 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7190 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7193 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7195 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7196 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7197 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7200 * config/i386/i386.c (ix86_fold_builtin): Handle
7201 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7204 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7205 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7207 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7210 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7211 (check_arch): Likewise.
7212 (check_fpu): Return the result rather than printing it.
7213 (end arch): Fix operator precedence.
7214 (end cpu): Likewise.
7215 (END): Print the result from check_fpu.
7217 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7218 Alan Hayward <alan.hayward@arm.com>
7219 David Sherwood <david.sherwood@arm.com>
7221 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7222 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7223 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7224 (*fcmuo<mode>_and): New patterns.
7226 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7228 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7229 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7230 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7231 (cmp_op, sve_imm_con): New code attributes.
7232 (SVE_COND_INT_CMP, imm_con): Delete.
7233 (cmp_op): Remove above unspecs from int attribute.
7234 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7236 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7237 comparison-specific unspecs.
7238 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7239 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7240 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7241 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7242 (*vec_fcm<cmp_op><mode>): Rename to...
7243 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7244 (*vec_fcmuo<mode>): Rename to...
7245 (*fcmuo<mode>): ...this and adjust likewise.
7246 (*pred_fcm<cmp_op><mode>): New pattern.
7247 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7248 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7250 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7252 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7253 (aarch64_emit_sve_predicated_cond): New function.
7254 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7255 (aarch64_emit_unspec_cond_or): Replace with...
7256 (aarch64_emit_sve_or_conds): ...this new function. Use
7257 aarch64_emit_sve_ptrue_op for the individual comparisons and
7258 aarch64_emit_binop to OR them together.
7259 (aarch64_emit_inverted_unspec_cond): Replace with...
7260 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7261 aarch64_emit_sve_ptrue_op for the comparison and
7262 aarch64_emit_unop to invert the result.
7263 (aarch64_expand_sve_vec_cmp_float): Update after the above
7264 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7266 2018-05-07 Nathan Sidwell <nathan@acm.org>
7268 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7269 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7270 (Backwards Compatibility): Likewise.
7272 2018-05-07 Luis Machado <luis.machado@linaro.org>
7276 2018-05-07 Luis Machado <luis.machado@linaro.org>
7278 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7279 <prefetch_dynamic_strides>: New const bool field.
7280 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7281 prefetch_dynamic_strides.
7282 (exynosm1_prefetch_tune): Likewise.
7283 (thunderxt88_prefetch_tune): Likewise.
7284 (thunderx_prefetch_tune): Likewise.
7285 (thunderx2t99_prefetch_tune): Likewise.
7286 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7288 (aarch64_override_options_internal): Update to set
7289 PARAM_PREFETCH_DYNAMIC_STRIDES.
7290 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7291 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7292 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7293 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7294 prefetch-dynamic-strides setting.
7296 2018-05-07 Luis Machado <luis.machado@linaro.org>
7298 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7299 <minimum_stride>: New const int field.
7300 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7301 minimum_stride field.
7302 (exynosm1_prefetch_tune): Likewise.
7303 (thunderxt88_prefetch_tune): Likewise.
7304 (thunderx_prefetch_tune): Likewise.
7305 (thunderx2t99_prefetch_tune): Likewise.
7306 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7307 (aarch64_override_options_internal): Update to set
7308 PARAM_PREFETCH_MINIMUM_STRIDE.
7309 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7310 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7311 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7312 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7313 stride is constant and is below the minimum stride threshold.
7315 2018-05-07 Luis Machado <luis.machado@linaro.org>
7317 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7320 2018-05-07 Luis Machado <luis.machado@linaro.org>
7322 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7323 <prefetch_dynamic_strides>: New const bool field.
7324 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7325 prefetch_dynamic_strides.
7326 (exynosm1_prefetch_tune): Likewise.
7327 (thunderxt88_prefetch_tune): Likewise.
7328 (thunderx_prefetch_tune): Likewise.
7329 (thunderx2t99_prefetch_tune): Likewise.
7330 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7332 (aarch64_override_options_internal): Update to set
7333 PARAM_PREFETCH_DYNAMIC_STRIDES.
7334 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7335 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7336 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7337 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7338 prefetch-dynamic-strides setting.
7340 2018-05-07 Luis Machado <luis.machado@linaro.org>
7342 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7343 <minimum_stride>: New const int field.
7344 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7345 minimum_stride field.
7346 (exynosm1_prefetch_tune): Likewise.
7347 (thunderxt88_prefetch_tune): Likewise.
7348 (thunderx_prefetch_tune): Likewise.
7349 (thunderx2t99_prefetch_tune): Likewise.
7350 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7351 (aarch64_override_options_internal): Update to set
7352 PARAM_PREFETCH_MINIMUM_STRIDE.
7353 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7354 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7355 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7356 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7357 stride is constant and is below the minimum stride threshold.
7359 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7362 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7363 the type is addressable. Don't force op into register if it has
7366 2018-05-05 Roland McGrath <mcgrathr@google.com>
7369 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7370 any section for which we don't know a specific type it should have,
7371 regardless of name. Previously this was done only for the exact
7372 names ".init_array", ".fini_array", and ".preinit_array".
7373 (default_elf_asm_named_section): Add comment about
7374 relationship with default_section_type_flags and SECTION_NOTYPE.
7375 (get_section): Don't consider it a type conflict if one side has
7376 SECTION_NOTYPE and the other doesn't, as long as neither has the
7377 SECTION_BSS et al used in the default_section_type_flags logic.
7379 2018-05-05 Tom de Vries <tom@codesourcery.com>
7382 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7383 (workaround_barsyncs): New function.
7384 (nvptx_reorg): Use workaround_barsyncs.
7385 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7386 (define_expand "nvptx_membar_cta"): New define_expand.
7387 (define_insn "*nvptx_membar_cta"): New insn.
7389 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7391 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7392 To improve optimization opportunities.
7393 * builtin-types.def: The new needed builtin types for the above.
7395 2018-05-04 Richard Biener <rguenther@suse.de>
7397 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7398 * gimple-ssa-store-merging.c
7399 (imm_store_chain_info::output_merged_store): Remove redundant create,
7400 release split_store vector contents on failure.
7401 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7402 scalar stmt vector on cache hit.
7404 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7406 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7408 * config.gcc (powerpc-xilinx-eabi*): Remove.
7409 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7411 (fusion_addis_mem_combo_load): Ditto.
7412 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7414 (rs6000_cpu_cpp_builtins): Ditto.
7415 * config/rs6000/rs6000-linux.c
7416 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7417 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7418 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7420 (rs6000_setup_reg_addr_masks): Ditto.
7421 (rs6000_init_hard_regno_mode_ok): Ditto.
7422 (rs6000_option_override_internal): Ditto.
7423 (legitimate_lo_sum_address_p): Ditto.
7424 (rs6000_legitimize_address): Ditto.
7425 (rs6000_legitimize_reload_address): Ditto.
7426 (rs6000_legitimate_address_p): Ditto.
7427 (abi_v4_pass_in_fpr): Ditto.
7428 (setup_incoming_varargs): Ditto.
7429 (rs6000_gimplify_va_arg): Ditto.
7430 (rs6000_split_multireg_move): Ditto.
7431 (rs6000_savres_strategy): Ditto.
7432 (rs6000_emit_prologue_components): Ditto.
7433 (rs6000_emit_epilogue_components): Ditto.
7434 (rs6000_emit_prologue): Ditto.
7435 (rs6000_emit_epilogue): Ditto.
7436 (rs6000_elf_file_end): Ditto.
7437 (rs6000_function_value): Ditto.
7438 (rs6000_libcall_value): Ditto.
7439 * config/rs6000/rs6000.h: Ditto.
7440 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7441 (TARGET_MINMAX): ... this. New.
7442 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7443 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7444 (*movsi_internal1_single): Delete.
7445 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7446 mfpu=, mxilinx-fpu): Delete.
7447 * config/rs6000/singlefp.h: Delete.
7448 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7449 * config/rs6000/t-rs6000: Ditto.
7450 * config/rs6000/t-xilinx: Delete.
7451 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7452 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7453 (VStype_simple): Delete.
7454 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7455 * config/rs6000/xfpu.h: Delete.
7456 * config/rs6000/xfpu.md: Delete.
7457 * config/rs6000/xilinx.h: Delete.
7458 * config/rs6000/xilinx.opt: Delete.
7459 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7460 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7462 2018-05-04 Tom de Vries <tom@codesourcery.com>
7465 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7468 2018-05-04 Richard Biener <rguenther@suse.de>
7471 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7472 (expand_complex_div_wide): Likewise.
7473 (expand_complex_operations_1): Likewise.
7474 (expand_complex_libcall): Preserve EH info of the original stmt.
7475 (tree_lower_complex): Handle removed blocks.
7476 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7477 on complex multiplication and division libcall builtins.
7479 2018-05-04 Richard Biener <rguenther@suse.de>
7482 * fold-const.c (negate_expr_p): Restrict negation of operand
7483 zero of a division to when we know that can happen without
7485 (fold_negate_expr_1): Likewise.
7487 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7490 * real.h (real_nextafter): Declare.
7491 * real.c (real_nextafter): New function.
7492 * fold-const-call.c (fold_const_nextafter): New function.
7493 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7494 CASE_CFN_NEXTTOWARD.
7495 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7496 even when arg1_mode is different from arg0_mode.
7498 2018-05-03 Nathan Sidwell <nathan@acm.org>
7500 * doc/extend.texi (Deprecated Features): Remove
7502 (Backwards Compatibility): Likewise.
7503 * doc/invoke.texi (C++ Language Options): Likewise.
7504 (C++ Dialect Options): Likewise.
7506 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7509 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7510 _mm512_mask_mullox_epi64): New intrinsics.
7512 2018-05-03 Tom de Vries <tom@codesourcery.com>
7515 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7516 dump files): Add offload-tree.
7518 2018-05-03 Richard Biener <rguenther@suse.de>
7520 PR tree-optimization/85615
7521 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7522 to loops not nested in BBs loop father to avoid creating multi-entry
7525 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7527 PR tree-optimization/70291
7528 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7529 arguments. Change return type to tree. Emit libcall as a new
7530 statement rather than replacing existing one when inplace_p is true.
7531 (expand_complex_multiplication_components): New function.
7532 (expand_complex_multiplication): Expand floating-point complex
7533 multiplication using the above.
7534 (expand_complex_division): Rename inner_type parameter to type.
7535 Update expand_complex_libcall call-site.
7536 (expand_complex_operations_1): Update expand_complex_multiplication
7537 and expand_complex_division call-sites.
7539 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7542 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7543 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7544 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7545 the highest significant bit of the shift count mask is clear. In
7546 check whether and[sq]i3 is needed verify that all significant bits
7547 of the shift count other than the highest are set.
7549 2018-05-02 Tom de Vries <tom@codesourcery.com>
7552 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7553 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7554 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7555 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7556 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7557 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7558 __builtin_goacc_parlevel_size.
7560 2018-05-02 Richard Biener <rguenther@suse.de>
7562 PR tree-optimization/85597
7563 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7564 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7567 2018-05-02 Tom de Vries <tom@codesourcery.com>
7570 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7571 dump files): Add ltrans-tree.
7573 2018-05-02 Tom de Vries <tom@codesourcery.com>
7576 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7577 dump files): Add wpa-ipa.
7579 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7581 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7582 powerpc*-*-linux*paired* target.
7583 * config/rs6000/750cl.h: Delete.
7584 * config/rs6000/paired.h: Delete.
7585 * config/rs6000/paired.md: Delete.
7586 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7588 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7589 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7590 comment. Remove paired float support.
7591 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7592 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7594 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7595 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7597 * config/rs6000/rs6000.c: Remove paired float support.
7598 (paired_expand_vector_init, paired_expand_vector_move,
7599 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7600 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7601 paired_expand_builtin, paired_expand_predicate_builtin,
7602 paired_init_builtins): Delete.
7603 * config/rs6000/rs6000.h: Remove paired float support.
7604 * config/rs6000/rs6000.md: Remove paired float support.
7605 (move_from_CR_ov_bit): Delete.
7606 * config/rs6000/rs6000.opt (mpaired): Delete.
7607 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7608 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7610 2018-05-02 Richard Biener <rguenther@suse.de>
7613 * gimplify.c (gimplify_save_expr): When in SSA form allow
7614 SAVE_EXPRs to compute to SSA vars.
7616 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7619 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7620 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7621 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7622 clobber operands[2], instead use a new pseudo. Formatting fixes.
7624 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7626 PR tree-optimization/85586
7627 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7628 exit early for statements in the same group if the accesses are
7631 2018-05-02 Tom de Vries <tom@codesourcery.com>
7634 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7637 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7639 PR tree-optimization/85143
7640 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7642 2018-05-01 Tom de Vries <tom@codesourcery.com>
7645 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7646 not found" error message.
7648 2018-05-01 Tom de Vries <tom@codesourcery.com>
7651 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7652 * vec.c (test_ordered_remove_if): New function.
7653 (vec_c_tests): Call test_ordered_remove_if.
7654 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7655 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7656 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7657 VEC_ORDERED_REMOVE_IF.
7659 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7661 PR tree-optimization/82665
7662 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7663 pointer subtraction where arguments come from a memchr call.
7665 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7667 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7668 --push-state --as-needed and --pop-state instead of --as-needed and
7669 --no-as-needed if ld supports it.
7670 * configure: Regenerated.
7673 * doc/install.texi2html: Replace _002d with - and _002a with * in
7674 generated html files using sed.
7676 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7679 * gcc-rich-location.c (blank_line_before_p): New function.
7680 (use_new_line): New function.
7681 (gcc_rich_location::add_fixit_insert_formatted): New function.
7682 * gcc-rich-location.h
7683 (gcc_rich_location::add_fixit_insert_formatted): New function.
7685 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7687 * selftest.c (assert_streq): Rename "expected" and "actual" to
7688 "val1" and "val2". Extend NULL-handling to cover both inputs
7689 symmetrically, while still requiring both to be non-NULL for a pass.
7690 * selftest.h (assert_streq): Rename "expected" and "actual" to
7692 (ASSERT_EQ): Likewise.
7693 (ASSERT_EQ_AT): Likewise.
7694 (ASSERT_KNOWN_EQ): Likewise.
7695 (ASSERT_KNOWN_EQ_AT): Likewise.
7696 (ASSERT_NE): Likewise.
7697 (ASSERT_MAYBE_NE): Likewise.
7698 (ASSERT_MAYBE_NE_AT): Likewise.
7699 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7700 the assertion to pass.
7701 (ASSERT_STREQ_AT): Likewise.
7703 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7705 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7706 interaction with -pie.
7708 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7710 * selftest.h: Fix alphabetization of per-source-file selftest
7713 2018-04-30 Jason Merrill <jason@redhat.com>
7715 PR c++/61982 - dead stores to destroyed objects.
7716 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7719 2018-04-30 Jason Merrill <jason@redhat.com>
7721 * tree.c (build_clobber): New.
7722 * tree.h: Declare it.
7723 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7725 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7727 * diagnostic-show-locus.c (layout::layout): Update for
7728 location_get_source_line returning a char_span.
7729 (struct char_span): Move to input.h.
7730 (struct correction): Update for fields in char_span becoming
7732 (struct source_line): Update for location_get_source_line
7733 returning a char_span.
7734 (layout::print_line): Likewise.
7735 * edit-context.c (edited_file::print_content): Likewise.
7736 (edited_file::print_diff_hunk): Likewise.
7737 (edited_file::print_run_of_changed_lines): Likewise.
7738 (edited_file::get_num_lines): Likewise.
7739 (edited_line::edited_line): Likewise.
7740 * final.c (asm_show_source): Likewise.
7741 * input.c (location_get_source_line): Convert return type
7742 from const char * to char_span, losing the final "line_len"
7744 (dump_location_info): Update for the above.
7745 (get_substring_ranges_for_loc): Likewise. Use a char_span
7746 when handling the literal within the line.
7747 (test_reading_source_line): Update for location_get_source_line
7748 returning a char_span.
7749 * input.h (class char_span): Move here from
7750 diagnostic-show-locus.c, converting from a struct to a class.
7751 Make data members private.
7752 (char_span::operator bool): New.
7753 (char_span::length): New.
7754 (char_span::get_buffer): New.
7755 (char_span::operator[]): New.
7756 (char_span::subspan): Make const.
7757 (char_span::xstrdup): New.
7758 (location_get_source_line): Convert return type from const char *
7759 to char_span, losing the final "line_size" param.
7761 2018-04-30 Jan Hubicka <jh@suse.cz>
7763 * lto-wrapper.c (ltrans_priorities): New static var.
7764 (cmp_priority): New.
7765 (run_gcc): Read priorities and if doing parallel build order
7766 the Makefile by them.
7768 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7770 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7772 2018-04-30 Richard Biener <rguenther@suse.de>
7774 * tree-cfg.c (verify_address): Remove base argument, add
7775 flag whether to check TREE_ADDRESSABLE and do that.
7776 (verify_expr): Remove.
7777 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7778 (verify_gimple_assign_single): Likewise.
7779 (verify_gimple_switch): Likewise.
7780 (verify_expr_location_1): Dereference tp once. Add (disabled)
7781 piece from verify_expr.
7782 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7784 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7786 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7788 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7790 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7791 (small_data_pattern): Likewise.
7792 (arc_rewrite_small_data): Likewise.
7793 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7794 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7795 (get_symbol_alignment): New function.
7796 (legitimate_small_data_address_p): Likewise.
7797 (legitimate_scaled_address): Update, call
7798 legitimate_small_data_address_p.
7799 (output_sdata): New static variable.
7800 (arc_print_operand): Update how we handle small data operands.
7801 (arc_print_operand_address): Likewise.
7802 (arc_legitimate_address_p): Update, use
7803 legitimate_small_data_address_p.
7804 (arc_rewrite_small_data_p): Remove.
7805 (arc_rewrite_small_data_1): Likewise.
7806 (arc_rewrite_small_data): Likewise.
7807 (small_data_pattern): Likewise.
7808 (compact_sda_memory_operand): Update to use
7809 legitimate_small_data_address_p and get_symbol_alignment.
7810 (prepare_move_operands): Don't rewite sdata pattern.
7811 (prepare_extend_operands): Remove.
7812 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7814 (zero_extendqisi2): Likewise.
7815 (zero_extendhisi2): Likewise.
7816 (extendqihi2): Likewise.
7817 (extendqisi2): Likewise.
7818 (extendhisi2): Likewise.
7822 * config/arc/constraints.md (Usd): Change it to memory constraint.
7824 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7826 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7827 as source of std instructions.
7828 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7829 allow 6-bit constants as source for store instructions.
7830 (movdi_insn): Update instruction pattern to allow 6-bit constants
7831 as source for store instructions.
7833 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7835 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7837 2018-04-30 Nathan Sidwell <nathan@acm.org>
7838 Sandra Loosemore <sandra@codesourcery.com>
7840 * dumpfile.c (dump_open): Allow '-' for stdout.
7841 * doc/invoke.texi (Developer Options): Document dump filename
7842 determination early. Document stdin/stdout selection.
7844 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7846 Microblaze Target: PIC data text relative
7848 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7849 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7851 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7852 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7853 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7854 New addressing mode for data-text relative position indepenedent code.
7855 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7856 'ADDRESS_SYMBOLIC_TXT_REL'.
7857 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7858 (microblaze_legitimate_pic_operand): Exclude function calls from
7859 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7860 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7862 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7863 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7864 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7865 for 'address + offset'.
7866 (microblaze_expand_prologue): Add new function prologue call for
7868 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7869 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7870 table in case of TARGET_PIC_DATA_TEXT_REL.
7871 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7872 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7873 Add new macros 'UNSPEC_TEXT',
7874 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7875 + exclude function calls from 'UNSPEC_PLT' in case of data text
7877 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7878 new target hook for generating address diff vector tables in case of
7880 * doc/tm.texi : Regenerate.
7881 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7882 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7883 of addr diff vector generation.
7884 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7885 target hook definition.
7886 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7887 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7888 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7889 option for data text relative.
7891 2018-04-30 Richard Biener <rguenther@suse.de>
7893 * tree-chrec.h (evolution_function_is_constant_p): Remove
7895 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7897 2018-04-30 Richard Biener <rguenther@suse.de>
7900 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7902 2018-04-30 Richard Biener <rguenther@suse.de>
7904 PR tree-optimization/28364
7905 PR tree-optimization/85275
7906 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7907 copying first exit test.
7909 2018-04-28 Mark Wielaard <mark@klomp.org>
7911 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7913 (dwarf_AT): Handle DW_AT_addr_base.
7914 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7916 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7919 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7920 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7921 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7922 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7924 2018-04-28 Richard Biener <rguenther@suse.de>
7926 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7927 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7928 to reflect use. Only add interesting stmts.
7930 2018-04-27 Martin Jambor <mjambor@suse.cz>
7933 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7934 the jump function allows for passing through aggregate values.
7936 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7938 * input.h (in_system_header_at): Convert from macro to inline
7940 (from_macro_expansion_at): Likewise.
7941 (from_macro_definition_at): Likewise.
7943 2018-04-27 Jeff Law <law@redhat.com>
7945 * config.gcc: Mark tile* targets as deprecated/obsolete.
7947 2018-04-27 Richard Biener <rguenther@suse.de>
7949 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7952 2018-04-27 Richard Biener <rguenther@suse.de>
7954 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7956 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7958 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7959 with Yd constraint. Set "preferred_for_speed" attribute from
7960 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7962 (*movdi_internal): Ditto.
7963 (movti_interunit splitters): Remove
7964 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7965 (movdi_interunit splitters): Ditto.
7966 * config/i386/constraints.md (Ye): Remove.
7967 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7969 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7972 * config/aarch64/constraints.md (Usg): Limit to 31.
7975 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7977 PR tree-optimization/85529
7978 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7979 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7980 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7981 zero extension or masking of the MSB bit.
7982 (optimize_range_tests): Add FIRST_BB argument, pass it through
7983 to optimize_range_tests_var_bound.
7984 (maybe_optimize_range_tests, reassociate_bb): Adjust
7985 optimize_range_tests callers.
7987 2018-04-26 Richard Biener <rguenther@suse.de>
7988 Jakub Jelinek <jakub@redhat.com>
7990 * cgraph.h (symbol_table): Just declare debug method here.
7991 * symtab.c (symbol_table::debug): Define.
7993 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7995 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7997 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7999 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8000 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8001 (*movdi_internal): Substitute Yi and Yj constraint with x
8002 and Ym and Yn constraint with y constraint. Update "isa"
8003 attribute and set "preferred_for_speed" attribute from
8004 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8005 (*movsi_internal): Ditto.
8006 (*movdf_internal): Ditto.
8007 (*movsf_internal): Ditto.
8008 (*zero_extendsidi2): Ditto.
8009 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8010 (sse2_loadld): Ditto.
8011 (*vec_extract<ssevecmodelower>_0): Ditto.
8012 (*vec_extractv4si_0_zext_sse4): Ditto.
8013 (vec_concatv2di): Ditto.
8014 (*vec_dup<mode>): Ditto.
8015 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8016 * config/i386/constraints.md (Yi): Remove.
8021 2018-04-26 Nathan Sidwell <nathan@acm.org>
8023 * dumpfile.c (dump_open): New.
8024 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8025 (dump_finish): Detect stdio/stderr by value not name.
8027 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8029 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8031 2018-04-26 Tom de Vries <tom@codesourcery.com>
8034 * config/nvptx/nvptx.c (verify_neutering_jumps)
8035 (verify_neutering_labels): New function
8036 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8038 2018-04-26 Tom de Vries <tom@codesourcery.com>
8041 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8042 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8045 2018-04-26 Richard Biener <rguenther@suse.de>
8046 Tom de Vries <tom@codesourcery.com>
8049 * lto-streamer-out.c (output_function): Fixup loops if required to match
8050 discovery done in the reader.
8052 2018-04-26 Richard Biener <rguenther@suse.de>
8054 PR tree-optimization/85116
8055 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8056 have a loop exit from the single latch predecessor. Remove
8057 case of header with just condition.
8058 (ch_base::copy_headers): Exclude infinite loops from any
8060 (pass_ch::execute): Record exits.
8062 2018-04-26 Richard Biener <rguenther@suse.de>
8064 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8065 prologue cost vector and pass it to vect_get_load_cost.
8066 (vect_get_peeling_costs_all_drs): Likewise.
8067 (vect_peeling_hash_get_lowest_cost): Likewise.
8068 (vect_enhance_data_refs_alignment): Likewise.
8070 2018-04-26 Richard Biener <rguenther@suse.de>
8073 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8074 checking of integer<->pointer conversions.
8075 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8076 sign-/zero-extending pointer types.
8077 (expand_omp_for_static_chunk): Likewise.
8079 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8080 Jean Lee <xiaoyur347@gmail.com>
8082 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8083 (TARGET_ASAN_SHADOW_OFFSET): Define.
8084 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8085 true for -fsanitize=address.
8087 2018-04-25 Mark Wielaard <mark@klomp.org>
8089 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8092 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8094 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8095 than "alu", remove explicit "memory" and "imm_disp" attributes.
8096 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8099 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8100 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8101 gen_lowpart_no_emit.
8103 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8106 * config/i386/i386.c (ix86_expand_builtin): Change memory
8107 operand to XI, extend p0 to Pmode.
8108 * config/i386/i386.md: Change unspec volatile and operand
8109 1 mode to XI, change operand 0 mode to P.
8111 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8113 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8114 GET_MODE_MASK before any checking.
8115 (nds32_can_use_bset_p): Likewise.
8116 (nds32_can_use_btgl_p): Likewise.
8118 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8120 * config/nds32/nds32-doubleword.md: New define_split pattern for
8121 illegal register number.
8123 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8125 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8127 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8129 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8131 2018-04-25 Richard Biener <rguenther@suse.de>
8133 * lto-streamer.h (LTO_major_version): Bump to 8.
8135 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8137 * BASE-VER: Set to 9.0.0.
8139 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8141 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8142 in __abskf2 and __powikf2.
8144 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8147 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8148 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8149 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8150 Use the above on operand 2. Reindent.
8151 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8153 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8156 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8158 * config/i386/i386.opt (mcet): Removed.
8159 * doc/install.texi: Remove -mcet documentation.
8160 * doc/invoke.texi: Likewise.
8162 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8165 * doc/install.texi: Remove -mcet from bootstrap-cet.
8167 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8170 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8171 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8175 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8176 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8177 containing a CONST_VECTOR.
8179 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8181 * doc/install.texi: Update newlib dependency for nvptx.
8183 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8186 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8187 instead of INTVAL when shifting x left.
8189 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8191 PR tree-optimization/85478
8192 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8193 vect_grouped_store_supported for single element vectors.
8195 2018-04-24 Richard Biener <rguenther@suse.de>
8198 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8199 load cost increase to the case of non-constant step.
8201 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8204 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8205 destination if any_malformed_asm.
8207 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8210 * expr.c (store_field): In the bitfield case, if the value comes from
8211 a function call and is returned in registers by means of a PARALLEL,
8212 do not change the mode of the temporary unless BLKmode and VOIDmode.
8214 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8216 PR rtl-optimization/85423
8217 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8218 dependencies to debug insns when the previous insn is non-debug.
8220 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8222 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8223 enums into a single definition.
8224 (fls): Fix predicates and printing.
8227 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8229 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8230 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8231 and short u6 immediate.
8232 (check_if_valid_sleep_operand): Remove.
8233 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8235 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8237 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8238 flag_always_save_lp condition.
8239 * config/nds32/nds32.opt (malways-save-lp): New option.
8241 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8243 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8244 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8245 * config/nds32/nds32.h
8246 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8247 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8249 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8251 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8252 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8254 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8255 Chung-Ju Wu <jasonwucj@gmail.com>
8257 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8259 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8260 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8262 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8264 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8266 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8268 * config/nds32/nds32-protos.h (nds32_data_alignment,
8269 nds32_local_alignment): Declare.
8270 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8271 nds32_local_alignment): New functions.
8272 (TARGET_CONSTANT_ALIGNMENT): Define.
8273 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8275 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8277 * config/nds32/nds32.c
8278 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8279 (TARGET_MODES_TIEABLE_P): Likewise.
8281 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8283 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8286 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8287 Chung-Ju Wu <jasonwucj@gmail.com>
8289 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8290 for unaligned access.
8291 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8293 * config/nds32/nds32-intrinsic.md: Likewise.
8294 * config/nds32/nds32_intrinsic.h: Likewise.
8295 * config/nds32/nds32.h (nds32_builtins): Likewise.
8296 * config/nds32/nds32.opt (munaligned-access): New option.
8297 * config/nds32/nds32.c (nds32_asm_file_start): Display
8298 flag_unaligned_access status.
8300 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8302 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8303 -mno-relax is present.
8304 * config/riscv/linux.h (LINK_SPEC): Ditto.
8306 2018-04-20 Martin Sebor <msebor@redhat.com>
8309 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8311 (gimple_fold_builtin_stxcpy_chk): Same.
8312 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8314 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8317 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8318 __powikf2 when long double is IEEE 128-bit.
8320 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8322 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8323 step to make sure stack always aligned.
8325 2018-04-20 Carl Love <cel@us.ibm.com>
8328 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8329 size check for arg0.
8331 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8332 Tom de Vries <tom@codesourcery.com>
8335 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8336 Emit insns for calls too.
8337 (nvptx_find_par): Always look for worker-level predecessor insn.
8338 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8340 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8341 (nvptx_process_pars): Propagate frames for calls.
8343 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8346 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8348 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8349 (ix86_handle_option): Don't handle OPT_mibt.
8350 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8352 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8354 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8355 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8356 (ix86_target_macros): Define __CET__ with flag_cf_protection
8357 for -fcf-protection.
8358 * config/i386/i386.c (isa2_opts): Remove -mibt.
8359 * config/i386/i386.h (TARGET_IBT): Removed.
8360 (TARGET_IBT_P): Likewise.
8361 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8362 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8363 * config/i386/i386.opt (mcet): Update help message.
8366 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8367 -mcet as an alias for -mshstk.
8369 2018-04-20 Richard Biener <rguenther@suse.de>
8372 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8373 complexity by forcing a single use of the multiply operand.
8375 2018-04-20 Martin Jambor <mjambor@suse.cz>
8378 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8379 recursion dependency to only apply to non-clones.
8381 2018-04-20 Martin Jambor <mjambor@suse.cz>
8384 * ipa-cp.c (create_specialized_node): Check that clones of
8385 self-recursive edges exist during IPA-CP.
8387 2018-04-19 Toon Moene <toon@moene.org>
8389 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8392 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8394 PR tree-optimization/85467
8395 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8396 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8397 VECTOR_CST element to type.
8399 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8402 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8403 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8404 (builtin_longjmp): Likewise.
8405 (save_stack_nonlocal): New pattern.
8406 (restore_stack_nonlocal): Likewise.
8408 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8411 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8412 Replace ASM_OUTPUT_LABEL with fprintf.
8414 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8417 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8418 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8419 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8420 define __IBT__ and __SHSTK__ for -fcf-protection.
8421 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8423 (ix86_trampoline_init): Likewise.
8424 (x86_output_mi_thunk): Likewise.
8425 (ix86_notrack_prefixed_insn_p): Likewise.
8426 (ix86_option_override_internal): Don't disallow -fcf-protection.
8427 * config/i386/i386.md (rdssp<mode>): Also enable for
8429 (incssp<mode>): Likewise.
8430 (nop_endbr): Likewise.
8431 * config/i386/i386.opt (mcet): Change help message to built-in
8435 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8436 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8437 enable CET built-in functions.
8439 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8441 * common/config/i386/i386-common.c
8442 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8443 OPTION_MASK_ISA_MOVDIRI_UNSET,
8444 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8445 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8446 * config.gcc (movdirintrin.h): New header.
8447 * config/i386/cpuid.h (bit_MOVDIRI,
8448 bit_MOVDIR64B): New bits.
8449 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8451 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8452 (VOID, PVOID, PCVOID)): New function types.
8453 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8454 __builtin_ia32_directstoreu_u64,
8455 __builtin_ia32_movdir64b): New builtins.
8456 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8457 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8459 (ix86_valid_target_attribute_inner_p): Ditto.
8460 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8461 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8462 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8463 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8464 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8465 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8466 (movdiri<mode>, movdir64b_<mode>): New.
8467 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8468 * config/i386/immintrin.h: Include movdirintrin.h.
8469 * config/i386/movdirintrin.h: New file.
8470 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8472 2018-04-19 Richard Biener <rguenther@suse.de>
8475 * cfg.c (clear_bb_flags): When loop state says we have
8476 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8478 2018-04-19 Richard Biener <rguenther@suse.de>
8480 PR tree-optimization/84737
8481 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8482 copying restrict info.
8483 (vect_setup_realignment): Use it.
8484 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8485 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8486 the first DR to all generated stores.
8487 (vectorizable_load): Likewise for loads.
8489 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8491 PR tree-optimization/85446
8492 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8493 the integral and pointer types to have the same precision.
8495 * doc/install.texi: Document --disable-cet being the default and
8498 2018-04-18 Martin Liska <mliska@suse.cz>
8500 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8503 2018-04-18 Martin Liska <mliska@suse.cz>
8506 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8509 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8510 arguments if they are comparable.
8512 2018-04-18 Martin Liska <mliska@suse.cz>
8515 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8518 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8521 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8524 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8525 ENDBR after calling __morestack.
8527 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8530 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8531 by using gcc_base_ver to generate a gcc_driver_version, and use
8532 it when generating GCC_DRIVER_NAME.
8533 * configure: Regenerate.
8535 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8538 * config.gcc: Obsolete powerpc*-*-*spe*.
8540 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8543 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8544 (stabstr_D): Change type of unum from unsigned int to
8545 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8548 2018-04-17 Jim Wilson <jimw@sifive.com>
8551 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8552 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8553 Set arg_pointer_offset after using pretend_args_size.
8555 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8557 PR rtl-optimization/85431
8558 * dse.c (record_store): Ignore zero width stores.
8561 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8562 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8563 __builtin_stack_restore rather than after it.
8564 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8565 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8566 argument instead of virtual_dynamic_stack_rtx.
8568 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8570 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8572 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8573 Add note to error message to explain internal mapping of overloaded
8574 built-in function name to non-overloaded built-in function name.
8575 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8578 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8581 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8582 where the inputs overlap with the output.
8584 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8586 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8587 (=v, v) alternative and explicit "memory" attribute.
8588 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8589 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8591 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8592 "sselog1" type instead of "sselog".
8593 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8594 "sselog". Remove explicit "memory" attribute.
8595 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8596 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8598 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8599 "sselog1" type instead of "sselog". Remove explicit "memory"
8601 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8602 use "sselog1" type instead of "sselog". Remove explicit "memory"
8604 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8605 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8607 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8608 "sselog1" type instead of "sselog". Remove explicit "memory"
8610 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8611 use "sselog1" type instead of "sselog". Remove explicit "memory"
8615 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8618 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8621 2018-04-17 Martin Jambor <mjambor@suse.cz>
8624 * ipa-cp.c (create_specialized_node): Call
8625 expand_all_artificial_thunks if necessary.
8627 2018-04-17 Martin Liska <mliska@suse.cz>
8630 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8631 in message, remote space in between '_G' and '('.
8633 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8636 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8637 avx512f_vmcmp<mode>3<round_saeonly_name>,
8638 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8639 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8640 avx512f_rndscale<mode><round_saeonly_name>,
8641 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8642 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8643 Use %<iptr>2 instead of %2 for -masm=intel.
8644 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8645 avx512f_vcvttss2usi<round_saeonly_name>,
8646 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8648 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8649 avx512f_vcvttsd2usi<round_saeonly_name>,
8650 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8651 Use %q1 instead of %1 for -masm=intel.
8652 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8653 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8654 of %3 for -masm=intel.
8655 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8657 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8659 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8661 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8662 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8664 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8666 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8667 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8668 %g1 and one with %0 and %1.
8669 (avx512er_vmrcp28<mode><round_saeonly_name>,
8670 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8672 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8673 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8674 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8675 of %0 and %{%4%} for -masm=intel.
8676 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8677 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8678 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8679 order of %0 and %{%5%}%{z%} for -masm=intel.
8681 2018-04-17 Jan Hubicka <jh@suse.cz>
8684 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8686 2018-04-17 Martin Liska <mliska@suse.cz>
8689 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8690 for target_clone error message. Make default implementation
8691 clone to be a local declaration.
8692 (separate_attrs): Add new argument and check for an empty
8694 (expand_target_clones): Handle it.
8695 (ipa_target_clone): Make redirection just for target_clones
8698 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8699 Tom de Vries <tom@codesourcery.com>
8702 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8703 tiled basic blocks without omp continue statements.
8705 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8708 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8709 vec_extract expression as having side effects to make sure it gets
8712 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8715 * config/i386/i386.c (get_builtin_code_for_version): Check
8718 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8721 * gcc/config.gcc: Support "skylake".
8722 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8724 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8725 (processor_target_table): Add "skylake".
8726 (ix86_option_override_internal): Add "skylake".
8727 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8728 PROCESSOR_CANNONLAKE.
8729 (get_builtin_code_for_version): Fix priority for
8730 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8731 PROCESSOR_SKYLAKE-AVX512.
8732 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8733 (processor_type): Add PROCESSOR_SKYLAKE.
8735 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8736 Jason Merrill <jason@redhat.com>
8739 * convert.c (convert_to_integer_1): Use direct recursion for
8740 enumeral types and types with a precision less than the number
8741 of bits in their mode.
8743 2018-04-16 Julia Koval <julia.koval@intel.com>
8746 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8747 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8749 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8752 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8753 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8754 and -mno-direct-move.
8756 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8759 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8760 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8761 Ensure negative shifts result in {0}.
8763 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8765 PR rtl-optimization/79916
8766 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8767 regs (if any) to define how to gnerate SD moves when LRA is in
8770 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8772 PR rtl-optimization/85393
8773 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8774 * except.c (expand_dw2_landing_pad_for_region): Make static.
8775 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8776 a label and unconditional jump to old_bb, rather than
8777 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8780 PR rtl-optimization/85376
8781 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8782 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8783 instead of a specific value.
8785 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8786 Bin Cheng <bin.cheng@arm.com>
8788 PR tree-optimization/82965
8789 PR tree-optimization/83991
8790 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8791 by_profile_only parameter.
8792 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8793 information if the loop was predicted to iterate too many times.
8794 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8796 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8799 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8802 2018-04-13 Martin Liska <mliska@suse.cz>
8803 Jakub Jelinek <jakub@redhat.com>
8806 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8807 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8808 * builtins.c (expand_builtin_memory_copy_args): Use
8809 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8810 handle dest_addr == pc_rtx.
8812 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8815 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8816 asked to not generate direct moves.
8817 (fix_trunc<mode>si2_stfiwx): Similar.
8818 (fix_trunc<mode>si2_internal): Similar.
8820 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8823 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8824 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8825 lookup if dest in some wider mode is known to be const0_rtx and
8826 if so, record permanent equivalence for it to be ZERO_EXTEND of
8827 the narrower mode destination.
8829 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8831 * lto-streamer-out.c (output_function): Revert 259346.
8832 * omp-expand.c (expand_oacc_for): Likewise.
8834 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8836 PR rtl-optimization/85354
8837 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8838 * sel-sched.c (sel_global_init): ... here.
8840 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8843 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8844 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8845 mode for PE-COFF targets.
8846 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8847 (i386_pe_asm_lto_end): Likewise.
8848 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8849 (TARGET_ASM_LTO_END): Likewise.
8850 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8851 (i386_pe_asm_lto_start): New function.
8852 (i386_pe_asm_lto_end): Likewise.
8854 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8855 Richard Biener <rguenther@suse.de>
8858 * lto-streamer-out.c (output_function): Fix CFG loop state before
8860 * omp-expand.c (expand_oacc_for): Handle calls to internal
8861 functions like regular functions.
8863 2018-04-12 Richard Biener <rguenther@suse.de>
8866 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8867 for the early LTO debug to properly generate references to it
8868 during DIE emission. Do not re-use that for the skeleton for
8870 (dwarf2out_early_finish): Likewise.
8872 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8875 * config/i386/sse.md
8876 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8877 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8878 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8879 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8880 and output is a reg, avoid creating invalid lowpart subreg, but
8881 instead split into a 512-bit move. Don't split if not AVX512VL,
8882 input is xmm16+ reg and output is a mem.
8883 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8884 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8885 xmm16+ reg and output is a mem.
8887 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8889 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8890 also for flag_dwarf2_cfi_asm.
8892 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8894 PR rtl-optimization/85342
8895 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8896 a bool scalar var inside of the loop instead. Don't try to update
8897 recog_data.operand after failed apply_change_group.
8899 2018-04-12 Tom de Vries <tom@codesourcery.com>
8902 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8903 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8904 array with flexible array member as array without given dimension.
8905 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8906 argument for undefined param to true.
8908 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8911 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8912 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8913 from PowerPC section.
8914 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8915 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8916 help text that is too long.
8917 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8918 help text that is too long.
8919 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8920 help text that is too long.
8922 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8924 * config/alpha/alpha.md (stack_probe_internal): Rename
8925 from "probe_stack". Update all callers.
8927 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8929 PR rtl-optimization/84566
8930 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8931 sched_macro_fuse_insns.
8933 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8936 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8937 (compute_block_dependences): ... from here.
8939 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8941 PR tree-optimization/85331
8942 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8943 from int to HOST_WIDE_INT.
8945 2018-04-11 Martin Jambor <mjambor@suse.cz>
8948 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8949 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8950 not the same as the source val.
8951 (cgraph_edge_brings_value_p): New parameter.
8952 (gather_edges_for_value): Pass destination value to
8953 cgraph_edge_brings_value_p.
8954 (perhaps_add_new_callers): Likewise.
8955 (get_info_about_necessary_edges): Likewise and exclude values brought
8956 only by self-recursive edges.
8957 (create_specialized_node): Redirect only clones of self-calling edges.
8958 (+self_recursive_pass_through_p): New function.
8959 (find_more_scalar_values_for_callers_subset): Use it.
8960 (find_aggregate_values_for_callers_subset): Likewise.
8961 (known_aggs_to_agg_replacement_list): Removed.
8962 (decide_whether_version_node): Re-calculate known constants for all
8963 remaining context clones.
8965 2018-04-11 Richard Biener <rguenther@suse.de>
8968 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8969 from early DWARF output.
8970 (dwarf2out_early_finish): Output line info unconditionally into
8971 early DWARF and add reference to it.
8973 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8976 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8977 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8978 other than V2DFmode using iptr mode attribute.
8979 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8981 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8983 PR rtl-optimization/84659
8984 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8986 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8989 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8991 (output_loc_list): Pass address of a dummy size variable even in the
8992 locview handling loop.
8993 (index_location_lists): Add comment on why skip_loc_list_entry can't
8996 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8999 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9002 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9005 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9006 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9007 and -mstring-compare-inline-limit.
9009 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9012 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9013 for stack clash protection in a register whenever we need it to be in
9016 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9018 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9019 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9021 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9024 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9026 (mlong-double-): Ditto.
9027 * config/rs6000/sysv4.opt (msdata=): Ditto.
9028 (mtls-size=): Ditto.
9030 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9032 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9033 erroneous entries for
9034 "vector int vec_ldl (int, long int *)", and
9035 "vector unsigned int vec_ldl (int, unsigned long int *)".
9036 Add comments and entries for
9037 "vector bool char vec_ldl (int, bool char *)",
9038 "vector bool short vec_ldl (int, bool short *)",
9039 "vector bool int vec_ldl (int, bool int *)",
9040 "vector bool long long vec_ldl (int, bool long long *)",
9041 "vector pixel vec_ldl (int, pixel *)",
9042 "vector long long vec_ldl (int, long long *)",
9043 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9044 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9045 type tree bool_long_long_type_node and correct definition of
9046 bool_V2DI_type_node to make reference to this new type tree.
9047 (rs6000_mangle_type): Replace erroneous reference to
9048 bool_long_type_node with bool_long_long_type_node.
9049 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9050 comments to emphasize sign distinctions for char and int types and
9051 replace RS6000_BTI_bool_long constant with
9052 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9053 use of RS6000_BTI_pixel.
9054 (bool_long_type_node): Remove this macro definition.
9055 (bool_long_long_type_node): New macro definition
9057 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9059 PR rtl-optimization/85300
9060 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9061 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9062 simplify_unary_operation fails.
9064 2018-04-10 Martin Liska <mliska@suse.cz>
9066 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9067 cgraph_edge and ipa_ref.
9069 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9073 * config/i386/sse.md
9074 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9075 computation of the VEC_MERGE selector from mask.
9076 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9077 Fix decoding of the VEC_MERGE selector into mask.
9079 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9081 PR tree-optimization/85286
9082 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9084 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9086 * final.c (final_1): Set insn_last_address as well as
9087 insn_current_address.
9089 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9092 * explow.c (emit_stack_probe): Call validize_mem on memory location
9093 before passing it to gen_probe_stack. Create address operand and
9094 legitimize it for the probe_stack_address case.
9096 2018-04-09 Jan Hubicka <jh@suse.cz>
9099 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9100 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9101 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9103 (free_lang_data): Rebuild type inheritance graph.
9105 2018-04-09 Martin Sebor <msebor@redhat.com>
9107 * invoke.texi (-finline-small-functions): Mention other optimization
9109 (-findirect-inlining, -fpartial-inlining): Same.
9110 (-finline-functions-called-once): Same.
9111 (-freorder-blocks-and-partition): Same.
9113 2018-04-09 Jan Hubicka <jh@suse.cz>
9116 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9117 jumps; choose last target that matches the criteria (i.e.
9118 no partition changes for non-crossing jumps).
9119 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9120 support for redirecting crossing jumps to non-crossing.
9122 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9124 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9125 also for naked functions.
9127 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9129 * config/arc/arc.md (add_shift): New pattern.
9130 (add_shift2): Likewise.
9131 (sub_shift): Likewise.
9132 (sub_shift_cmp0_noout): Likewise.
9133 (compare_si_ashiftsi): Likewise.
9134 (xbfu_cmp0_noout): New combine pattern.
9135 (xbfu_cmp0"): Likewise.
9136 (movsi_set_cc_insn): Place the predicable variant first.
9137 (commutative_binary_cmp0_noout): Remove clobber.
9138 (commutative_binary_cmp0): New pattern.
9139 (noncommutative_binary_cmp0): Likewise.
9140 (noncommutative_binary_cmp0_noout): Likewise.
9141 (noncommutative_binary_comparison_result_used): Removed.
9142 (rsub_cmp0): New pattern.
9143 (rsub_cmp0_noout): Likewise.
9144 (extzvsi): Changed, keep only meaningful variants.
9145 (SQH, SEZ): New iterators.
9146 (SQH_postfix): New mode attribute.
9147 (SEZ_prefix): New code attribute.
9148 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9149 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9150 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9152 (noncommutative_operator): Check the availability of barrel
9155 2018-04-09 Richard Biener <rguenther@suse.de>
9157 PR tree-optimization/85284
9158 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9159 Only use the niter constraining form of simple_iv when the exit
9162 2018-04-09 Tom de Vries <tom@codesourcery.com>
9165 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9166 (define_expand "*memory_barrier"): New define_expand.
9167 (define_insn "memory_barrier"): New insn.
9169 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9171 PR rtl-optimization/80463
9172 PR rtl-optimization/83972
9173 PR rtl-optimization/83480
9175 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9176 correct producer for the insn.
9177 (tidy_control_flow): Fixup seqnos in case of debug insns.
9179 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9181 PR rtl-optimization/83913
9183 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9184 different sched-times when merging exprs.
9186 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9188 PR rtl-optimization/83962
9190 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9191 tidy_fallthru_edge and tidy_control_flow.
9193 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9195 PR rtl-optimization/83530
9197 * sel-sched.c (force_next_insn): New global variable.
9198 (remove_insn_for_debug): When force_next_insn is true, also leave only
9199 next insn in the ready list.
9200 (sel_sched_region): When the region wasn't scheduled, make another pass
9201 over it with force_next_insn set to 1.
9203 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9205 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9207 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9208 for interrupt control.
9209 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9210 functions for interrupt control.
9211 * config/nds32/nds32-intrinsic.md: Likewise.
9212 * config/nds32/nds32_intrinsic.h: Likewise.
9213 * config/nds32/nds32.h (nds32_builtins): Likewise.
9215 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9217 * config/nds32/nds32.c (nds32_init_machine_status,
9218 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9219 strict_aligned_p field.
9220 (nds32_expand_to_rtl_hook): New function.
9221 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9222 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9224 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9225 Chung-Ju Wu <jasonwucj@gmail.com>
9227 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9228 * config/nds32/nds32-n7.md: New file.
9229 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9230 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9232 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9233 * config/nds32/nds32.md (pipeline_model): Add n7.
9234 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9235 * config/nds32/pipelines.md: Include n7 settings.
9237 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9238 Chung-Ju Wu <jasonwucj@gmail.com>
9240 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9241 * config/nds32/nds32-e8.md: New file.
9242 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9243 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9245 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9246 * config/nds32/nds32.md (pipeline_model): Add e8.
9247 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9248 * config/nds32/pipelines.md: Include e8 settings.
9250 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9251 Chung-Ju Wu <jasonwucj@gmail.com>
9253 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9254 * config/nds32/nds32-n8.md: New file.
9255 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9256 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9258 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9259 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9260 * config/nds32/nds32.md (pipeline_model): Add n8.
9261 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9262 * config/nds32/pipelines.md: Include n8 settings.
9264 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9265 Chung-Ju Wu <jasonwucj@gmail.com>
9267 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9268 * config/nds32/nds32-n9-2r1w.md: New file.
9269 * config/nds32/nds32-n9-3r2w.md: New file.
9270 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9271 nds32_register_ports): New or modify for cpu n9.
9272 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9274 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9275 * config/nds32/nds32-utils.c: New file.
9276 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9277 TARGET_MUL_SLOW): Define.
9278 * config/nds32/nds32.md (pipeline_model): New attribute.
9279 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9280 New options that support cpu n9.
9281 * config/nds32/pipelines.md: Include n9 settings.
9282 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9284 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9286 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9287 information if necessary.
9288 (output_cond_branch_compare_zero): Likewise.
9289 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9290 (nds32_target_alignment): Refine for alignment.
9291 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9292 (FUNCTION_BOUNDARY): Modify.
9293 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9295 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9297 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9299 * config/nds32/constants.md (unspec_volatile_element): Add values for
9300 TLB operation and data prefetch.
9301 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9302 functions for TLB operation and data prefetch.
9303 * config/nds32/nds32-intrinsic.md: Likewise.
9304 * config/nds32/nds32_intrinsic.h: Likewise.
9305 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9306 (nds32_print_operand): Likewise.
9307 * config/nds32/nds32.h (nds32_builtins): Likewise.
9309 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9310 Andrew Pinski <pinsika@gcc.gnu.org>
9313 * match.pd: Use constant_boolean_node of correct type instead of
9314 boolean_true_node or boolean_false_node for simplifying
9315 pointer comparisons to zero.
9317 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9319 PR tree-optimization/80021
9320 * tree.c (verify_type_variant): Make error call in verify_variant_match
9321 translatable and remove final full stop.
9323 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9325 * config/nds32/constants.md (unspec_volatile_element): Add
9326 UNSPEC_VOLATILE_EH_RETURN.
9327 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9328 nds32_output_stack_pop): Support dwarf exception handling process.
9329 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9330 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9331 exception handling process.
9332 (nds32_compute_stack_frame): Likewise.
9333 (nds32_return_addr_rtx): Likewise.
9334 (nds32_initial_elimination_offset): Likewise.
9335 (nds32_expand_prologue): Likewise.
9336 (nds32_expand_epilogue): Likewise.
9337 (nds32_dynamic_chain_address): New function.
9338 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9340 (DYNAMIC_CHAIN_ADDRESS): Define.
9341 (EH_RETURN_DATA_REGNO): Define.
9342 (EH_RETURN_STACKADJ_RTX): Define.
9343 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9344 patterns for dwarf exception handling.
9346 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9348 * config/nds32/nds32.h: Clean up obsolete macros.
9350 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9352 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9353 Add enum values for particular instructions.
9354 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9355 particular intrinsic functions.
9356 * config/nds32/nds32-intrinsic.md: Likewise.
9357 * config/nds32/nds32_intrinsic.h: Likewise.
9358 * config/nds32/nds32.h (nds32_builtins): Likewise.
9359 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9360 (btst, ave): New patterns for particular instructions.
9362 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9364 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9365 Add enum values for atomic load/store and memory sync.
9366 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9368 * config/nds32/nds32-intrinsic.md: Likewise.
9369 * config/nds32/nds32_intrinsic.h: Likewise.
9370 * config/nds32/nds32.h (nds32_builtins): Likewise.
9372 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9374 PR tree-optimization/85257
9375 * fold-const.c (native_encode_vector): If not all elts could fit
9376 and off is -1, return 0 rather than offset.
9377 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9378 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9379 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9380 adjust buffer in native_interpret_expr call.
9382 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9384 * config/nds32/constants.md (unspec_volatile_element): Add cache
9385 control enum values.
9386 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9387 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9388 * config/nds32/nds32.c (nds32_cctl_names): New.
9389 (nds32_print_operand): Handle cache control register names.
9390 * config/nds32/nds32.h (nds32_builtins): New enum values.
9391 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9393 * config/nds32/nds32.md (type): Add mmu.
9394 * config/nds32/pipelines.md (simple_insn): Add mmu.
9396 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9398 * config/nds32/nds32.md (type): Remove call.
9399 * config/nds32/pipelines.md (simple_insn): Likewise.
9401 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9403 * config/nds32/constants.md (unspec_volatile_element): Add
9404 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9405 UNSPEC_VOLATILE_FMFCFG.
9406 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9407 description for fmfcfg and fmfcsr.
9408 (bdesc_1arg): Add fmtcsr.
9409 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9410 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9411 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9412 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9413 unspec_fmfcfg): New patterns.
9414 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9415 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9416 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9417 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9418 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9419 __nds32__fmfcfg): Define.
9421 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9423 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9424 intrinsic register names.
9425 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9426 intrinsic register enum values and macros.
9428 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9430 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9431 for load/store addressing form.
9432 (nds32_print_operand_address): Likewise.
9434 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9437 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9438 based on LABEL_REF. Remove useless assertion.
9439 (pic_address_needs_scratch): Fix formatting.
9440 (sparc_legitimize_pic_address): Minor tweaks.
9441 (sparc_delegitimize_address): Adjust assertion accordingly.
9442 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9443 into symbolic_operand.
9444 (movsi_high_pic_label_ref): Likewise.
9445 (movsi_lo_sum_pic_label_ref): Likewise.
9446 (movdi_pic_label_ref): Likewise.
9447 (movdi_high_pic_label_ref): Likewise.
9448 (movdi_lo_sum_pic_label_ref): Likewise.
9450 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9452 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9453 custom LIB_SPEC setup.
9455 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9456 Kito Cheng <kito.cheng@gmail.com>
9458 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9459 * config/riscv/freebsd.h: New.
9461 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9463 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9464 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9467 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9468 Kito Cheng <kito.cheng@gmail.com>
9470 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9471 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9472 * config/nds32/nds32-protos.h (nds32_output_call,
9473 nds32_output_return): Declare.
9474 * config/nds32/nds32.md: Refine all the call and return patterns.
9476 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9479 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9480 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9482 PR rtl-optimization/84872
9483 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9484 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9487 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9489 * expr.c (copy_blkmode_to_reg): Revert 254862.
9490 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9492 2018-04-06 Richard Biener <rguenther@suse.de>
9495 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9496 after seeing a component reference with an adjacent field. Treat
9497 refs to arrays at struct end of external decls similar to
9498 refs to unconstrained commons.
9500 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9503 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9504 look through SAVE_EXPRs with non-side-effects argument. Adjust
9506 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9509 2018-04-06 Richard Biener <rguenther@suse.de>
9512 * alias.c (find_base_term): New wrapper around find_base_term
9513 unwinding CSELIB_VAL_PTR changes.
9514 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9517 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9519 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9521 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9522 constant definitions.
9523 ("nop"): lr 0,0 -> nopr r0
9524 ("nop_lr0", "nop_lr1"): New insn definitions.
9526 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9528 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9529 NDS32_V3PUSH_AVAILABLE_P macro.
9531 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9532 Chung-Ju Wu <jasonwucj@gmail.com>
9534 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9535 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9536 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9537 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9538 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9539 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9540 * config/nds32/constraints.md: New constraints and checking for hard
9541 float configuration.
9542 * config/nds32/iterators.md: New mode iterator and attribute for hard
9543 float configuration.
9544 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9546 * config/nds32/nds32-fpu.md: New file.
9547 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9548 deal with hard float code generation.
9549 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9551 (abi_type, float_reg_number): New enum type.
9552 * config/nds32/nds32-predicates.c: New predicates for hard float.
9553 * config/nds32/nds32-protos.h: Declare functions for hard float.
9554 * config/nds32/nds32.c: Implementation for hard float configuration.
9555 * config/nds32/nds32.h: Definitions for hard float configuration.
9556 * config/nds32/nds32.md: Include hard float machine description and
9557 modify patterns for hard float configuration.
9558 * config/nds32/nds32.opt: New options for hard float configuration.
9559 * config/nds32/predicates.md: New predicates for hard float
9562 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9564 * common/config/nds32/nds32-common.c
9565 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9567 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9570 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9571 CONSTRUCTOR_ELT (ctor, ...)->value.
9573 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9576 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9578 2018-04-05 Tom de Vries <tom@codesourcery.com>
9581 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9584 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9585 Kito Cheng <kito.cheng@gmail.com>
9587 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9588 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9589 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9590 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9592 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9593 Kito Cheng <kito.cheng@gmail.com>
9595 * config/nds32/constraints.md (Ufe): New memory constraint.
9596 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9597 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9598 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9600 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9601 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9603 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9605 * config/nds32/nds32.md: Use optimize_size in the condition for
9606 alu-shift instructions.
9608 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9610 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9612 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9614 * config/nds32/nds32.md (negsi2): Refine pattern.
9616 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9617 Chung-Ju Wu <jasonwucj@gmail.com>
9619 * config/nds32/iterators.md (shift_rotate): New code iterator.
9620 (shift): New code attribute.
9621 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9622 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9623 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9624 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9625 bit-wise operations.
9626 (andsi3, *andsi3): Ditto.
9627 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9628 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9629 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9630 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9631 nds32_ior_operand, nds32_xor_operand): New predicates.
9633 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9635 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9636 (addsi3, subsi3): ... this.
9638 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9640 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9642 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9644 * config/nds32/nds32.md: Adjust indention.
9646 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9648 * config/nds32/nds32.md (feature): New attribute.
9650 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9652 * config/nds32/nds32.md (subtype): New attribute.
9654 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9657 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9658 expansion to perform a bitwise AND of the argument followed by a
9659 boolean negation of the result.
9661 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9663 PR rtl-optimization/84878
9664 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9665 the basic block. Assert the use reference is not artificial and that
9666 it has an associated insn.
9668 2018-04-04 Michael Matz <matz@suse.de>
9670 * builtins.c (compute_objsize): Pass correct operand
9671 to array_at_struct_end_p.
9673 2018-04-04 Richard Biener <rguenther@suse.de>
9676 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9677 from contexts for DINFO_LEVEL_TERSE and below.
9679 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9681 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9682 resiter_operand condition.
9683 * config/nds32/nds32.md (*move<mode>): Ditto.
9685 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9686 Monk Chiang <sh.chiang04@gmail.com>
9688 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9690 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9692 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9694 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9695 Kito Cheng <kito.cheng@gmail.com>
9697 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9698 nds32_cond_code_str, output_cond_branch,
9699 output_cond_branch_compare_zero, nds32_expand_cbranch,
9700 nds32_expand_cstore, nds32_expand_movcc,
9701 nds32_output_cbranchsi4_equality_zero,
9702 nds32_output_cbranchsi4_equality_reg,
9703 nds32_output_cbranchsi4_equality_reg_or_const_int,
9704 nds32_output_cbranchsi4_greater_less_zero: New functions.
9705 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9706 nds32_expand_cstore, nds32_expand_movcc,
9707 nds32_output_cbranchsi4_equality_zero,
9708 nds32_output_cbranchsi4_equality_reg,
9709 nds32_output_cbranchsi4_equality_reg_or_const_int,
9710 nds32_output_cbranchsi4_greater_less_zero): Declare.
9711 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9712 nds32_rimm11s_operand): New predicates.
9713 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9714 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9717 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9719 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9720 * config/nds32/nds32.md: Ditto.
9721 * config/nds32/pipelines.md: Ditto.
9723 2018-04-04 Richard Biener <rguenther@suse.de>
9725 PR tree-optimization/85168
9726 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9727 propagating abnormals.
9729 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9731 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9733 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9734 Kito Cheng <kito.cheng@gmail.com>
9736 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9737 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9738 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9739 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9740 * config/nds32/nds32.md (sibcall_internal): New.
9741 (sibcall_register): Remove.
9742 (sibcall_immediate): Remove.
9743 (sibcall_value_internal): New.
9744 (sibcall_value_register): Remove.
9745 (sibcall_value_immediate): Remove.
9746 * config/nds32/predicates.md (nds32_general_register_operand): New.
9747 (nds32_call_address_operand): New.
9749 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9751 PR rtl-optimization/85167
9752 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9753 bb_defs if *split_p, instead preinitialize it to NULL.
9755 PR tree-optimization/85156
9756 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9757 evaluating the argument multiple times.
9759 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9761 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9763 (_mm_cvtpd_ps): Likewise.
9764 (_mm_cvttpd_epi32): Likewise.
9765 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9766 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9767 vector, pixel, and bool following altivec.h include.
9769 2018-04-03 Martin Sebor <msebor@redhat.com>
9771 * doc/extend.texi (Common Function Attributes): Clarify.
9772 (const attribute): Likewise.
9773 (pure attribute): Likewise.
9775 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9778 * config/i386/i386.c (ix86_expand_vector_set): Use
9779 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9781 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9783 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9784 instructions when changing rounding bits to preserve precision bits
9785 in the x87 control word.
9787 2018-04-03 Martin Liska <mliska@suse.cz>
9789 PR tree-optimization/82491
9790 * rtl.h (strip_offset_and_add): Replace += suboffset with
9791 poly_uint64 () + suboffset.
9793 2018-03-29 Martin Liska <mliska@suse.cz>
9794 Martin Jambor <mjambor@suse.cz>
9797 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9798 param_type is not an integral or pointer type.
9800 2018-04-03 Richard Biener <rguenther@suse.de>
9802 * sese.h (recompute_all_dominators): Remove.
9804 2018-04-02 Martin Sebor <msebor@redhat.com>
9806 * doc/invoke.texi (-Wrestrict): Fix typos.
9808 2018-04-02 Jim Wilson <jimw@sifive.com>
9810 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9811 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9812 (<optab>di3, <optab>si3_extend): Likewise.
9813 (<optab>si3_mask, <optab>si3_mask_1): New.
9814 (<optab>di3_mask, <optab>di3_mask_1): New.
9815 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9816 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9817 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9819 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9821 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9824 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9826 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9827 (nds32_canonicalize_comparison): New function.
9829 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9830 Kito Cheng <kito.cheng@gmail.com>
9831 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9833 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9834 * config/nds32/constants.md (unspec_volatile_element): Add
9835 UNSPEC_VOLATILE_RELAX_GROUP.
9836 * config/nds32/nds32-relax-opt.c: New file.
9837 * config/nds32/nds32-predicates.c
9838 (nds32_symbol_load_store_p): New function.
9839 * config/nds32/nds32-protos.h
9840 (nds32_symbol_load_store_p): Declare function.
9841 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9842 * config/nds32/nds32.c
9843 (nds32_register_pass): New function to register pass.
9844 (nds32_register_passes): New function to register passes.
9845 * config/nds32/nds32.md (relax_group): New pattern.
9846 * config/nds32/nds32.opt (mrelax-hint): New option.
9847 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9849 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9851 * config/nds32/t-nds32: Modify files dependency.
9853 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9855 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9856 (PROFILE_HOOK): Define its implementation.
9858 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9860 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9861 type and 32-bit size.
9863 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9866 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9867 (V_128_256): New mode iterator.
9868 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9869 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9870 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9872 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9873 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9875 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9878 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9879 NaN inputs correctly.
9881 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9884 * config/rs6000/vsx.md (??r): New mode attribute.
9885 (*vsx_mov<mode>_64bit): Use it.
9886 (*vsx_mov<mode>_32bit): Likewise.
9888 2018-03-30 Martin Sebor <msebor@redhat.com>
9890 PR tree-optimization/84818
9891 * builtins.c (check_access): Use warning_n.
9893 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9896 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9898 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9901 2018-03-30 Julia Koval <julia.koval@intel.com>
9904 * x86-tune.def (movx, partial_reg_dependency): Enable for
9907 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9910 * lra-constraints.c (process_alt_operands): Move setting
9911 this_alternative_matches below.
9913 2018-03-29 Martin Liska <mliska@suse.cz>
9916 * doc/invoke.texi: Document how LTO works with debug info.
9917 Describe auto-load support of binutils. Mention 'x86-64'
9918 as valid option value of -march option.
9920 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9922 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9925 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9926 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9927 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9930 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9933 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9934 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9935 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9936 for RS6000_BTM_POWERPC64.
9937 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9938 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9939 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9944 2018-03-28 Carl Love <cel@us.ibm.com>
9947 2017-09-27 Carl Love <cel@us.ibm.com>
9949 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9950 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9951 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9954 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9956 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9957 instead of __vector bool.
9958 (_mm_max_pu8): Likewise.
9959 (_mm_min_pi16): Likewise.
9961 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9964 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9965 (DIVWEUO): Likewise.
9967 (DIVDEUO): Likewise.
9968 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9969 DIVWEUO and DIVDEUO.
9970 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9971 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9972 (div_extend): Likewise.
9973 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9975 (__builtin_divweuo): Likewise.
9976 (__builtin_divdeo): Likewise.
9977 (__builtin_divdeuo): Likewise.
9979 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9982 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9983 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9985 PR tree-optimization/82004
9986 * gimple-match-head.c (optimize_pow_to_exp): New function.
9987 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9988 Don't fold to exp if optimize_pow_to_exp is false.
9990 2018-03-28 Martin Liska <mliska@suse.cz>
9993 * calls.c (initialize_argument_information): Fix trailing space.
9994 * common.opt: Fix typo and provide better explanation for
9995 -fsanitize-coverage option.
9996 * config/i386/i386.opt: Fix typo.
9998 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9999 Martin Liska <mliska@suse.cz>
10002 * gimplify.c (asan_poison_variable): Don't do the check for
10003 gimplify_omp_ctxp here.
10004 (gimplify_decl_expr): Do it here.
10005 (gimplify_target_expr): Likewise.
10007 2018-03-28 Martin Liska <mliska@suse.cz>
10010 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10011 chkp_type_bounds_count if MPX is not enabled.
10013 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10015 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10017 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10020 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10021 function to create the function decl for complex long double
10022 multiply and divide for -mabi=ieeelongdouble.
10023 (init_float128_ieee): Call it.
10025 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10028 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10029 -fcf-protection=branch -mibt.
10030 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10032 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10035 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10036 arm_word_relocations.
10038 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10041 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10042 extern array declarations.
10044 2018-03-27 Richard Biener <rguenther@suse.de>
10046 PR middle-end/84067
10047 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10048 explicit single_use checks.
10050 2018-03-27 Richard Biener <rguenther@suse.de>
10052 PR tree-optimization/85082
10053 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10056 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10058 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10059 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10060 Turn on fasynchronous-unwind-tables and funwind-tables.
10062 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10065 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10066 (*bmi_blsr_<mode>_ccz): Ditto.
10068 2018-03-26 Tom de Vries <tom@codesourcery.com>
10070 PR tree-optimization/85063
10071 * omp-general.c (offloading_function_p): New function. Factor out
10073 * omp-offload.c (pass_omp_target_link::gate): ... here.
10074 * omp-general.h (offloading_function_p): Declare.
10075 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10076 with attribute omp declare target for offloading functions.
10078 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10080 PR tree-optimization/84005
10081 * tree-data-ref.h (get_base_for_alignment): Declare.
10082 * tree-data-ref.c (get_base_for_alignment_1): New function.
10083 (get_base_for_alignment): Likewise.
10084 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10085 get_base_for_alignment to find a suitable base object, instead
10086 of always using drb->base_address.
10088 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10090 PR inline-asm/85022
10091 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10092 known size by default.
10094 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10096 PR inline-asm/85030
10097 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10098 and non BLKmode operands.
10100 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10103 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10104 Clean up attributes.
10106 2018-03-23 Richard Biener <rguenther@suse.de>
10109 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10110 we are going to emit early debug for LTO.
10112 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10114 PR inline-asm/85034
10115 * function.c (match_asm_constraints_1): Don't optimize if input
10116 doesn't satisfy general_operand predicate for output's mode.
10118 PR inline-asm/85022
10119 * alias.c (write_dependence_p): Don't require for x_canonicalized
10120 non-VOIDmode if x has VOIDmode.
10123 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10124 just don't try to optimize it rather than assert it never happens.
10126 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10128 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10129 macro expansions for definition of ST_INTERNAL_<mode> and
10130 LD_INTERNAL_<mode> builtins.
10131 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10133 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10135 (altivec_expand_st_builtin): Likewise.
10136 (altivec_expand_builtin): Remove calls to deleted functions.
10137 (rs6000_address_for_altivec): Delete this function.
10138 * config/rs6000/vector.md: Remove expands for
10139 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10141 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10144 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10145 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10146 re-computing once computed.
10147 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10148 (arm_init_machine_status): Initialize
10149 machine->static_chain_stack_bytes.
10151 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10154 * doc/extend.texi: Add four new prototypes for vec_ld.
10155 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10156 definitions for more logical presentation.
10157 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10158 entries for V1TI variants of __builtin_altivec_ld builtin.
10159 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10160 handling of V1TI variant of LVX icode pattern.
10161 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10162 (rs6000_gimple_fold_builtin): Likewise.
10163 (altivec_init_builtins): Add code to define
10164 __builtin_altivec_lvx_v1ti function.
10166 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10168 PR inline-asm/84941
10169 * function.c (match_asm_constraints_1): Don't do the optimization
10170 if input isn't a REG, SUBREG, MEM or constant.
10172 2018-03-22 Tom de Vries <tom@codesourcery.com>
10174 PR tree-optimization/84956
10175 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10176 bb_has_abnormal_pred.
10178 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10181 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10182 DECL_INITIAL (decl) to decl at the end.
10183 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10184 adjust the comment.
10186 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10188 * doc/extend.texi (__builtin_tgmath): Document when complex
10189 integer types are treated as _Complex _Float64.
10191 2018-03-21 Tom de Vries <tom@codesourcery.com>
10193 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10195 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10197 PR tree-optimization/84960
10198 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10199 if it is ENTRY block, move them into single succ of ENTRY in that case.
10201 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10203 PR tree-optimization/84811
10204 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10205 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10206 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10207 (known_subrange_p): Update accordingly. Cast each value involved
10208 in the size comparison, rather than casting the result of the
10211 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10213 PR tree-optimization/84982
10214 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10215 by flipping the least significant bit rather than all bits from
10216 bitpos to bitpos + bitsize - 1.
10218 2018-03-21 Nathan Sidwell <nathan@acm.org>
10220 * doc/extend.texi (Deprecated Features): Remove mention of
10221 long-deleted deprecations.
10223 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10226 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10227 * configure: Regenerate.
10229 2018-03-21 Tom de Vries <tom@codesourcery.com>
10231 PR tree-optimization/83126
10232 * tree-parloops.c (num_phis): New function.
10233 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10235 2018-03-21 Nathan Sidwell <nathan@acm.org>
10237 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10238 mention anon-struct/union members and trailing attributes.
10240 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10242 PR tree-optimization/84969
10243 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10244 builtin memset partitions if they set different rhs values.
10246 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10248 PR rtl-optimization/84989
10249 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10250 VEC_DUPLICATE with scalar result mode.
10252 2018-03-21 Martin Liska <mliska@suse.cz>
10255 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10256 not intended return statement.
10258 2018-03-21 Martin Liska <mliska@suse.cz>
10261 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10262 (chkp_find_bound_slots_1): Limit number of iterations.
10264 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10267 * Minor grammar fixes for x86 options.
10269 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10272 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10273 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10275 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10278 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10279 (altivec_lvx_<mode>_1op): Likewise.
10280 (altivec_stvx_<mode>_2op): Likewise.
10281 (altivec_stvx_<mode>_1op): Likewise.
10282 (altivec_lvx_<VM2:mode>): New define_expand.
10283 (altivec_stvx_<VM2:mode>): Likewise.
10284 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10285 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10286 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10287 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10288 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10289 (rs6000_gen_lvx): Likewise.
10290 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10291 (altivec_expand_stv_builtin): Likewise.
10292 (altivec_expand_builtin): Likewise.
10293 * config/rs6000/vector.md: Likewise.
10295 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10298 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10301 2018-03-20 Richard Biener <rguenther@suse.de>
10304 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10305 sign-conversions as zero, fall back to standard scalar_stmt
10308 2018-03-20 Martin Liska <mliska@suse.cz>
10311 * predict.c (rebuild_frequencies): Handle case when we have
10312 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10314 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10317 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10318 flag_section_anchors.
10319 * varasm.c (use_blocks_for_decl_p): Remove hack for
10320 dw2_force_const_mem.
10323 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10325 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10326 be created, use lowpart_subreg of operands[0] rather than operands[0]
10328 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10329 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10330 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10331 and n constraint instead of aarch64_shift_imm_di and Usd.
10332 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10333 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10335 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10338 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10339 to favor GPR over NEON registers.
10340 (<shift>di3_neon): Likewise.
10342 2018-03-20 Tom de Vries <tom@codesourcery.com>
10345 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10346 (nvptx_process_pars): Emit bar.sync asap and alap.
10348 2018-03-20 Tom de Vries <tom@codesourcery.com>
10351 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10352 seen_label if seen_label is already set.
10354 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10357 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10358 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10359 Use 1U instead of 1. Formatting fixes.
10362 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10363 instead of TREE_TYPE (s1) for the return value.
10365 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10367 PR tree-optimization/84946
10368 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10369 bitsize + bitsize in poly_uint64 rather than poly_int64.
10372 * dwarf2asm.c: Include fold-const.c.
10373 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10374 of decl rather than decl itself.
10376 PR rtl-optimization/84643
10377 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10379 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10382 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10383 calling assemble_variable.
10385 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10388 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10389 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10391 2018-03-19 Jim Wilson <jimw@sifive.com>
10394 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10395 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10396 (riscv_first_stack_step): Likewise.
10397 (riscv_option_override): Use STACK_BOUNDARY instead of
10398 MIN_STACK_BOUNDARY.
10399 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10400 MIN_STACK_BOUNDARY.
10401 (BIGGEST_ALIGNMENT): Set to 128.
10402 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10403 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10406 2018-03-19 Richard Biener <rguenther@suse.de>
10408 PR tree-optimization/84933
10409 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10410 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10412 2018-03-19 Richard Biener <rguenther@suse.de>
10414 PR tree-optimization/84859
10415 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10416 (cond_if_else_store_replacement): Perform sinking operation on
10417 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10418 Generalize what a BB with a single eligible store is.
10420 2018-03-19 Richard Biener <rguenther@suse.de>
10422 PR tree-optimization/84929
10423 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10424 chrec_is_positive against non-chrec arg.
10426 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10429 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10431 2018-03-18 Martin Liska <mliska@suse.cz>
10433 PR rtl-optimization/84635
10434 * regrename.c (build_def_use): Use matches_mode only when
10437 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10439 PR tree-optimization/84913
10440 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10441 vectorize chains of COND_EXPRs.
10443 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10445 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10447 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10449 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10451 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10453 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10455 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10456 Kito Cheng <kito.cheng@gmail.com>
10458 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10459 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10460 (nds32_adjust_reg_alloc_order): New function.
10461 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10463 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10465 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10466 nds32_print_operand, nds32_print_operand_address): Use
10467 HOST_WIDE_INT_PRINT_DEC instead.
10469 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10471 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10473 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10476 * config/i386/i386.c (initial_ix86_tune_features,
10477 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10478 unsigned long long.
10479 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10480 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10481 rather than 1u << ix86_tune. Formatting fix.
10482 (ix86_option_override_internal): Change ix86_arch_mask from
10483 unsigned int to unsigned HOST_WIDE_INT, initialize to
10484 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10485 (ix86_function_specific_restore): Likewise.
10487 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10490 * postreload.c (reload_combine_recognize_pattern): Perform
10491 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10492 truncate_int_for_mode the result for the destination's mode.
10495 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10496 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10498 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10501 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10504 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10505 regno_allocno_class_array and sorted_pseudos.
10506 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10507 insns where regno is used.
10509 2018-03-16 Martin Liska <mliska@suse.cz>
10512 * multiple_target.c (create_dispatcher_calls): Redirect
10513 reference in the symbol table.
10515 2018-03-16 Martin Liska <mliska@suse.cz>
10518 * multiple_target.c (create_dispatcher_calls): Redirect also
10521 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10525 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10526 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10527 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10529 2018-03-16 Julia Koval <julia.koval@intel.com>
10531 * doc/invoke.texi (Skylake Server): Add CLWB.
10532 Cannonlake): Remove CLWB.
10534 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10536 PR tree-optimization/84841
10537 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10539 (FLOAT_ONE_CONST_TYPE): Define.
10540 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10541 (sort_by_operand_rank): Put entries with higher constant_type last
10542 rather than first to match comments.
10544 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10546 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10549 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10552 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10553 check and use address of target always.
10555 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10558 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10559 (indirect_thunk_bnd_needed): Likewise.
10560 (indirect_thunks_used): Likewise.
10561 (indirect_thunks_bnd_used): Likewise.
10562 (indirect_return_needed): New.
10563 (indirect_return_bnd_needed): Likewise.
10564 (output_indirect_thunk_function): Add a bool argument for
10566 (output_indirect_thunk_function): Don't generate alias for
10567 function return thunk.
10568 (ix86_code_end): Call output_indirect_thunk_function to generate
10569 function return thunks.
10570 (ix86_output_function_return): Set indirect_return_bnd_needed
10571 and indirect_return_needed instead of indirect_thunk_bnd_needed
10572 and indirect_thunk_needed.
10574 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10576 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10577 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10578 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10580 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10581 Paul Hua <paul.hua.gm@gmail.com>
10584 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10586 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10588 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10589 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10590 resp. SFmode cases.
10592 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10595 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10596 instead of GET_MODE_SIZE when comparing Units.
10598 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10601 * varasm.c (hash_section): Return an unchangeble hash value
10602 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10603 Return !aarch64_can_use_per_function_literal_pools_p ().
10605 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10608 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10609 rather than address of cmode as last argument to prepare_cmp_insn.
10611 2018-03-15 Julia Koval <julia.koval@intel.com>
10613 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10614 F_AVX512VNNI, F_AVX512BITALG): New.
10616 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10619 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10620 insn for floating-point loads and stores.
10622 2018-03-14 Carl Love <cel@us.ibm.com>
10624 * config/rs6000/rs6000-c.c: Add macro definitions for
10625 ALTIVEC_BUILTIN_VEC_PERMXOR.
10626 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10627 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10628 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10629 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10631 * config/doc/extend.texi: Add prototypes for vec_permxor.
10633 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10636 * diagnostic-show-locus.c (class layout_point): Convert m_line
10637 from int to linenum_type.
10638 (line_span::comparator): Use linenum "compare" function when
10639 comparing line numbers.
10640 (test_line_span): New function.
10641 (layout_range::contains_point): Convert param "row" from int to
10643 (layout_range::intersects_line_p): Likewise.
10644 (layout::will_show_line_p): Likewise.
10645 (layout::print_source_line): Likewise.
10646 (layout::should_print_annotation_line_p): Likewise.
10647 (layout::print_annotation_line): Likewise.
10648 (layout::print_leading_fixits): Likewise.
10649 (layout::annotation_line_showed_range_p): Likewise.
10650 (struct line_corrections): Likewise for field m_row.
10651 (line_corrections::line_corrections): Likewise for param "row".
10652 (layout::print_trailing_fixits): Likewise.
10653 (layout::get_state_at_point): Likewise.
10654 (layout::get_x_bound_for_row): Likewise.
10655 (layout::print_line): Likewise.
10656 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10657 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10658 * input.c (selftest::test_linenum_comparisons): New function.
10659 (selftest::input_c_tests): Call it.
10660 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10661 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10662 * selftest.h (ASSERT_GT): New macro.
10663 (ASSERT_GT_AT): New macro.
10664 (ASSERT_LT): New macro.
10665 (ASSERT_LT_AT): New macro.
10667 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10669 PR rtl-optimization/84780
10670 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10672 2018-03-14 Martin Liska <mliska@suse.cz>
10674 * tree.c (record_node_allocation_statistics): Use
10675 get_stats_node_kind.
10676 (get_stats_node_kind): New function extracted from
10677 record_node_allocation_statistics.
10678 (free_node): Use get_stats_node_kind.
10680 2018-03-14 Richard Biener <rguenther@suse.de>
10682 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10683 that the value-set of ANTIC_IN doesn't grow.
10686 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10688 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10689 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10691 2018-03-14 Julia Koval <julia.koval@intel.com>
10693 * config.gcc (icelake-client, icelake-server): New.
10695 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10696 (initial_ix86_arch_features): Ditto.
10697 (PTA_SKYLAKE): Add SGX.
10698 (PTA_ICELAKE): Remove.
10699 (PTA_ICELAKE_CLIENT): New.
10700 (PTA_ICELAKE_SERVER): New.
10701 (ix86_option_override_internal): Split up icelake on icelake client and
10703 (get_builtin_code_for_version): Ditto.
10704 (fold_builtin_cpu): Ditto.
10705 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10706 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10707 * config/i386/i386.h (processor_type): Ditto.
10708 * doc/invoke.texi: Ditto.
10710 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10713 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10714 INTEGER_CST offset, add it together with bitpos / 8 and
10715 sign extend based on POINTER_SIZE.
10719 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10722 * config/i386/constraints.md (Yc): New register constraint.
10723 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10724 Use Yc constraint for alternative 2 of operand 0. Remove
10725 preferred_for_speed attribute.
10727 2018-03-14 Richard Biener <rguenther@suse.de>
10729 PR tree-optimization/84830
10730 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10731 with the old one to avoid oscillations.
10733 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10736 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10738 (assign_by_spills): Return a flag of reload assignment failure.
10739 Do not process the reload assignment failures. Do not spill other
10740 reload pseudos if they has the same reg class. Update n if
10742 (lra_assign): Add a return arg. Set up from the result of
10743 assign_by_spills call.
10744 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10745 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10746 usage_insns if it is not NULL.
10747 (spill_hard_reg_in_range): New function.
10748 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10749 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10750 function prototypes.
10751 (lra_assign): Change prototype.
10752 * lra.c (lra): Add code to deal with fails by splitting hard reg
10755 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10757 * config/riscv/riscv.opt (mrelax): New option.
10758 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10759 "norelax" when riscv_mrelax is disabled.
10760 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10762 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10765 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10766 reassociation for int modes.
10768 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10770 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10771 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10773 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10774 * config/aarch64/aarch64-sve.md
10775 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10776 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10777 (*extend<mode><Vwide>2): Rename to...
10778 (aarch64_sve_extend<mode><Vwide>2): ...this.
10779 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10780 renaming the old pattern to...
10781 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10783 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10784 define_expand, renaming the old pattern to...
10785 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10786 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10787 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10788 account when deciding which SVE instruction the optab should use.
10789 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10791 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10793 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10794 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10795 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10796 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10797 (tlsdesc_small_<mode>): Turn a define_expand and use
10798 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10799 (tlsdesc_small_advsimd_<mode>): ...this.
10800 (tlsdesc_small_sve_<mode>): New pattern.
10802 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10804 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10805 (UNSPEC_UMUL_HIGHPART): New constants.
10806 (MUL_HIGHPART): New int iteraor.
10807 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10808 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10810 (*<su>mul<mode>3_highpart): New define_insn.
10812 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10815 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10818 2018-03-13 Martin Liska <mliska@suse.cz>
10821 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10823 (sem_item_optimizer::~sem_item_optimizer): Release it.
10824 (sem_item_optimizer::merge_classes): Register variable aliases.
10825 (sem_item_optimizer::fixup_pt_set): New function.
10826 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10827 * ipa-icf.h: Declare new variables and functions.
10829 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10831 PR middle-end/84834
10832 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10833 integer_pow2p@2 and test integer_pow2p in condition.
10834 (A < 0 ? C : 0): Similarly for @1.
10836 PR middle-end/84831
10837 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10838 characters starting at p contain '\0' character, don't look beyond
10842 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10843 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10846 * reg-stack.c (change_stack): Change update_end var from int to
10847 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10848 also call set_block_for_insn on the newly added insns and rescan.
10851 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10852 on the last operand.
10855 * tree.c (stabilize_reference_1): Return save_expr (e) for
10856 STATEMENT_LIST even if it doesn't have side-effects.
10858 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10860 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10862 2018-03-12 Renlin Li <renlin.li@arm.com>
10864 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10865 aarch64_output_scalar_simd_mov_immediate.
10867 2018-03-12 Martin Sebor <msebor@redhat.com>
10869 PR tree-optimization/83456
10870 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10871 for perfectly overlapping calls to memcpy.
10872 (gimple_fold_builtin_memory_chk): Same.
10873 (gimple_fold_builtin_strcpy): Handle no-warning.
10874 (gimple_fold_builtin_stxcpy_chk): Same.
10875 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10877 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10879 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10880 parameter. Use it for SFmode.
10881 (rs6000_function_arg_advance_1): Adjust.
10882 (rs6000_function_arg): Adjust.
10883 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10885 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10887 PR rtl-optimization/84169
10888 PR rtl-optimization/84780
10889 * combine.c (can_combine_p): Check for a 2-insn combination whether
10890 the destination register is used between the two insns, too.
10892 2018-03-12 Richard Biener <rguenther@suse.de>
10894 PR tree-optimization/84803
10895 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10896 for refs DR analysis didn't process.
10898 2018-03-12 Richard Biener <rguenther@suse.de>
10900 PR tree-optimization/84777
10901 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10902 force-vectorize loops ignore whether we are optimizing for size.
10904 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10906 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10907 (TARGET_MD_ASM_ADJUST): Define.
10909 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10910 Kito Cheng <kito.cheng@gmail.com>
10911 Chung-Ju Wu <jasonwucj@gmail.com>
10913 * config/nds32/nds32.c (nds32_compute_stack_frame,
10914 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10915 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10916 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10917 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10918 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10919 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10920 * config/nds32/nds32.md (prologue, epilogue): Use macro
10921 NDS32_V3PUSH_AVAILABLE_P to do checking.
10923 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10926 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10927 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10928 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10929 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10930 addition of most attributes on !orig_type_die or the attribute not
10931 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10933 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10934 Chung-Ju Wu <jasonwucj@gmail.com>
10936 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10937 __NDS32_VH__ macro.
10938 * config/nds32/nds32.opt (mvh): New option.
10940 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10941 Chung-Ju Wu <jasonwucj@gmail.com>
10943 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10945 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10946 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10949 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10950 Chung-Ju Wu <jasonwucj@gmail.com>
10952 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10954 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10955 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10957 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10958 Kito Cheng <kito.cheng@gmail.com>
10959 Chung-Ju Wu <jasonwucj@gmail.com>
10961 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10962 UNSPEC_FFMISM and UNSPEC_FLMISM.
10963 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10964 for ffb, ffmism and flmism.
10965 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10966 (unspec_ffmism): Ditto.
10967 (unspec_flmism): Ditto.
10968 (nds32_expand_builtin_impl): Check if string extension is available.
10969 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10970 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10972 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10975 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10978 * lra-assigns.c (assign_by_spills): Return a flag of reload
10979 assignment failure. Do not process the reload assignment
10980 failures. Do not spill other reload pseudos if they has the same
10982 (lra_assign): Add a return arg. Set up from the result of
10983 assign_by_spills call.
10984 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10985 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10986 usage_insns if it is not NULL.
10987 (spill_hard_reg_in_range): New function.
10988 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10989 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10990 function prototypes.
10991 (lra_assign): Change prototype.
10992 * lra.c (lra): Add code to deal with fails by splitting hard reg
10995 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10998 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11000 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11003 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11004 (dw_val_node): Add val_symbolic_view.
11005 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11006 (symview_upper_bound): New.
11007 (new_line_info_table): Initialize symviews_since_reset.
11008 (dwarf2out_source_line): Count symviews_since_reset and set
11009 symview_upper_bound.
11010 (dw_val_equal_p): Handle symview.
11011 (add_AT_symview): New.
11012 (print_dw_val): Handle symview.
11013 (attr_checksum, attr_checksum_ordered): Likewise.
11014 (same_dw_val_p, size_of_die): Likewise.
11015 (value_format, output_die): Likewise.
11016 (add_high_low_attributes): Use add_AT_symview for entry_view.
11017 (dwarf2out_finish): Reset symview_upper_bound, clear
11020 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11023 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11024 Add strict argument and use it.
11025 (rs6000_split_multireg_move): Update for new strict argument.
11026 (mem_operand_gpr): Disallow all non-offsettable addresses.
11027 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11029 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11032 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11033 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11034 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11037 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11038 decl, use remap_type if we want to use the type.
11040 2018-03-09 Martin Sebor <msebor@redhat.com>
11042 PR tree-optimization/84526
11043 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11045 (builtin_access::generic_overlap): Be prepared to handle non-array
11048 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11050 PR rtl-optimization/84682
11051 * lra-constraints.c (process_address_1): Check is_address flag
11052 for address constraints.
11053 (process_alt_operands): Likewise.
11054 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11055 preprocess_constraints.
11056 * recog.h (preprocess_constraints): Add oploc parameter.
11058 * recog.c (preprocess_constraints): Test address_operand for
11059 CT_ADDRESS constraints.
11061 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11064 * lra-assigns.c (assign_by_spills): Return a flag of reload
11065 assignment failure. Do not process the reload assignment
11066 failures. Do not spill other reload pseudos if they has the same
11068 (lra_assign): Add a return arg. Set up from the result of
11069 assign_by_spills call.
11070 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11071 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11072 usage_insns if it is not NULL.
11073 (spill_hard_reg_in_range): New function.
11074 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11075 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11076 function prototypes.
11077 (lra_assign): Change prototype.
11078 * lra.c (lra): Add code to deal with fails by splitting hard reg
11081 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11084 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11085 Accept complain bool parameter. Only emit errors if it is true.
11086 (arm_parse_cpu_option_name): Likewise.
11087 (arm_target_thumb_only): Adjust callers of the above.
11088 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11089 prototype to take a default true bool parameter.
11090 (arm_parse_arch_option_name): Likewise.
11092 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11093 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11097 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11098 * configure: Regenerate.
11099 * configure.ac ("linker --version-script option"): New.
11100 ("linker soname option"): New.
11102 2018-03-09 Richard Biener <rguenther@suse.de>
11104 PR tree-optimization/84775
11105 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11106 immediate uses of predicate stmts and mark them modified.
11109 PR tree-optimization/84178
11110 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11112 (version_loop_for_if_conversion): Delay update_ssa call.
11113 (tree_if_conversion): Delay update_ssa until after predicate
11116 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11119 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11120 when the function accesses prior frames.
11122 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11125 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11126 gen_llsym, otherwise call maybe_gen_llsym.
11128 PR inline-asm/84742
11129 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11130 has ',' character inside of it.
11132 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11135 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11136 as clobbering CC_REGNUM.
11138 2018-03-08 Richard Biener <rguenther@suse.de>
11140 PR middle-end/84552
11141 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11142 (follow_copies_to_constant): Do not follow SSA names registered
11145 2018-03-08 Richard Biener <rguenther@suse.de>
11147 PR tree-optimization/84178
11148 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11150 (version_loop_for_if_conversion): Delay update_ssa call.
11151 (tree_if_conversion): Delay update_ssa until after predicate
11154 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11156 PR tree-optimization/84178
11157 * tree-if-conv.c (release_bb_predicate): Remove the
11158 the assertion that the stmts have NULL use_ops.
11159 Discard the statements, asserting that they haven't
11160 yet been added to a BB.
11162 2018-03-08 Richard Biener <rguenther@suse.de>
11164 PR tree-optimization/84746
11165 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11166 (phi_translate): Pass in destination ANTIC_OUT set.
11167 (phi_translate_1): Likewise. For a simplified result lookup
11168 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11169 (phi_translate_set): Adjust.
11170 (do_pre_regular_insertion): Likewise.
11171 (do_pre_partial_partial_insertion): Likewise.
11173 2018-03-08 Martin Liska <mliska@suse.cz>
11175 PR gcov-profile/84735
11176 * doc/gcov.texi: Document usage of profile files.
11177 * gcov-io.h: Document changes in the format.
11179 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11183 * dwarf2out.c (struct dw_line_info_table): Update comments for
11185 (FORCE_RESET_NEXT_VIEW): New.
11186 (FORCE_RESETTING_VIEW_P): New.
11187 (RESETTING_VIEW_P): Check for -1 too.
11188 (ZERO_VIEW_P): Likewise.
11189 (new_line_info_table): Force-reset next view.
11190 (dwarf2out_begin_function): Likewise.
11191 (dwarf2out_source_line): Simplify zero_view_p initialization.
11192 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11193 view directly. Omit view when omitting .loc at line 0.
11195 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11197 PR tree-optimization/84740
11198 * tree-switch-conversion.c (process_switch): Call build_constructors
11199 only if info.phi_count is non-zero.
11201 PR tree-optimization/84739
11202 * tree-tailcall.c (find_tail_calls): Check call arguments against
11203 DECL_ARGUMENTS (current_function_decl) rather than
11204 DECL_ARGUMENTS (func) when checking for tail recursion.
11206 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11208 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11209 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11210 Volker Reichelt's entry and add entries for people that perform
11211 GCC fuzzy testing and report numerous bugs.
11213 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11216 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11217 readonly data in sdata, if that is disabled.
11218 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11219 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11220 -mreadonly-in-sdata option.
11222 2018-03-07 Martin Sebor <msebor@redhat.com>
11224 PR tree-optimization/84468
11225 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11226 basic block when looking for nul assignment.
11228 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11231 * except.h (output_function_exception_table): Adjust prototype.
11232 * except.c (output_function_exception_table): Remove FNNAME parameter
11233 and add SECTION parameter. Ouput one part of the table at a time.
11234 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11235 the first part of the exception table and emit unwind directives.
11236 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11237 (i386_pe_seh_cold_init): Likewise.
11238 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11239 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11240 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11241 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11242 * config/i386/winnt.c: Include except.h.
11243 (struct seh_frame_state): Add reg_offset, after_prologue and
11244 in_cold_section fields.
11245 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11246 (i386_pe_seh_cold_init): New function.
11247 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11248 to seh->in_cold_section.
11249 (seh_emit_push): Record the offset of the push.
11250 (seh_emit_save): Record the offet of the save.
11251 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11252 Test seh->after_prologue to disregard the epilogue.
11253 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11254 (i386_pe_end_cold_function): New function.
11256 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11259 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11260 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11263 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11265 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11266 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11268 PR middle-end/84723
11269 * multiple_target.c: Include tree-inline.h and intl.h.
11270 (expand_target_clones): Diagnose and fail if node->definition and
11271 !tree_versionable_function_p (node->decl).
11273 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11275 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11277 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11278 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11279 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11281 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11284 * combine.c (try_combine): Use reg_or_subregno instead of handling
11285 just paradoxical SUBREGs and REGs.
11287 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11289 * config/arc/arc.c (arc_finalize_pic): Remove function.
11290 (arc_must_save_register): We use single base PIC register, remove
11291 checks to save/restore the PIC register.
11292 (arc_expand_prologue): Likewise.
11293 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11295 (arc_verify_short): Likewise.
11296 (arc_attr_type): Likewise.
11297 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11298 (walk_stores): Likewise.
11299 (arc_address_cost): Make it static.
11300 (arc_verify_short): Likewise.
11301 (branch_dest): Likewise.
11302 (arc_attr_type): Likewise.
11303 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11304 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11305 (arc_final_prescan_insn): Remove inserting the nops due to
11306 hardware hazards. It is done in reorg step.
11307 (insn_length_variant_t): Remove.
11308 (insn_length_parameters_t): Likewise.
11309 (arc_insn_length_parameters): Likewise.
11310 (arc_get_insn_variants): Likewise.
11311 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11313 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11315 PR inline-asm/84683
11316 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11319 PR tree-optimization/84687
11320 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11322 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11324 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11326 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11327 Rename to ppc_speculation_barrier.
11328 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11329 __builtin_ppc_speculation_barrier.
11331 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11334 * combine.c (combine_simplify_rtx): Don't try to simplify if
11335 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11338 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11340 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11341 to 32 bytes when compiling for POWER9.
11343 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11346 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11347 regparm >= 3 with no arg reg available also for calls with
11348 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11351 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11353 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11355 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11358 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11360 2018-03-05 Richard Biener <rguenther@suse.de>
11362 PR tree-optimization/84486
11363 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11364 When inserting a __builtin_assume_aligned call set the LHS
11365 SSA name alignment info accordingly.
11367 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11369 PR tree-optimization/84114
11370 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11371 Avoid reassociation of FLOAT_MODE addition.
11373 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11375 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11376 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11377 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11378 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11379 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11380 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11383 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11384 __builtin_ia32_wbinvd): New builtins.
11385 (SPECIAL_ARGS2): New.
11386 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11387 (SPECIAL_ARGS2): New.
11388 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11389 (ix86_valid_target_attribute_inner_p): Ditto.
11390 (ix86_init_mmx_sse_builtins): Add special_args2.
11391 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11392 TARGET_WBNOINVD_P): New.
11393 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11394 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11395 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11396 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11397 * config/i386/pconfigintrin.h: New file.
11398 * config/i386/wbnoinvdintrin.h: Ditto.
11399 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11401 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11403 2018-03-05 Richard Biener <rguenther@suse.de>
11405 PR tree-optimization/84670
11406 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11408 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11409 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11410 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11411 doesn't grow if all successors (recursively) were visited at least
11414 2018-03-05 Richard Biener <rguenther@suse.de>
11416 PR tree-optimization/84650
11417 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11418 if executed in the loop pipeline.
11420 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11422 * doc/configfiles.texi (Configuration Files): Move info about
11423 conditionalizing $target-protos.h to...
11424 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11425 differs from $target-protos.h.
11427 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11428 Chung-Ju Wu <jasonwucj@gmail.com>
11430 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11431 * config/nds32/nds32-multiple.md (setmemsi): Define.
11432 * config/nds32/nds32-memory-manipulation.c
11433 (nds32_gen_dup_4_byte_to_word_value): New.
11434 (emit_setmem_word_loop): New.
11435 (emit_setmem_byte_loop): New.
11436 (nds32_expand_setmem_loop): New.
11437 (nds32_expand_setmem_loop_v3m): New.
11438 (nds32_expand_setmem_unroll): New.
11439 (nds32_expand_setmem): New.
11441 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11442 Chung-Ju Wu <jasonwucj@gmail.com>
11444 * config/nds32/nds32-memory-manipulation.c
11445 (nds32_emit_load_store): New.
11446 (nds32_emit_post_inc_load_store): New.
11447 (nds32_emit_mem_move): New.
11448 (nds32_emit_mem_move_block): New.
11449 (nds32_expand_movmemsi_loop_unknown_size): New.
11450 (nds32_expand_movmemsi_loop_known_size): New.
11451 (nds32_expand_movmemsi_loop): New.
11452 (nds32_expand_movmemsi_unroll): New.
11453 (nds32_expand_movmemqi): Rename ...
11454 (nds32_expand_movmemsi): ... to this.
11455 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11456 (movmemsi): ... to this.
11457 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11458 (nds32_expand_movmemsi): ... to this.
11460 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11461 Monk Chiang <sh.chiang04@gmail.com>
11462 Chung-Ju Wu <jasonwucj@gmail.com>
11464 * config/nds32/nds32-protos.h
11465 (nds32_expand_load_multiple): New arguments.
11466 (nds32_expand_store_multiple): Ditto.
11467 (nds32_valid_multiple_load_store): Rename ...
11468 (nds32_valid_multiple_load_store_p): ... to this.
11469 * config/nds32/nds32-memory-manipulation.c
11470 (nds32_expand_load_multiple): Refine implementation.
11471 (nds32_expand_store_multiple): Ditto.
11472 * config/nds32/nds32-multiple.md
11473 (load_multiple): Update nds32_expand_load_multiple interface.
11474 (store_multiple): Update nds32_expand_store_multiple interface.
11475 * config/nds32/nds32-predicates.c
11476 (nds32_valid_multiple_load_store): Rename ...
11477 (nds32_valid_multiple_load_store_p): ... to this and refine
11479 * config/nds32/predicates.md
11480 (nds32_load_multiple_and_update_address_operation): New predicate.
11481 (nds32_store_multiple_and_update_address_operation): New predicate.
11483 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11484 Chung-Ju Wu <jasonwucj@gmail.com>
11486 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11487 (combo): New attribute.
11488 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11490 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11492 * config/nds32/nds32.opt: Change -mcmodel= default value.
11494 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11495 Monk Chiang <sh.chiang04@gmail.com>
11496 Chung-Ju Wu <jasonwucj@gmail.com>
11498 * config/nds32/constants.md (unspec_element): New enum.
11499 * config/nds32/constraints.md (Umw): New constraint.
11500 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11501 * config/nds32/nds32-intrinsic.md: Likewise.
11502 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11503 (nds32_valid_smw_lwm_base_p): New.
11504 (nds32_output_smw_single_word): New.
11505 (nds32_output_lmw_single_word): New.
11506 (nds32_expand_unaligned_load): New.
11507 (nds32_expand_unaligned_store): New.
11508 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11509 (nds32_output_smw_single_word): Declare.
11510 (nds32_output_lmw_single_word): Declare.
11511 (nds32_expand_unaligned_load): Declare.
11512 (nds32_expand_unaligned_store): Declare.
11513 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11514 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11515 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11516 NDS32_BUILTIN_UASTORE_DW.
11517 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11520 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11521 Kito Cheng <kito.cheng@gmail.com>
11522 Chung-Ju Wu <jasonwucj@gmail.com>
11524 * config/nds32/nds32-intrinsic.c
11525 (nds32_expand_builtin_null_ftype_reg): Delete.
11526 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11527 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11528 (nds32_read_argument): New.
11529 (nds32_legitimize_target): Ditto.
11530 (nds32_legitimize_argument): Ditto.
11531 (nds32_check_constant_argument): Ditto.
11532 (nds32_expand_unop_builtin): Ditto.
11533 (nds32_expand_unopimm_builtin): Ditto.
11534 (nds32_expand_binop_builtin): Ditto.
11535 (nds32_builtin_decl_impl): Ditto.
11536 (builtin_description): Ditto.
11537 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11538 (nds32_init_builtins_impl): Ditto.
11539 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11540 (nds32_builtin_decl): New.
11541 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11542 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11544 2018-03-02 Jeff Law <law@redhat.com>
11546 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11547 (redundant_insn, fill_simple_delay_slots): Likewise.
11548 (fill_slots_from_thread): Likewise.
11549 * resource.c (mark_referenced_resources): Likewise.
11550 (mark_set_resources, find_dead_or_set_registers): Likewise.
11552 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11554 * substring-locations.h (format_warning_va): Formatting fix for
11555 ATTRIBUTE_GCC_DIAG.
11556 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11558 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11559 * substring-locations.c: Include intl.h.
11560 (format_warning_va): Turned into small wrapper around
11561 format_warning_n_va, renamed to ...
11562 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11563 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11565 (format_warning_at_substring_n): New function.
11566 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11567 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11568 format_warning_at_substring with just a shorter name instead of
11569 const function pointer.
11570 (fmtwarn_n): New function.
11571 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11572 appropriate, get rid of all the fmtstr temporaries, move conditionals
11573 with G_() wrapped string literals directly into fmtwarn arguments,
11574 cast dir.len to (int), formatting fixes.
11576 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11578 * doc/invoke.texi: Remove "Cilk Plus" references.
11580 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11581 Richard Biener <rguenther@suse.de>
11584 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11585 for error or warning attributes if CALL_FROM_THUNK_P is set.
11588 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11591 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11592 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11595 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11596 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11598 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11599 instead of -1U in last predictors element's probability member.
11601 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11604 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11605 arguments if they are comparable.
11607 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11609 PR tree-optimization/84634
11610 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11611 masks and masked_loop_p with a single loop_masks, making sure it's
11612 null for bb vectorization.
11614 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11616 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11617 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11618 loop->force_vectorize to check whether there is no alias.
11620 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11623 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11625 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11627 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11628 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11629 instead of a loop around prev_real_insn.
11630 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11633 PR inline-asm/84625
11634 * config/i386/i386.c (ix86_print_operand): Use conditional
11635 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11638 2018-03-02 Richard Biener <rguenther@suse.de>
11640 PR tree-optimization/84427
11641 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11642 (bitmap_set_subtract_values): Rewrite to handle multiple
11645 (prune_clobbered_mems): Likewise.
11646 (phi_translate): Take edge instead of pred/phiblock.
11647 (phi_translate_1): Likewise.
11648 (phi_translate_set): Likewise. Insert all translated
11649 exprs for a value into the set, keeping possibly multiple
11650 expressions per value.
11651 (compute_antic_aux): Adjust for phi_translate changes.
11652 When intersecting union the expressions and prune those
11653 not in the final value set, keeping possibly multiple
11654 expressions per value. Do not use value-insertion
11655 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11656 all expressions. Add verification that the value-sets
11657 only shrink during iteration.
11658 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11659 (do_pre_regular_insertion): Likewise.
11660 (do_pre_partial_partial_insertion): Likewise.
11662 2018-03-02 Richard Biener <rguenther@suse.de>
11665 * config/darwin.c (saved_debug_info_level): New static global.
11666 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11667 (darwin_asm_lto_end): Restore debug info generation settings.
11669 2018-03-01 Martin Liska <mliska@suse.cz>
11672 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11673 volatile arguments.
11675 2018-03-01 Richard Biener <rguenther@suse.de>
11678 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11679 type annotation with LTO when debug was disabled at compile-time.
11681 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11683 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11685 (mips_final_postscan_insn): Likewise.
11687 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11689 PR rtl-optimization/84528
11690 * alias.c (init_alias_target): Add commentary.
11691 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11692 a unique base value if the frame pointer is not eliminated
11693 to the stack pointer.
11695 2018-03-01 Tom de Vries <tom@codesourcery.com>
11697 PR rtl-optimization/83327
11698 * lra-int.h (hard_regs_spilled_into): Declare.
11699 * lra.c (hard_regs_spilled_into): Define.
11700 (init_reg_info): Init hard_regs_spilled_into.
11701 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11702 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11703 (process_bb_lives): Handle hard_regs_spilled_into.
11704 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11705 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11707 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11709 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11710 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11711 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11712 * config/rs6000/aix72.h: New file.
11714 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11716 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11717 instead of warning_at with conditional singular and plural messages
11721 * stor-layout.c (update_alignment_for_field): For
11722 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11723 && !DECL_PACKED (field), do the alignment update, just use
11724 only desired_align instead of MAX (type_align, desired_align)
11726 (place_field): Don't do known_align < desired_align handling
11727 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11728 is non-NULL, instead do it after rli->prev_field handling and
11729 only if not within a bitfield word. For DECL_PACKED (field)
11730 use type_align of BITS_PER_UNIT.
11732 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11734 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11735 superfluous parentheses and trailing spaces.
11737 2018-02-28 Richard Biener <rguenther@suse.de>
11739 PR tree-optimization/84584
11740 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11741 SCoPs with fake exit edge.
11743 2018-02-28 Martin Liska <mliska@suse.cz>
11746 * timevar.c (timer::print): Fix format to properly print 100%
11749 2018-02-28 Richard Biener <rguenther@suse.de>
11751 PR middle-end/84607
11752 * genmatch.c (capture_info::walk_match): Do not mark
11753 captured expressions without operands as expr_p given
11754 they act more like predicates and should be subject to
11755 "lost tail" side-effect preserving.
11757 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11759 PR rtl-optimization/81611
11760 * auto-inc-dec.c (attempt_change): Move dead note from
11761 mem_insn if it's the next use of regno
11762 (find_address): Take address use of reg holding
11763 non-incremented value. Add parm to limit search to the named
11765 (merge_in_block): Attempt to use a mem insn that is the next
11766 use of the original regno.
11768 2018-02-27 Martin Sebor <msebor@redhat.com>
11771 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11772 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11774 2018-02-27 Martin Sebor <msebor@redhat.com>
11776 PR translation/84207
11777 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11778 n argument to unsigned HOST_WIDE_INT.
11779 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11780 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11781 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11782 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11784 2018-02-27 Richard Biener <rguenther@suse.de>
11786 PR tree-optimization/84512
11787 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11788 Do not use the estimate returned from record_stmt_cost for
11789 the scalar iteration cost but sum properly using add_stmt_cost.
11791 2018-02-27 Richard Biener <rguenther@suse.de>
11793 PR tree-optimization/84466
11794 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11795 Adjust last change to less strictly validate use operands.
11797 2018-02-27 Martin Liska <mliska@suse.cz>
11799 PR gcov-profile/84548
11800 * gcov.c (process_file): Allow partial overlap and consider it
11801 also as group functions.
11802 (output_lines): Properly calculate range of lines for a group.
11804 2018-02-27 Martin Liska <mliska@suse.cz>
11806 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11807 'ggc' suffixes. Change first column width.
11808 (timer::print): Fix formatting of the column.
11810 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11812 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11813 preserve inline entry blocks for the sake of debug inline
11814 entry point markers alone.
11815 (remove_unused_locals): Suggest in comments a better place to
11816 force the preservation of inline entry blocks that are
11817 otherwise unused, but do not preserve them.
11819 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11821 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11823 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11826 * config/i386/constraints.md (Bs): Replace
11827 ix86_indirect_branch_register with
11828 TARGET_INDIRECT_BRANCH_REGISTER.
11830 * config/i386/i386.md (indirect_jump): Likewise.
11831 (tablejump): Likewise.
11832 (*sibcall_memory): Likewise.
11833 (*sibcall_value_memory): Likewise.
11834 Peepholes of indirect call and jump via memory: Likewise.
11835 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11836 (*sibcall_value_GOT_32): Likewise.
11837 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11838 (GOT_memory_operand): Likewise.
11839 (call_insn_operand): Likewise.
11840 (sibcall_insn_operand): Likewise.
11841 (GOT32_symbol_operand): Likewise.
11842 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11844 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11846 PR rtl-optimization/83496
11847 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11848 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11849 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11850 redundant insn, if any.
11851 (relax_delay_slots): Likewise.
11852 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11854 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11856 PR tree-optimization/83965
11857 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11858 that grouped statements are part of a reduction chain. Return
11859 true if the statement is not marked as a reduction itself but
11860 is part of a group.
11861 (vect_recog_dot_prod_pattern): Don't check whether the statement
11862 is part of a group here.
11863 (vect_recog_sad_pattern): Likewise.
11864 (vect_recog_widen_sum_pattern): Likewise.
11866 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11869 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11871 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11874 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11876 (ix86_output_indirect_function_return): New prototype.
11877 (ix86_split_simple_return_pop_internal): Likewise.
11878 * config/i386/i386.c (indirect_return_via_cx): New.
11879 (indirect_return_via_cx_bnd): Likewise.
11880 (indirect_thunk_name): Handle return va CX_REG.
11881 (output_indirect_thunk_function): Create alias for
11882 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11883 (ix86_output_indirect_jmp): Remove the bool argument.
11884 (ix86_output_indirect_function_return): New function.
11885 (ix86_split_simple_return_pop_internal): Likewise.
11886 * config/i386/i386.md (*indirect_jump): Don't pass false
11887 to ix86_output_indirect_jmp.
11888 (*tablejump_1): Likewise.
11889 (simple_return_pop_internal): Change it to define_insn_and_split.
11890 Call ix86_split_simple_return_pop_internal to split it for
11891 -mfunction-return=.
11892 (simple_return_indirect_internal): Call
11893 ix86_output_indirect_function_return instead of
11894 ix86_output_indirect_jmp.
11896 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11899 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11900 memset and value initialization afterwards.
11902 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11904 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11906 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11909 * common/config/aarch64/aarch64-common.c
11910 (aarch_option_optimization_table[]): Switch
11911 off fomit-frame-pointer
11913 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11914 Chung-Ju Wu <jasonwucj@gmail.com>
11916 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11918 (store_multiple): Ditto.
11920 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11922 * config.gcc: Add --with-cpu support for nds32 target.
11923 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11924 * config/nds32/nds32.opt: Add -mcpu= option.
11926 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11928 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11929 isel=yes): Warn for these deprecated options.
11931 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11933 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11934 ISA_2_5_MASKS_EMBEDDED.
11936 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11938 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11939 p->max as pointers rather than using iterative_hash_expr.
11941 2018-02-23 Carl Love <cel@us.ibm.com>
11943 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11944 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11946 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11947 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11948 P8V_BUILTIN_VEC_VUNSIGNED2.
11950 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11953 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11954 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11956 * lra-constraints.c (curr_insn_transform): Set up
11957 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11959 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11960 * lra-eliminations.c (spill_pseudos): Ditto.
11961 (process_insn_for_elimination): Ditto.
11962 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11963 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11964 LRA_NON_CLOBBERED_ALT.
11966 2018-02-22 Martin Sebor <msebor@redhat.com>
11968 PR tree-optimization/84480
11969 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11970 to maybe_diag_stxncpy_trunc. Call it.
11971 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11972 from gimple_fold_builtin_strcpy. Print inlining stack.
11973 (handle_builtin_stxncpy): Print inlining stack.
11974 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11976 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11979 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11980 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11981 and -fcheck-pointer-bounds are used together.
11982 (indirect_thunk_prefix): New enum.
11983 (indirect_thunk_need_prefix): New function.
11984 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11985 "_nt" instead of "_bnd" for NOTRACK prefix.
11986 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11987 (output_indirect_thunk_function): Likewise.
11989 (ix86_code_end): Update output_indirect_thunk_function calls.
11990 (ix86_output_indirect_branch_via_reg): Replace
11991 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11992 (ix86_output_indirect_branch_via_push): Likewise.
11993 (ix86_output_function_return): Likewise.
11994 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11995 incompatible with -fcf-protection=branch and
11996 -fcheck-pointer-bounds.
11998 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12001 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12002 Change gcc_assert call to output_operand_lossage.
12004 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12006 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12008 2018-02-22 DJ Delorie <dj@redhat.com>
12009 Sebastian Perta <sebastian.perta@renesas.com>
12010 Oleg Endo <olegendo@gcc.gnu.org>
12012 * config/rx/rx.c (rx_rtx_costs): New function.
12013 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12015 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12017 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12019 2018-02-22 Martin Liska <mliska@suse.cz>
12022 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12023 Add "native" as a possible value.
12025 2018-02-22 Martin Liska <mliska@suse.cz>
12028 * config/i386/i386.c (ix86_option_override_internal):
12029 Add "native" as a possible value for -march and -mtune.
12031 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12034 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12035 to all type variants.
12037 PR tree-optimization/84503
12038 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12039 width as info->bitpos + info->bitsize - start.
12040 (merged_store_group::merge_overlapping): Simplify width computation.
12041 (check_no_overlap): New function.
12042 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12043 start + width and last_order of the group, fail if check_no_overlap
12045 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12046 to group if check_no_overlap fails.
12048 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12050 * config/rs6000/altivec.md: Delete contraint arguments to
12051 define_expand, define_split, and define_peephole2, and in
12052 define_insn_and_split if always unused.
12053 * config/rs6000/darwin.md: Ditto.
12054 * config/rs6000/dfp.md: Ditto.
12055 * config/rs6000/rs6000.md: Ditto.
12056 * config/rs6000/sync.md: Ditto.
12057 * config/rs6000/vector.md: Ditto.
12058 * config/rs6000/vsx.md: Ditto.
12060 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12062 * config/rs6000/altivec.md: Write output control strings as braced
12063 blocks instead of double-quoted strings.
12064 * config/rs6000/darwin.md: Ditto.
12065 * config/rs6000/rs6000.md: Ditto.
12066 * config/rs6000/vector.md: Ditto.
12067 * config/rs6000/vsx.md: Ditto.
12069 2018-02-21 Jason Merrill <jason@redhat.com>
12071 PR c++/84314 - ICE with templates and fastcall attribute.
12072 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12074 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12076 * ipa-cp.c (determine_versionability): Fix comment typos.
12078 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12081 * ipa-cp.c (determine_versionability): Do not version functions caling
12084 2018-02-21 Martin Liska <mliska@suse.cz>
12087 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12088 Add "native" as a possible value.
12089 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12090 the macro when native cpu detection is available.
12092 2018-02-21 Martin Liska <mliska@suse.cz>
12095 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12096 Add "native" as a possible value.
12097 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12098 when native cpu detection is available.
12100 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12101 Martin Sebor <msebor@redhat.com>
12103 PR tree-optimization/84478
12104 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12106 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12107 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12108 support which is conservatively correct, for 2 only stay conservative
12109 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12110 argument to the 2 argument get_range_strlen, adjust 6 arg
12111 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12113 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12114 (gimple_fold_builtin_strlen): Pass true as last argument to
12117 2018-02-20 Martin Sebor <msebor@redhat.com>
12119 PR middle-end/84095
12120 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12121 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12122 (builtin_memref::builtin_memref): Factor out parts into
12123 set_base_and_offset and call it.
12125 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12127 PR middle-end/84406
12128 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12129 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12130 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12131 search at the associated MODE_INT.
12133 2018-02-20 Jeff Law <law@redhat.com>
12135 PR middle-end/82123
12136 PR tree-optimization/81592
12137 PR middle-end/79257
12138 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12139 for range data rather than using global data.
12140 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12141 range data rather than using global data.
12142 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12143 pass it to children as needed.
12144 (struct directive::fmtresult): Similarly.
12145 (struct directive::set_width): Similarly.
12146 (struct directive::set_precision): Similarly.
12147 (format_integer, format_directive, parse_directive): Similarly.
12148 (format_none): Accept unnamed vr_values parameter.
12149 (format_percent, format_floating, format_character): Similarly.
12150 (format_string, format_plain): Similarly.
12151 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12152 the EVRP range analyzer for range data rather than using global data.
12153 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12154 gimple-ssa-evrp-analyze.h
12155 (class sprintf_dom_walker): Add after_dom_children member function.
12156 Add evrp_range_analyzer member.
12157 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12158 range analyzer as needed.
12159 (sprintf_dom_walker::after_dom_children): New member function.
12160 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12162 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12163 (evrp_range_analyzer::pop_to_marker): Likewise.
12165 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12167 PR tree-optimization/84419
12168 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12169 with the required type if its current type is compatible but
12172 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12174 PR middle-end/82004
12175 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12176 after vectorization.
12178 2018-02-20 Martin Liska <mliska@suse.cz>
12181 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12182 possible values if we don't have a hint.
12184 2018-02-20 Martin Liska <mliska@suse.cz>
12188 * final.c (shorten_branches): Build align_tab array with one
12190 * opts.c (finish_options): Add alignment option limit check.
12191 (MAX_CODE_ALIGN): Likewise.
12192 (MAX_CODE_ALIGN_VALUE): Likewise.
12193 * doc/invoke.texi: Document maximum allowed option value for
12194 all -falign-* options.
12196 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12199 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12200 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12201 * var-tracking.c (emit_note_insn_var_location): Remove all references
12202 to NOTE_INSN_CALL_ARG_LOCATION.
12203 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12204 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12205 Use copy_rtx_if_shared.
12206 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12207 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12208 (dwarf2out_var_location): Remove handling of
12209 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12211 * final.c (final_scan_insn): Remove all references to
12212 NOTE_INSN_CALL_ARG_LOCATION.
12213 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12214 before dumping final insns.
12215 * except.c (emit_note_eh_region_end): Remove all references to
12216 NOTE_INSN_CALL_ARG_LOCATION.
12217 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12218 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12219 * config/arc/arc.c (hwloop_optimize): Likewise.
12220 * config/arm/arm.c (create_fix_barrier): Likewise.
12221 * config/s390/s390.c (s390_chunkify_start): Likewise.
12222 * config/sh/sh.c (find_barrier): Likewise.
12223 * config/i386/i386.c (rest_of_insert_endbranch,
12224 ix86_seh_fixup_eh_fallthru): Likewise.
12225 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12226 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12227 * config/frv/frv.c (frv_function_prologue): Likewise.
12228 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12230 (note_outside_basic_block_p): Remove all references to
12231 NOTE_INSN_CALL_ARG_LOCATION.
12232 * gengtype.c (adjust_field_rtx_def): Likewise.
12233 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12235 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12236 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12239 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12242 PR tree-optimization/84452
12243 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12244 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12247 2018-02-19 Martin Liska <mliska@suse.cz>
12250 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12252 2018-02-19 Martin Liska <mliska@suse.cz>
12253 Richard Sandiford <richard.sandiford@linaro.org>
12255 PR tree-optimization/82491
12256 * gimple-fold.c (get_base_constructor): Make earlier bail out
12259 2018-02-19 Carl Love <cel@us.ibm.com>
12261 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12262 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12264 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12265 P8V_BUILTIN_VEC_NEG.
12267 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12269 * config/rl78/rl78.md (movdf): New define expand.
12271 2018-02-19 Martin Liska <mliska@suse.cz>
12274 * doc/invoke.texi: Fix typo.
12275 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12277 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12279 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12280 handle rs6000_single_float and rs6000_double_float specially for
12283 2018-02-16 Jeff Law <law@redhat.com>
12285 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12286 (pop_regs): Use it.
12288 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12291 * ipa-inline.c (inline_small_functions): Fix a typo.
12293 2018-02-16 Nathan Sidwell <nathan@acm.org>
12295 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12297 2018-02-16 Carl Love <cel@us.ibm.com>
12299 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12300 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12301 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12302 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12303 expansion to P8V_BUILTIN_VEC_FLOAT2.
12305 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12307 PR rtl-optimization/70023
12308 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12309 src_regno into account.
12311 2018-02-16 Carl Love <cel@us.ibm.com>
12313 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12314 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12315 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12316 * config/rs6000/rs6000.c: Remove case statements for
12317 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12318 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12319 and P9V_BUILTIN_VEC_VINSERT4B.
12320 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12321 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12322 * config/rs6000/vsx.md:
12323 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12326 2018-02-16 Carl Love <cel@us.ibm.com>
12328 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12330 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12332 * config/rs6000/rs6000-c.c: Add the definitions for
12333 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12334 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12335 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12336 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12337 definition for insert4b and define insn *insert3b_internal.
12338 * doc/extend.texi: Add documentation for vec_extract4b.
12340 2018-02-16 Nathan Sidwell <nathan@acm.org>
12342 * doc/extend.texi (Backwards Compatibility): Mention friend
12343 injection. Note for-scope is deprecated.
12344 * doc/invoke.texi (-ffriend-injection): Deprecate.
12346 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12348 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12349 that moved to I2, also allow destinations that are a paradoxical
12350 subreg (instead of a normal reg).
12352 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12355 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12358 2018-02-16 Richard Biener <rguenther@suse.de>
12360 PR tree-optimization/84037
12361 PR tree-optimization/84016
12363 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12364 Adjust vec_construct for the fact we need additional higher latency
12365 128bit inserts for AVX256 and AVX512 vector builds.
12366 (ix86_add_stmt_cost): Scale vector construction cost for
12369 2018-02-16 Richard Biener <rguenther@suse.de>
12371 PR tree-optimization/84417
12372 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12373 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12374 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12376 2018-02-16 Martin Liska <mliska@suse.cz>
12379 * internal-fn.def (ASAN_CHECK): Set proper flags.
12380 (ASAN_MARK): Likewise.
12382 2018-02-16 Julia Koval <julia.koval@intel.com>
12384 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12385 from PTA_CANNONLAKE.
12387 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12390 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12391 Use ++iter rather than iter++ for std::list iterators.
12392 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12393 defer deleting them until all nodes in the forest are processed. Do
12394 free even leaf nodes. Change to_process into auto_vec.
12397 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12398 * vec.h (vec_default_construct): Use memset instead of placement new
12399 if BROKEN_VALUE_INITIALIZATION is defined.
12400 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12401 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12404 PR rtl-optimization/83723
12405 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12406 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12407 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12409 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12411 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12413 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12415 PR rtl-optimization/81443
12416 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12417 from inner REGs to paradoxical SUBREGs.
12419 2018-02-16 Richard Biener <rguenther@suse.de>
12421 PR tree-optimization/84399
12422 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12423 For operands we can analyze at their definition make sure we can
12424 analyze them at each use as well.
12426 2018-02-16 Richard Biener <rguenther@suse.de>
12428 PR tree-optimization/84190
12429 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12430 volatile accesses if the decl isn't volatile.
12432 2018-02-15 Jason Merrill <jason@redhat.com>
12434 PR c++/84314 - ICE with templates and fastcall attribute.
12435 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12436 TYPE_CANONICAL on an existing type.
12438 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12440 PR tree-optimization/84383
12441 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12442 dstoff nor call operand_equal_p if dstbase is NULL.
12444 PR tree-optimization/84334
12445 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12446 also a CONSTANT_CLASS_P, punt.
12448 2018-02-14 Jim Wilson <jimw@sifive.com>
12450 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12451 first SMALL_OPERAND check. New local min_second_step. Move assert
12452 to where locals are set. Add TARGET_RVC support.
12453 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12455 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12457 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12459 2018-02-14 Martin Sebor <msebor@redhat.com>
12461 PR tree-optimization/83698
12462 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12463 arrays constrain the offset range to their bounds.
12464 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12465 (builtin_access::overlap): Avoid setting the size of overlap if it's
12467 (maybe_diag_overlap): Also consider arrays when deciding what values
12468 of offsets to include in diagnostics.
12470 2018-02-14 Martin Sebor <msebor@redhat.com>
12473 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12474 that correspond to the kind of a declaration.
12476 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12479 * config/pa/pa.md: Load address of PIC label using the linkage table
12480 if the label is nonlocal.
12482 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12484 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12485 warning message if user requests -maltivec=be.
12486 * doc/invoke.texi: Document deprecation of -maltivec=be.
12488 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12491 * config/rs6000/rs6000-c.c: Update definitions for
12492 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12493 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12495 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12498 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12499 add _get_ssp intrinsics. Remove argument from
12500 __builtin_ia32_rdssp[d|q].
12501 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12502 * config/i386/i386-builtin.def: Remove argument from
12503 __builtin_ia32_rdssp[d|q].
12504 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12505 ix86_expand_special_args_builtin for _rdssp[d|q].
12506 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12507 Clear register before usage.
12508 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12509 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12511 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12513 PR tree-optimization/84357
12514 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12515 operand 1 of an ARRAY_REF too.
12517 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12520 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12521 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12523 (set_of_reg): New struct.
12524 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12525 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12526 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12528 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12529 Split into bitclr, bitset, bitinvert patterns if appropriate.
12530 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12531 use rx_fuse_in_memory_bitop.
12532 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12533 to named insn, correct maximum insn length.
12535 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12538 * machmode.def: Define a complex mode for PARTIAL_INT.
12539 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12541 * doc/rtl.texi: Document CSPImode.
12542 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12544 (msp430_hard_regno_nregs_with_padding): Likewise.
12546 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12549 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12551 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12553 PR rtl-optimization/84169
12554 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12555 we generated a parallel as new i3 and we split that to new i2 and i3
12556 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12557 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12558 those to i2, not i1. Partially rewrite this scan code.
12560 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12563 * stor-layout.c (place_field): For variable length fields, adjust
12564 offset_align afterwards not just based on the field's alignment,
12565 but also on the size.
12567 PR middle-end/84309
12568 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12569 of exps and logs in the use_exp2 case.
12571 2018-02-13 Jeff Law <law@redhat.com>
12573 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12574 entry for "vector".
12576 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12579 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12583 * common.opt (gas-loc-support, gas-locview-support): New.
12584 (ginline-points, ginternal-reset-location-views): New.
12585 * doc/invoke.texi: Document them. Use @itemx where intended.
12586 (gvariable-location-views): Adjust.
12587 * target.def (reset_location_view): New.
12588 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12589 (TARGET_RESET_LOCATION_VIEW): New.
12590 * doc/tm.texi: Rebuilt.
12591 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12592 (dwarf2out_default_as_locview_support): New.
12593 (output_asm_line_debug_info): Use option variables.
12594 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12595 (output_loc_list): Likewise.
12596 (add_high_low_attributes): Check option variables.
12597 Don't output entry view attribute in strict mode.
12598 (gen_inlined_subroutine_die): Check option variables.
12599 (dwarf2out_inline_entry): Likewise.
12600 (init_sections_and_labels): Likewise.
12601 (dwarf2out_early_finish): Likewise.
12602 (maybe_reset_location_view): New, from...
12603 (dwarf2out_var_location): ... here. Call it.
12604 * debug.h (dwarf2out_default_as_loc_support): Declare.
12605 (dwarf2out_default_as_locview_support): Declare.
12606 * hooks.c (hook_int_rtx_insn_0): New.
12607 * hooks.h (hook_int_rtx_insn_0): Declare.
12608 * toplev.c (process_options): Take -gas-loc-support and
12609 -gas-locview-support from dwarf2out. Enable
12610 -gvariable-location-views by default only with locview
12611 assembler support. Enable -ginternal-reset-location-views by
12612 default only if the target defines the corresponding hook.
12613 Enable -ginline-points by default if location views are
12614 enabled; force it disabled if statement frontiers are
12616 * tree-inline.c (expand_call_inline): Check option variables.
12617 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12619 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12621 PR tree-optimization/84321
12622 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12623 handling. Also check whether the anti-range contains any values
12624 that satisfy the mask; switch to a VR_RANGE if not.
12626 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12629 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12631 2018-02-13 Martin Jambor <mjambor@suse.cz>
12634 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12635 of call statements, also set location of a load to a temporary.
12637 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12639 * config/rl78/rl78.c (add_vector_labels): New function.
12640 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12641 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12642 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12643 which checks that no arguments are passed.
12644 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12645 * doc/extend.texi: Documentation for the new attribute.
12647 2018-02-13 Andreas Schwab <schwab@suse.de>
12649 * config/riscv/linux.h (CPP_SPEC): Define.
12651 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12654 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12655 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12656 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12657 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12658 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12659 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12660 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12661 temporarily for AES and PCLMUL builtins.
12663 PR tree-optimization/84339
12664 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12665 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12668 PR middle-end/84309
12669 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12670 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12671 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12673 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12675 * omp-simd-clone.h: New file.
12676 * omp-simd-clone.c: Include omp-simd-clone.h.
12677 (expand_simd_clones): No longer static.
12678 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12679 cgraph.h and omp-simd-clone.h.
12680 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12681 (vect_recog_widen_shift_pattern): Formatting fix.
12682 (vect_pattern_recog_1): Don't check optab for calls.
12685 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12686 operands[2] into a REG before using gen_lowpart on it.
12688 2018-02-12 Jeff Law <law@redhat.com>
12691 * config/sh/sh.c (find_barrier): Consider a sibling call
12694 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12695 successfully back substituting a reg.
12697 2018-02-12 Richard Biener <rguenther@suse.de>
12699 PR tree-optimization/84037
12700 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12701 parameter, move visited init to caller.
12702 (vect_slp_analyze_operations): Separate cost from validity
12703 check, initialize visited once for all instances.
12704 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12706 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12708 (vect_model_promotion_demotion_cost): Likewise.
12709 (vectorizable_bswap): Guard cost modeling with !slp_node
12710 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12712 (vectorizable_call): Likewise.
12713 (vectorizable_conversion): Likewise.
12714 (vectorizable_assignment): Likewise.
12715 (vectorizable_shift): Likewise.
12716 (vectorizable_operation): Likewise.
12717 (vectorizable_store): Likewise.
12718 (vectorizable_load): Likewise.
12719 (vectorizable_condition): Likewise.
12720 (vectorizable_comparison): Likewise.
12722 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12725 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12726 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12727 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12729 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12731 PR middle-end/83665
12732 * params.def (inline-min-speedup): Increase from 8 to 15.
12733 (max-inline-insns-auto): Decrease from 40 to 30.
12734 * ipa-split.c (consider_split): Add some buffer for function to
12735 be considered inlining candidate.
12736 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12739 2018-02-12 Richard Biener <rguenther@suse.de>
12741 PR tree-optimization/84037
12742 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12743 matched stmts if we cannot swap the non-matched ones.
12745 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12747 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12748 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12749 _mm_maskz_scalef_round_ss): New intrinsics.
12750 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12751 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12752 __builtin_ia32_scalefss_round): Remove.
12753 (__builtin_ia32_scalefsd_mask_round,
12754 __builtin_ia32_scalefss_mask_round): New intrinsics.
12755 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12756 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12757 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12758 "<round_constraint>")): Changed to ...
12759 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12760 "<round_scalar_constraint>")): ... this.
12761 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12762 %0, %1, %2<round_op3>}"): Changed to ...
12763 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12764 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12765 %2<round_scalar_mask_op3>}"): ... this.
12766 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12768 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12770 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12771 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12772 (_mm_maskz_sqrt_round_ss): New intrinsics.
12773 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12774 (__builtin_ia32_sqrtsd_mask_round)
12775 (__builtin_ia32_sqrtss_mask_round): New builtins.
12776 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12777 (__builtin_ia32_sqrtss_round): Remove.
12778 (__builtin_ia32_sqrtsd_mask_round)
12779 (__builtin_ia32_sqrtss_mask_round): New builtins.
12780 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12781 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12782 ((match_operand:VF_128 1 "vector_operand"
12783 "xBm,<round_constraint>")): Changed to ...
12784 ((match_operand:VF_128 1 "vector_operand"
12785 "xBm,<round_scalar_constraint>")): ... this.
12786 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12787 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12788 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12789 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12790 %<iptr>1<round_scalar_mask_op3>}): ... this.
12791 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12792 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12794 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12797 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12798 Cast vec_cmpeq result to correct type.
12799 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12800 Cast vec_cmpgt result to correct type.
12802 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12804 * final.c (final_scan_insn_1): Renamed from...
12805 (final_scan_insn): ... this. New wrapper, to recover
12806 seen from the outermost call in recursive ones.
12807 * config/sparc/sparc.c (output_return): Drop seen from call.
12808 (output_sibcall): Likewise.
12809 * config/visium/visium.c (output_branch): Likewise.
12811 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12813 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12816 2018-02-10 Alan Modra <amodra@gmail.com>
12819 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12820 Specify LR as an input.
12822 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12825 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12826 remove_member_access_dummy_vars): New functions.
12827 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12828 lower_omp_1, execute_lower_omp): Use them.
12830 PR rtl-optimization/84308
12831 * shrink-wrap.c (spread_components): Release todo vector.
12833 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12835 PR rtl-optimization/57193
12836 * ira-color.c (struct allocno_color_data): Add member
12837 conflict_allocno_hard_prefs.
12838 (update_conflict_allocno_hard_prefs): New.
12839 (bucket_allocno_compare_func): Add a preference based on
12840 conflict_allocno_hard_prefs.
12841 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12842 (color_allocnos): Remove a dead code. Initiate
12843 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12845 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12848 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12849 constraint from =wa to wa. Avoid a subreg on the output operand,
12850 instead use a pseudo and subreg it in a move.
12851 (p9_xxbrd_<mode>): Changed to ...
12852 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12853 (p9_xxbrd_v2df): New expander.
12854 (p9_xxbrw_<mode>): Changed to ...
12855 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12856 (p9_xxbrw_v4sf): New expander.
12858 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12860 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12862 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12865 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12866 multiply in 32-bit mode.
12867 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12868 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12871 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12873 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12874 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12875 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12876 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12878 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12881 * dwarf2out.c (is_trivial_indirect_ref): New function.
12882 (dwarf2out_late_global_decl): Do not generate a location
12883 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12884 and that are not defined in the current unit.
12886 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12888 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12889 instead of a libcall for UNORDERED.
12891 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12894 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12895 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12897 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12900 * config/s390/s390.c (s390_set_current_function): Invoke
12901 s390_indirect_branch_settings also if fndecl didn't change.
12903 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12905 * config/rs6000/rs6000.md (blockage): Set length to zero.
12907 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12909 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12911 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12914 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12915 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12919 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12920 PARALLEL incoming that failed vt_get_decl_and_offset check.
12922 PR middle-end/84237
12923 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12924 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12926 (get_variable_section): For decls in named .bss* sections pass true as
12927 second argument to bss_initializer_p.
12929 2018-02-09 Marek Polacek <polacek@redhat.com>
12930 Jakub Jelinek <jakub@redhat.com>
12933 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12934 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12935 Sync some changes from cxx_fold_indirect_ref.
12937 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12939 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12941 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12942 (BLOCK_INLINE_ENTRY_LABEL): New.
12943 (dwarf2out_var_location): Disregard inline entry markers.
12944 (inline_entry_data): New struct.
12945 (inline_entry_data_hasher): New hashtable type.
12946 (inline_entry_data_hasher::hash): New.
12947 (inline_entry_data_hasher::equal): New.
12948 (inline_entry_data_table): New variable.
12949 (add_high_low_attributes): Add DW_AT_entry_pc and
12950 DW_AT_GNU_entry_view attributes if a pending entry is found
12951 in inline_entry_data_table. Add old entry_pc attribute only
12952 if debug nonbinding markers are disabled.
12953 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12954 markers are enabled.
12955 (block_within_block_p, dwarf2out_inline_entry): New.
12956 (dwarf2out_finish): Check that no entries remained in
12957 inline_entry_data_table.
12958 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12959 (final_scan_insn, notice_source_line): Likewise.
12960 (rest_of_clean_state): Skip inline entry markers.
12961 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12963 * gimple.c (gimple_build_debug_inline_entry): New.
12964 * gimple.h (enum gimple_debug_subcode): Add
12965 GIMPLE_DEBUG_INLINE_ENTRY.
12966 (gimple_build_debug_inline_entry): Declare.
12967 (gimple_debug_inline_entry_p): New.
12968 (gimple_debug_nonbind_marker_p): Adjust.
12969 * insn-notes.def (INLINE_ENTRY): New.
12970 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12971 inline entry marker notes.
12972 (print_insn): Likewise.
12973 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12974 (INSN_DEBUG_MARKER_KIND): Likewise.
12975 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12976 * tree-inline.c (expand_call_inline): Build and insert
12977 debug_inline_entry stmt.
12978 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12979 inline entry blocks early, if nonbind markers are enabled.
12980 (dump_scope_block): Dump fragment info.
12981 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12982 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12983 (gimple_build_debug_inline_entry): New.
12984 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12985 Enable/disable inline entry points too.
12986 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12987 (DEBUG_INSN): Describe inline entry markers.
12989 * common.opt (gvariable-location-views): New.
12990 (gvariable-location-views=incompat5): New.
12991 * config.in: Rebuilt.
12992 * configure: Rebuilt.
12993 * configure.ac: Test assembler for view support.
12994 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12995 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12996 * dwarf2out.c (var_loc_view): New typedef.
12997 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12998 (dwarf2out_locviews_in_attribute): New.
12999 (dwarf2out_locviews_in_loclist): New.
13000 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13001 (enum dw_line_info_opcode): Add LI_adv_address.
13002 (struct dw_line_info_table): Add view.
13003 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13004 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13005 (zero_view_p): New variable.
13006 (ZERO_VIEW_P): New macro.
13007 (output_asm_line_debug_info): New.
13008 (struct var_loc_node): Add view.
13009 (add_AT_view_list, AT_loc_list): New.
13010 (add_var_loc_to_decl): Add view param. Test it against last.
13011 (new_loc_list): Add view params. Record them.
13012 (AT_loc_list_ptr): Handle loc and view lists.
13013 (view_list_to_loc_list_val_node): New.
13014 (print_dw_val): Handle dw_val_class_view_list.
13015 (size_of_die): Likewise.
13016 (value_format): Likewise.
13017 (loc_list_has_views): New.
13018 (gen_llsym): Set vl_symbol too.
13019 (maybe_gen_llsym, skip_loc_list_entry): New.
13020 (dwarf2out_maybe_output_loclist_view_pair): New.
13021 (output_loc_list): Output view list or entries too.
13022 (output_view_list_offset): New.
13023 (output_die): Handle dw_val_class_view_list.
13024 (output_dwarf_version): New.
13025 (output_compilation_unit_header): Use it.
13026 (output_skeleton_debug_sections): Likewise.
13027 (output_rnglists, output_line_info): Likewise.
13028 (output_pubnames, output_aranges): Update version comments.
13029 (output_one_line_info_table): Output view numbers in asm comments.
13030 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13031 Call maybe_gen_llsym.
13032 (loc_list_from_tree_1): Adjust.
13033 (add_AT_location_description): Create view list attribute if
13034 needed, check it's absent otherwise.
13035 (convert_cfa_to_fb_loc_list): Adjust.
13036 (maybe_emit_file): Call output_asm_line_debug_info for test.
13037 (dwarf2out_var_location): Reset views as needed. Precompute
13038 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13039 attribute. Set view.
13040 (new_line_info_table): Reset next view.
13041 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13042 (dwarf2out_source_line): Likewise. Output view resets and labels to
13043 the assembler, or select appropriate line info opcodes.
13044 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13045 (optimize_string_length): Catch it. Adjust.
13046 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13047 dw_val_class_view_list, and remove it if no longer needed.
13048 (hash_loc_list): Hash view numbers.
13049 (loc_list_hasher::equal): Compare them.
13050 (optimize_location_lists): Check whether a view list symbol is
13051 needed, and whether the locview attribute is present, and
13052 whether they match. Remove the locview attribute if no longer
13054 (index_location_lists): Call skip_loc_list_entry for test.
13055 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13056 Use output_dwarf_version.
13057 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13058 (struct dw_val_node): Add val_view_list.
13059 * final.c (SEEN_NEXT_VIEW): New.
13060 (set_next_view_needed): New.
13061 (clear_next_view_needed): New.
13062 (maybe_output_next_view): New.
13063 (final_start_function): Rename to...
13064 (final_start_function_1): ... this. Take pointer to FIRST,
13065 add SEEN parameter. Emit param bindings in the initial view.
13066 (final_start_function): Reintroduce SEEN-less interface.
13067 (final): Rename to...
13068 (final_1): ... this. Take SEEN parameter. Output final pending
13069 next view at the end.
13070 (final): Reintroduce seen-less interface.
13071 (final_scan_insn): Output pending next view before switching
13072 sections or ending a block. Mark the next view as needed when
13073 outputting variable locations. Notify debug backend of section
13074 changes, and of location view changes.
13075 (rest_of_handle_final): Adjust.
13076 * toplev.c (process_options): Autodetect value for debug variable
13077 location views option. Warn on incompat5 without -gdwarf-5.
13078 * doc/invoke.texi (gvariable-location-views): New.
13079 (gvariable-location-views=incompat5): New.
13080 (gno-variable-location-views): New.
13082 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13084 PR tree-optimization/84136
13085 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13086 that the result of find_edge is non-NULL.
13088 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13091 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13092 storing integer register in SImode. Fix cost of 256 and 512
13093 byte aligned SSE register store.
13095 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13097 * config/i386/i386.c (ix86_multiplication_cost): Fix
13098 multiplication cost for TARGET_AVX512DQ.
13100 2018-02-08 Marek Polacek <polacek@redhat.com>
13102 PR tree-optimization/84238
13103 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13106 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13108 PR tree-optimization/84265
13109 * tree-vect-stmts.c (vectorizable_store): Don't treat
13110 VMAT_CONTIGUOUS accesses as grouped.
13111 (vectorizable_load): Likewise.
13113 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13115 PR tree-optimization/81635
13116 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13117 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13118 (test_round_for_mask): New functions.
13119 (wide_int_cc_tests): Call test_round_for_mask.
13120 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13121 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13122 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13123 range returned by get_range_info.
13125 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13128 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13129 * symtab.c: Include builtins.h
13130 (symtab_node::output_to_lto_symbol_table_p): Move here
13131 from lto-streamer-out.c:output_symbol_p.
13132 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13133 (output_symbol_p): Move all logic to symtab.c
13134 (produce_symtab): Update.
13136 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13138 * config/s390/s390-opts.h (enum indirect_branch): Define.
13139 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13140 (s390_indirect_branch_via_thunk)
13141 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13142 (enum s390_indirect_branch_type): Define.
13143 * config/s390/s390.c (struct s390_frame_layout, struct
13144 machine_function): Remove.
13145 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13146 (indirect_branch_table_label_no, indirect_branch_table_name):
13148 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13149 (enum s390_indirect_branch_option): Define.
13150 (s390_return_addr_from_memory): New function.
13151 (s390_handle_string_attribute): New function.
13152 (s390_attribute_table): Add new attribute handler.
13153 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13154 (s390_indirect_branch_via_thunk): New function.
13155 (s390_indirect_branch_via_inline_thunk): New function.
13156 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13157 sibling call optimization for non z10 compiles.
13158 (s390_emit_call): Force indirect branch target to be a single
13159 register. Add r1 clobber for non-z10 compiles.
13160 (s390_emit_epilogue): Emit return jump via return_use expander.
13161 (s390_reorg): Handle JUMP_INSNs as execute targets.
13162 (s390_option_override_internal): Perform validity checks for the
13163 new command line options.
13164 (s390_indirect_branch_attrvalue): New function.
13165 (s390_indirect_branch_settings): New function.
13166 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13167 (s390_output_indirect_thunk_function): New function.
13168 (s390_code_end): Implement target hook.
13169 (s390_case_values_threshold): Implement target hook.
13170 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13172 * config/s390/s390.h (struct s390_frame_layout)
13173 (struct machine_function): Move here from s390.c.
13174 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13175 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13176 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13177 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13178 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13179 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13180 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13181 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13182 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13183 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13184 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13185 (mnemonic attribute): Add values which aren't recognized
13187 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13188 pattern for branch conversion. Fix mnemonic attribute.
13189 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13190 indirect branch via thunk if requested.
13191 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13192 ("*indirect_jump"): Disable for branch conversion using out of
13194 ("indirect_jump_via_thunk<mode>_z10")
13195 ("indirect_jump_via_thunk<mode>")
13196 ("indirect_jump_via_inlinethunk<mode>_z10")
13197 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13198 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13199 ("casesi_jump_via_inlinethunk<mode>_z10")
13200 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13201 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13202 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13203 ("*indirect2_jump"): Disable for branch conversion.
13204 ("casesi_jump"): Turn into expander and expand patterns for branch
13206 ("return_use"): New expander.
13207 ("*return"): Emit return via thunk and rename it to ...
13208 ("*return<mode>"): ... this one.
13209 * config/s390/s390.opt: Add new options and and enum for the
13212 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13214 * lra-constraints.c (match_reload): Unconditionally use
13215 gen_lowpart_SUBREG, rather than selecting between that
13216 and equivalent gen_rtx_SUBREG code.
13218 2018-02-08 Richard Biener <rguenther@suse.de>
13220 PR tree-optimization/84233
13221 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13222 changed flag instead of boguously re-using phi_inserted.
13224 2018-02-08 Martin Jambor <mjambor@suse.cz>
13226 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13227 static local variables.
13229 2018-02-08 Richard Biener <rguenther@suse.de>
13231 PR tree-optimization/84278
13232 * tree-vect-stmts.c (vectorizable_store): When looking for
13233 smaller vector types to perform grouped strided loads/stores
13234 make sure the mode is supported by the target.
13235 (vectorizable_load): Likewise.
13237 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13239 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13240 Increase LDP/STP opportunities by adding adjacent callee-saves.
13242 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13244 PR rtl-optimization/84068
13245 PR rtl-optimization/83459
13246 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13248 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13250 PR tree-optimization/84224
13251 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13252 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13253 non-zero arguments.
13255 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13258 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13259 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13260 count, remove one USE.
13262 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13264 * doc/install.texi (Configuration): Document the
13265 --with-long-double-format={ibm,ieee} PowerPC configuration
13269 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13270 Convert from define_expand to be define_insn_and_split. Rework
13271 float/double/_Float128 conversions to QI/HI/SImode to work with
13272 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13273 conversions to QI/HImode types did a store and then a load to
13274 truncate the value. For conversions to VSX registers, don't split
13275 the insn, instead emit the code directly. Use the code iterator
13276 any_fix to combine signed and unsigned conversions.
13277 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13278 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13279 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13280 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13281 (fix_<mode>di2_hw): Likewise.
13282 (fixuns_<mode>di2_hw): Likewise.
13283 (fix_<mode>si2_hw): Likewise.
13284 (fixuns_<mode>si2_hw): Likewise.
13285 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13286 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13287 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13288 fix<uns>_trunc<SFDF:mode>si2_p8.
13289 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13290 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13291 (fix<uns>_<mode>_mem): Likewise.
13292 (fctiw<u>z_<mode>_mem): Likewise.
13293 (fix<uns>_<mode>_mem): Likewise.
13294 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13295 the register allocator from doing a direct move to the GPRs to do
13296 a store, and instead use the ISA 3.0 store byte/half-word from
13297 vector register instruction. For IEEE 128-bit floating point,
13298 also optimize stores of 32-bit ints.
13299 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13301 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13303 * genextract.c (push_pathstr_operand): New function to support
13305 (walk_rtx): Call push_pathstr_operand.
13306 (print_path): Support [a-zA-Z].
13308 2018-02-07 Richard Biener <rguenther@suse.de>
13310 PR tree-optimization/84037
13311 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13312 (cse_and_gimplify_to_preheader): Declare.
13313 (vect_get_place_in_interleaving_chain): Likewise.
13314 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13316 (_loop_vec_info::~_loop_vec_info): Delete it.
13317 (cse_and_gimplify_to_preheader): New function.
13318 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13319 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13320 (vectorizable_load): Likewise. For grouped stores always base
13321 the IV on the first element.
13322 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13323 condition before gimplifying.
13325 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13327 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13328 *DIV_EXPR and *MOD_EXPR.
13330 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13333 * config/i386/i386.c (ix86_option_override_internal): Mask out
13334 the CF_SET bit when checking -fcf-protection.
13336 2018-02-07 Tom de Vries <tom@codesourcery.com>
13339 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13342 2018-02-07 Richard Biener <rguenther@suse.de>
13344 PR tree-optimization/84204
13345 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13348 PR tree-optimization/84205
13349 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13350 special-case isl_ast_op_zdiv_r.
13352 PR tree-optimization/84223
13353 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13354 Only add conditions from within the region.
13355 (gather_bbs::after_dom_children): Adjust.
13357 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13360 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13361 * config/avr/avr.md: Only post-reload split REG-REG moves if
13362 either register is GENERAL_REG_P.
13364 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13366 PR tree-optimization/84235
13367 * tree-ssa-scopedtables.c
13368 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13369 if the subtraction is performed in floating point type where NaNs are
13370 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13371 build 1. Formatting fix.
13373 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13376 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13377 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13378 and skip it regardless of bb boundaries. Use CALL_P macro,
13379 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13380 unnecessarily, formatting fix.
13382 2018-02-06 Michael Collison <michael.collison@arm.com>
13384 * config/arm/thumb2.md:
13385 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13386 (*thumb_mov_notscc): Ditto.
13388 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13391 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13392 unsigned_fix, not "s".
13394 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13396 * configure.ac (gcc_fn_eh_frame_ro): New function.
13397 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13398 correct .eh_frame permissions.
13399 * configure: Regenerate.
13401 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13403 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13404 irrelevant options.
13406 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13408 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13409 Display warning message for -mno-speculate-indirect-jumps.
13411 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13413 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13415 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13417 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13419 PR tree-optimization/84225
13420 * tree-eh.c (find_trapping_overflow): Only call
13421 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13423 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13426 * config/i386/i386.c: Reimplement the check of possible options
13427 -mibt/-mshstk conbination. Change error messages.
13428 * doc/invoke.texi: Fix a typo: remove extra '='.
13430 2018-02-06 Marek Polacek <polacek@redhat.com>
13432 PR tree-optimization/84228
13433 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13435 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13438 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13439 emitted arch directives.
13440 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13441 __ARM_FEATURE_COPROC before changing architectures.
13443 2018-02-06 Richard Biener <rguenther@suse.de>
13445 * config/i386/i386.c (print_reg): Fix typo.
13446 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13448 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13450 * configure: Regenerate.
13452 2018-02-05 Martin Sebor <msebor@redhat.com>
13454 PR tree-optimization/83369
13455 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13458 2018-02-05 Martin Liska <mliska@suse.cz>
13460 * doc/invoke.texi: Cherry-pick upstream r323995.
13462 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13464 * ira.c (ira_init_register_move_cost): Adjust comment.
13466 2018-02-05 Martin Liska <mliska@suse.cz>
13468 PR gcov-profile/84137
13469 * doc/gcov.texi: Fix typo in documentation.
13471 2018-02-05 Martin Liska <mliska@suse.cz>
13473 PR gcov-profile/83879
13474 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13475 using dlopen functionality.
13477 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13479 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13480 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13481 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13482 _mm_maskz_range_round_ss): New intrinsics.
13483 (__builtin_ia32_rangesd128_round)
13484 (__builtin_ia32_rangess128_round): Remove.
13485 (__builtin_ia32_rangesd128_mask_round,
13486 __builtin_ia32_rangess128_mask_round): New builtins.
13487 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13488 __builtin_ia32_rangess128_round): Remove.
13489 (__builtin_ia32_rangesd128_mask_round,
13490 __builtin_ia32_rangess128_mask_round): New builtins.
13491 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13492 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13493 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13494 "<round_saeonly_constraint>")): Changed to ...
13495 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13496 "<round_saeonly_scalar_constraint>")): ... this.
13497 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13498 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13499 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13500 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13501 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13503 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13505 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13507 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13508 Remove all values except native, 8540 and 8548.
13510 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13512 * config/i386/i386.c (ix86_output_function_return): Pass
13513 INVALID_REGNUM, instead of -1, as invalid register number to
13514 indirect_thunk_name and output_indirect_thunk.
13516 2018-02-02 Julia Koval <julia.koval@intel.com>
13518 * config.gcc: Add -march=icelake.
13519 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13520 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13521 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13522 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13523 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13524 (processor_target_table): Add icelake.
13525 (ix86_option_override_internal): Handle new PTAs.
13526 (get_builtin_code_for_version): Handle icelake.
13527 (M_INTEL_COREI7_ICELAKE): New.
13528 (fold_builtin_cpu): Handle icelake.
13529 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13530 * doc/invoke.texi: Add -march=icelake.
13532 2018-02-02 Julia Koval <julia.koval@intel.com>
13534 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13535 to wide_int_bitmask.
13536 * wide-int-bitmask.h: New.
13538 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13541 * config/i386/i386.md: Replace Pmode with word_mode in
13542 builtin_setjmp_setup and builtin_longjmp to support x32.
13544 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13548 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13550 (rs6000_supported_cpu_names): New static variable.
13551 (linux_cpu_translation_table): Likewise.
13552 (elf_platform) <cpu>: Define new static variable and use it.
13553 Translate kernel AT_PLATFORM name to canonical name if needed.
13554 Error if platform name is unknown.
13556 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13559 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13561 2018-02-01 Jeff Law <law@redhat.com>
13564 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13565 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13566 the scratch if RELEASE_VIA_POP is false.
13567 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13568 If we have to save a temporary register, decrement SIZE appropriately.
13569 Pass new arguments to release_scratch_register_on_entry.
13570 (ix86_adjust_stack_and_probe): Likewise.
13571 (ix86_emit_probe_stack_range): Pass new arguments to
13572 release_scratch_register_on_entry.
13574 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13576 PR rtl-optimization/84157
13577 * combine.c (change_zero_ext): Use REG_P predicate in
13578 front of HARD_REGISTER_P predicate.
13580 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13582 * config/avr/avr.c (avr_option_override): Move disabling of
13583 -fdelete-null-pointer-checks to...
13584 * common/config/avr/avr-common.c (avr_option_optimization_table):
13587 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13589 PR tree-optimization/81635
13590 * tree-data-ref.c (split_constant_offset_1): For types that
13591 wrap on overflow, try to use range info to prove that wrapping
13594 2018-02-01 Renlin Li <renlin.li@arm.com>
13597 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13598 TAILCALL_ADDR_REGS.
13599 (aarch64_register_move_cost): Likewise.
13600 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13601 TAILCALL_ADDR_REGS.
13602 (REG_CLASS_NAMES): Likewise.
13603 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13604 TAILCALL_ADDR_REGS. Remove IP registers.
13605 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13607 2018-02-01 Richard Biener <rguenther@suse.de>
13609 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13610 for specifying RPO order and allow NULL for that.
13611 * domwalk.c (dom_walker::dom_walker): Likewise.
13612 (dom_walker::walk): Handle NULL RPO order.
13613 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13615 (rewrite_update_dom_walker): Likewise.
13616 (mark_def_dom_walker): Likewise.
13618 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13620 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13621 (aarch64_maybe_expand_sve_subreg_move): Declare.
13622 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13623 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13625 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13626 that are semantically a reverse operation.
13627 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13628 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13629 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13631 (aarch64_can_change_mode_class): For big-endian, forbid changes
13632 between two SVE modes if they have different element sizes.
13634 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13636 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13637 the TImode handling for big-endian targets.
13639 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13641 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13642 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13644 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13645 Remove BSWAP handing for big-endian targets and use the form of
13646 LD1RQ appropariate for the mode.
13648 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13650 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13651 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13652 duplicated element.
13654 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13657 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13658 check for operands that need to go through aarch64_sve_reload_be.
13660 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13662 PR tree-optimization/81661
13663 PR tree-optimization/84117
13664 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13665 * tree-eh.c: Include gimplify.h.
13666 (find_trapping_overflow, replace_trapping_overflow,
13667 rewrite_to_non_trapping_overflow): New functions.
13668 * tree-vect-loop.c: Include tree-eh.h.
13669 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13670 * tree-data-ref.c: Include tree-eh.h.
13671 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13673 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13675 PR rtl-optimization/84123
13676 * combine.c (change_zero_ext): Check if hard register satisfies
13677 can_change_dest_mode before calling gen_lowpart_SUBREG.
13679 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13682 * ira.c (ira_init_register_move_cost): Remove assert.
13684 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13686 PR rtl-optimization/84071
13687 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13688 * doc/tm.texi: Regenerate.
13690 2018-01-31 Richard Biener <rguenther@suse.de>
13692 PR tree-optimization/84132
13693 * tree-data-ref.c (analyze_miv_subscript): Properly
13694 check whether evolution_function_is_affine_multivariate_p
13695 before calling gcd_of_steps_may_divide_p.
13697 2018-01-31 Julia Koval <julia.koval@intel.com>
13700 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13701 * config/i386/i386.md (rdpid_rex64) New.
13702 (rdpid): Make 32bit only.
13704 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13707 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13708 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13710 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13713 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13715 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13717 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13719 PR rtl-optimization/84071
13720 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13721 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13723 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13725 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13726 (arc_attribute_table): Add 'aux' attribute.
13727 (arc_in_small_data_p): Consider aux like variables.
13728 (arc_is_aux_reg_p): New function.
13729 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13730 (arc_get_aux_arg): New function.
13731 (prepare_move_operands): Handle aux-register access.
13732 (arc_handle_aux_attribute): New function.
13733 * doc/extend.texi (ARC Variable attributes): Add subsection.
13735 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13737 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13738 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13739 (arc_attribute_table): Add 'uncached' attribute.
13740 (arc_print_operand): Print '.di' flag for uncached memory
13742 (arc_in_small_data_p): Do not consider for small data the uncached
13744 (arc_is_uncached_mem_p): New function.
13745 * config/arc/predicates.md (compact_store_memory_operand): Check
13746 for uncached memory accesses.
13747 (nonvol_nonimm_operand): Likewise.
13748 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13750 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13753 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13754 falign-loops=): Add Optimization flag.
13756 2018-01-30 Jeff Law <law@redhat.com>
13759 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13760 INT_REGISTERS_SAVED. Check it prior to calling
13761 get_scratch_register_on_entry.
13762 (ix86_adjust_stack_and_probe): Similarly.
13763 (ix86_emit_probe_stack_range): Similarly.
13764 (ix86_expand_prologue): Corresponding changes.
13766 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13769 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13770 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13772 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13775 * lra-constraints.c (curr_insn_transform): Process AND in the
13778 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13780 PR rtl-optimization/83986
13781 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13782 dependence against last_pending_memory_flush in addition to
13783 pending_jump_insns.
13785 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13787 PR tree-optimization/81611
13788 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13791 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13794 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13797 2018-01-30 Richard Biener <rguenther@suse.de>
13798 Jakub Jelinek <jakub@redhat.com>
13800 PR tree-optimization/84111
13801 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13802 inner loops added during recursion, as they don't have up-to-date
13805 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13808 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13809 (can_inline_edge_by_limits_p): ... here.
13810 (can_early_inline_edge_p, check_callers,
13811 update_caller_keys, update_callee_keys, recursive_inlining,
13812 add_new_edges_to_heap, speculation_useful_p,
13813 inline_small_functions,
13814 inline_small_functions, flatten_function,
13815 inline_to_all_callers_1): Update.
13817 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13819 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13822 2018-01-30 Richard Biener <rguenther@suse.de>
13824 PR tree-optimization/83008
13825 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13826 invariant and constant vector uses in stmts when they need
13827 more than one stmt.
13829 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13832 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13833 * configure: Regenerate.
13835 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13837 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13839 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13840 Use gen_rtx_REG rather than gen_lowpart.
13842 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13844 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13845 rather than 0 when creating partial subregs.
13847 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13849 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13852 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13855 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13856 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13857 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13858 flags. This restores the settings used before the 2017-07-24.
13859 Turning off pre increment/decrement/modify allows IVOPTS to
13860 optimize DF/SF loops where the index is an int.
13862 2018-01-29 Richard Biener <rguenther@suse.de>
13863 Kelvin Nilsen <kelvin@gcc.gnu.org>
13866 * tree-vect-stmts.c (vectorizable_call): Don't call
13867 targetm.vectorize_builtin_md_vectorized_function if callee is
13870 2018-01-22 Carl Love <cel@us.ibm.com>
13872 * doc/extend.tex: Fix typo in second arg in
13873 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13875 2018-01-29 Richard Biener <rguenther@suse.de>
13877 PR tree-optimization/84086
13878 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13879 (flush_ssaname_freelist): When SSA names were released reset
13880 the SCEV hash table.
13882 2018-01-29 Richard Biener <rguenther@suse.de>
13884 PR tree-optimization/84057
13885 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13886 removed paths when removing edges.
13888 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13890 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13891 -mfunction-return=@var{choice}.
13893 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13895 PR diagnostic/84034
13896 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13897 Handle CR like TAB.
13898 (layout::print_source_line): Likewise.
13899 (test_get_line_width_without_trailing_whitespace): Add test cases.
13901 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13903 PR middle-end/84040
13904 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13907 2018-01-26 Jim Wilson <jimw@sifive.com>
13909 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13911 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13914 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13916 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13917 and CMP + SUB-immediate -> SUBS.
13919 2018-01-26 Martin Sebor <msebor@redhat.com>
13921 PR tree-optimization/83896
13922 * tree-ssa-strlen.c (get_string_len): Rename...
13923 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13924 Avoid assuming length is constant.
13925 (handle_char_store): Use HOST_WIDE_INT for string length.
13927 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13930 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13931 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13933 2018-01-26 Richard Biener <rguenther@suse.de>
13935 PR rtl-optimization/84003
13936 * dse.c (record_store): Only record redundant stores when
13937 the earlier store aliases at least all accesses the later one does.
13939 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13941 PR rtl-optimization/83985
13942 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13943 REG_CFA_RESTORE insns.
13944 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13945 REG_CFA_RESTORE insns here.
13948 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13949 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13951 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13953 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13954 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13955 (arc_init): Likewise.
13956 (arc_override_options): Likewise.
13957 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13959 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13961 * config/arc/arc.h (TARGET_DBNZ): Define.
13962 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13963 properly set the tune attribute.
13964 (dbnz): Use TARGET_DBNZ guard.
13965 * config/arc/arc.opt (mtune): Add core3 option.
13967 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13969 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13970 recognize new pic like addresses.
13971 (arc_delegitimize_address): Clean up.
13973 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13975 * config/arc/arc-arches.def: Option mrf16 valid for all
13977 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13978 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13979 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13980 * config/arc/arc-tables.opt: Regenerate.
13981 * config/arc/arc.c (arc_conditional_register_usage): Handle
13982 reduced register file case.
13983 (arc_file_start): Set must have build attributes.
13984 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13985 mrf16 option value.
13986 * config/arc/arc.opt (mrf16): Add new option.
13987 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13988 * config/arc/genmultilib.awk: Handle new mrf16 option.
13989 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13990 * config/arc/t-multilib: Regenerate.
13991 * doc/invoke.texi (ARC Options): Document mrf16 option.
13993 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13995 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13996 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13997 (arc_attribute_table): Add 'secure_call' attribute.
13998 (arc_print_operand): Print secure call operand.
13999 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14001 (arc_is_secure_call_p): New function. * config/arc/arc.md
14002 (call_i): Add support for sjli instruction.
14003 (call_value_i): Likewise.
14004 * config/arc/constraints.md (Csc): New constraint.
14006 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14007 John Eric Martin <John.Martin@emmicro-us.com>
14009 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14010 * config/arc/arc.c (_arc_jli_section): New struct.
14011 (arc_jli_section): New type.
14012 (rc_jli_sections): New static variable.
14013 (arc_handle_jli_attribute): New function.
14014 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14015 (arc_file_end): New function.
14016 (TARGET_ASM_FILE_END): Define.
14017 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14018 (arc_add_jli_section): New function.
14019 (jli_call_scan): Likewise.
14020 (arc_reorg): Call jli_call_scan.
14021 (arc_output_addsi): Remove 'S' from printing asm operand.
14022 (arc_is_jli_call_p): New function.
14023 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14025 (movhi_insn): Likewise.
14026 (movsi_insn): Likewise.
14027 (movsi_set_cc_insn): Likewise.
14028 (loadqi_update): Likewise.
14029 (load_zeroextendqisi_update): Likewise.
14030 (load_signextendqisi_update): Likewise.
14031 (loadhi_update): Likewise.
14032 (load_zeroextendhisi_update): Likewise.
14033 (load_signextendhisi_update): Likewise.
14034 (loadsi_update): Likewise.
14035 (loadsf_update): Likewise.
14036 (movsicc_insn): Likewise.
14037 (bset_insn): Likewise.
14038 (bxor_insn): Likewise.
14039 (bclr_insn): Likewise.
14040 (bmsk_insn): Likewise.
14041 (bicsi3_insn): Likewise.
14042 (cmpsi_cc_c_insn): Likewise.
14043 (movsi_ne): Likewise.
14044 (movsi_cond_exec): Likewise.
14045 (clrsbsi2): Likewise.
14046 (norm_f): Likewise.
14056 (call_i): Remove 'S' asm letter, add jli instruction.
14057 (call_value_i): Likewise.
14058 * config/arc/arc.op (mjli-always): New option.
14059 * config/arc/constraints.md (Cji): New constraint.
14060 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14062 (subsf3_fpx): Likewise.
14063 (mulsf3_fpx): Likewise.
14064 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14066 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14067 function attrbutes.
14068 * doc/invoke.texi (ARC): Document mjli-always option.
14070 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14072 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14073 avoid addition with 0 and use incw and decw where possible.
14075 2018-01-26 Richard Biener <rguenther@suse.de>
14077 PR tree-optimization/81082
14078 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14079 association if it requires casting to unsigned.
14080 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14081 from fold_plusminus_mult_expr to catch important cases late when
14082 range info is available.
14084 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14086 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14087 * configure.ac (hidden_linkonce): New test.
14088 * configure: Regenerate.
14089 * config.in: Regenerate.
14091 2018-01-26 Julia Koval <julia.koval@intel.com>
14093 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14094 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14095 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14096 _mm_mask_bitshuffle_epi64_mask): Fix type.
14097 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14098 USI_FTYPE_V4DI_V4DI_USI): Remove.
14099 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14100 __builtin_ia32_vpshufbitqmb256_mask,
14101 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14102 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14103 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14105 2018-01-26 Alan Modra <amodra@gmail.com>
14108 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14109 UNSPEC_VBPERMQ. Sort other unspecs.
14111 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14113 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14115 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14117 PR middle-end/83055
14118 * predict.c (drop_profile): Do not push/pop cfun; update also
14120 (handle_missing_profiles): Fix logic looking for zero profiles.
14122 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14124 PR middle-end/83977
14125 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14126 on functions with #pragma omp declare simd or functions with simd
14128 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14129 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14130 Remove trailing \n from warning_at calls.
14132 2018-01-25 Tom de Vries <tom@codesourcery.com>
14135 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14136 for neutered workers.
14138 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14141 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14142 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14144 2018-01-24 Jeff Law <law@redhat.com>
14147 * i386.c (get_probe_interval): Move to earlier point.
14148 (ix86_compute_frame_layout): If -fstack-clash-protection and
14149 the frame is larger than the probe interval, then use pushes
14150 to save registers rather than reg->mem moves.
14151 (ix86_expand_prologue): Remove conditional for int_registers_saved
14154 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14157 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14158 min/max for never referenced object.
14160 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14162 PR middle-end/83977
14163 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14165 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14166 attributes from DECL_ATTRIBUTES (decl) without affecting
14167 DECL_ATTRIBUTES (current_function_decl).
14168 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14169 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14171 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14173 PR tree-optimization/83979
14174 * fold-const.c (fold_comparison): Use constant_boolean_node
14175 instead of boolean_{true,false}_node.
14177 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14179 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14182 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14184 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14185 Simplify the clause that sets the length attribute.
14186 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14187 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14188 clause that sets the length attribute.
14189 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14191 2018-01-24 Tom de Vries <tom@codesourcery.com>
14194 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14195 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14196 Add strict parameter.
14197 (prevent_branch_around_nothing): Insert dummy insn between branch to
14198 label and label with no ptx insn inbetween.
14199 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14201 2018-01-24 Tom de Vries <tom@codesourcery.com>
14204 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14205 for neutered threads in warp.
14206 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14208 2018-01-24 Richard Biener <rguenther@suse.de>
14210 PR tree-optimization/83176
14211 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14214 2018-01-24 Richard Biener <rguenther@suse.de>
14216 PR tree-optimization/82819
14217 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14218 code generating pluses that are no-ops in the target precision.
14220 2018-01-24 Richard Biener <rguenther@suse.de>
14222 PR middle-end/84000
14223 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14225 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14227 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14228 to merge probabilities.
14229 * predict.c (probably_never_executed): Also mark as cold functions
14230 with global 0 profile and guessed local profile.
14231 * profile-count.c (profile_probability::combine_with_count): New
14233 * profile-count.h (profile_probability::operator*,
14234 profile_probability::operator*=, profile_probability::operator/,
14235 profile_probability::operator/=): Reduce precision to adjusted
14236 and set value to guessed on contradictory divisions.
14237 (profile_probability::combine_with_freq): Remove.
14238 (profile_probability::combine_wiht_count): Declare.
14239 (profile_count::force_nonzero):: Set to adjusted.
14240 (profile_count::probability_in):: Set quality to adjusted.
14241 * tree-ssa-tail-merge.c (replace_block_by): Use
14242 combine_with_count.
14244 2018-01-23 Andrew Waterman <andrew@sifive.com>
14245 Jim Wilson <jimw@sifive.com>
14247 * config/riscv/riscv.c (riscv_stack_boundary): New.
14248 (riscv_option_override): Set riscv_stack_boundary. Handle
14249 riscv_preferred_stack_boundary_arg.
14250 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14251 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14252 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14253 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14254 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14255 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14257 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14260 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14261 of struct ix86_frame.
14262 (ix86_expand_epilogue): Likewise. Add a local variable for
14263 the reg_save_offset field in struct ix86_frame.
14265 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14267 PR tree-optimization/82604
14268 * tree-loop-distribution.c (enum partition_kind): New enum item
14269 PKIND_PARTIAL_MEMSET.
14270 (partition_builtin_p): Support above new enum item.
14271 (generate_code_for_partition): Ditto.
14272 (compute_access_range): Differentiate cases that equality can be
14273 proven at all loops, the innermost loops or no loops.
14274 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14275 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14276 (finalize_partitions, distribute_loop): Don't fuse partition of
14277 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14278 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14279 parloop is enabled.
14281 2018-01-23 Martin Liska <mliska@suse.cz>
14283 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14284 order to ignore the predictor.
14285 (PRED_POLYMORPHIC_CALL): Likewise.
14286 (PRED_RECURSIVE_CALL): Likewise.
14288 2018-01-23 Martin Liska <mliska@suse.cz>
14290 * tree-profile.c (tree_profiling): Print function header to
14291 aware reader which function we are working on.
14292 * value-prof.c (gimple_find_values_to_profile): Do not print
14293 not interesting value histograms.
14295 2018-01-23 Martin Liska <mliska@suse.cz>
14297 * profile-count.h (enum profile_quality): Add
14298 profile_uninitialized as the first value. Do not number values
14299 as they are zero based.
14300 (profile_count::verify): Update sanity check.
14301 (profile_probability::verify): Likewise.
14303 2018-01-23 Nathan Sidwell <nathan@acm.org>
14305 * doc/invoke.texi (ffor-scope): Deprecate.
14307 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14309 PR tree-optimization/83510
14310 * domwalk.c (set_all_edges_as_executable): New function.
14311 (dom_walker::dom_walker): Convert bool param
14312 "skip_unreachable_blocks" to enum reachability. Move setup of
14313 edge flags to set_all_edges_as_executable and only do it when
14314 reachability is REACHABLE_BLOCKS.
14315 * domwalk.h (enum dom_walker::reachability): New enum.
14316 (dom_walker::dom_walker): Convert bool param
14317 "skip_unreachable_blocks" to enum reachability.
14318 (set_all_edges_as_executable): New decl.
14319 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14320 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14322 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14323 but converting true to REACHABLE_BLOCKS.
14324 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14326 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14327 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14328 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14330 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14331 if check_all_array_refs will be called.
14333 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14335 * tree.c (selftest::test_location_wrappers): Add more test
14338 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14340 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14341 (selftest::test_bit_in_range): Likewise.
14343 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14346 * doc/sourcebuild.texi (vect_float): Say that the selector
14347 only describes the situation when -funsafe-math-optimizations is on.
14348 (vect_float_strict): Document.
14350 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14352 PR tree-optimization/83965
14353 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14354 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14355 instead of checking only for a reduction.
14356 (vect_recog_widen_sum_pattern): Likewise.
14358 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14360 * predict.c (probably_never_executed): Only use precise profile info.
14361 (compute_function_frequency): Skip after inlining hack since we now
14362 have quality checking.
14364 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14366 * profile-count.h (profile_probability::very_unlikely,
14367 profile_probability::unlikely, profile_probability::even): Set
14368 precision to guessed.
14370 2018-01-23 Richard Biener <rguenther@suse.de>
14372 PR tree-optimization/83963
14373 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14374 Properly terminate dominator walk when crossing the exit edge not
14375 when visiting its source block.
14377 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14380 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14381 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14383 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14385 PR tree-optimization/83957
14386 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14387 semicolon after for body surrounded by braces.
14389 PR tree-optimization/83081
14390 * profile-count.h (profile_probability::split): New method.
14391 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14392 Use profile_probability::split.
14393 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14394 when splitting a single conditional jump into 2.
14396 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14398 PR tree-optimization/69452
14399 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14404 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14405 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14406 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14408 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14410 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14412 * config/rl78/rl78.md (movdi): New define_expand.
14413 * config/rl78/rl78.c (rl78_split_movdi): New function.
14415 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14418 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14420 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14421 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14422 128-bit to produce an UNSPEC move to get the double word with the
14423 signbit and then a shift directly to do signbit.
14424 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14425 implementation with a new version that just does either a direct
14426 move or a regular move. Move memory interface to separate insns.
14427 Move insns so they are next to the expander.
14428 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14429 with signbit move. Split big and little endian case.
14430 (signbit<mode>2_dm_mem_le): Likewise.
14431 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14432 (signbit<mode>2_dm2): Likewise.
14434 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14436 * config/rl78/rl78.md (anddi3): New define_expand.
14438 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14440 * config/rl78/rl78.md (umindi3): New define_expand.
14442 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14444 * config/rl78/rl78.md (smindi3): New define_expand.
14446 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14448 * config/rl78/rl78.md (smaxdi3): New define_expand.
14450 2018-01-22 Carl Love <cel@us.ibm.com>
14452 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14453 LVX_V1TI): Add macro expansion.
14454 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14455 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14456 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14457 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14458 Change check to determine if the instruction is a byte reversing
14459 entry. Fix typo in comment.
14460 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14461 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14462 Add def_builtin calls for new builtins.
14463 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14464 Add define_insn expansion.
14466 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14468 * config/rl78/rl78.md (umaxdi3): New define_expand.
14470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14472 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14473 for non-QImode registers.
14475 2018-01-22 Richard Biener <rguenther@suse.de>
14477 PR tree-optimization/83963
14478 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14479 including the loop exit block.
14480 (scop_detection::merge_sese): Likewise.
14481 (scop_detection::add_scop): Do it here instead.
14483 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14485 * doc/sourcebuild.texi (arm_softfloat): Document.
14487 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14490 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14491 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14492 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14494 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14495 David Edelsohn <dje.gcc@gmail.com>
14498 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14499 Change "crset eq" to "crset 2".
14500 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14501 (*call_indirect_aix<mode>_nospec): Likewise.
14502 (*call_value_indirect_aix<mode>_nospec): Likewise.
14503 (*call_indirect_elfv2<mode>_nospec): Likewise.
14504 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14505 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14506 change assembly output from . to $.
14507 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14508 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14509 (*tablejump<mode>_internal1_nospec): Likewise.
14511 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14514 * config/sh/sh_optimize_sett_clrt.cc:
14515 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14517 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14519 PR tree-optimization/83940
14520 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14521 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14522 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14523 use it to pass back the definition type.
14524 (vect_check_store_rhs): Likewise rhs_dt_out.
14525 (vect_build_gather_load_calls): Add a mask_dt argument and use
14526 it instead of a call to vect_is_simple_use.
14527 (vectorizable_store): Update calls to vect_check_load_store_mask
14528 and vect_check_store_rhs. Use the dt returned by the latter instead
14529 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14530 instead of calls to vect_is_simple_use. Pass the scalar rather
14531 than the vector operand to vect_is_simple_use when handling
14532 second and subsequent copies of an rhs value.
14533 (vectorizable_load): Update calls to vect_check_load_store_mask
14534 and vect_build_gather_load_calls. Use the cached mask_dt and
14535 gs_info.offset_dt instead of calls to vect_is_simple_use.
14537 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14539 PR middle-end/83945
14540 * tree-emutls.c: Include gimplify.h.
14541 (lower_emutls_2): New function.
14542 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14543 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14544 it before further processing.
14547 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14548 UINTVAL (trueop1) instead of INTVAL (op1).
14550 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14554 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14555 INCOMING_FRAME_SP_OFFSET if not defined.
14556 (scan_trace): Add ENTRY argument. If true and
14557 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14558 emit a note to adjust the CFA offset.
14559 (create_cfi_notes): Adjust scan_trace callers.
14560 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14561 INCOMING_FRAME_SP_OFFSET in the CIE.
14562 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14563 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14565 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14566 * doc/tm.texi: Regenerated.
14568 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14570 PR rtl-optimization/83147
14571 * lra-constraints.c (remove_inheritance_pseudos): Use
14572 lra_substitute_pseudo_within_insn.
14574 2018-01-19 Tom de Vries <tom@codesourcery.com>
14575 Cesar Philippidis <cesar@codesourcery.com>
14578 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14580 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14583 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14584 spaces for function labels.
14586 2018-01-19 Martin Liska <mliska@suse.cz>
14588 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14589 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14590 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14591 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14592 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14593 (PRED_CONST_RETURN): Change from 69 to 65.
14594 (PRED_NULL_RETURN): Change from 91 to 71.
14595 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14596 (PRED_LOOP_GUARD): Change from 66 to 73.
14598 2018-01-19 Martin Liska <mliska@suse.cz>
14600 * predict.c (predict_insn_def): Add new assert.
14601 (struct branch_predictor): Change type to signed integer.
14602 (test_prediction_value_range): Amend test to cover
14603 PROB_UNINITIALIZED.
14604 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14605 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14606 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14607 (PRED_LOOP_IV_COMPARE): Likewise.
14608 * predict.h (PROB_UNINITIALIZED): Define new constant.
14610 2018-01-19 Martin Liska <mliska@suse.cz>
14612 * predict.c (dump_prediction): Add new format for
14613 analyze_brprob.py script which is enabled with -details
14615 * profile-count.h (precise_p): New function.
14617 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14619 PR tree-optimization/83922
14620 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14621 there are no statements that need masking.
14622 (vect_active_double_reduction_p): New function.
14623 (vect_analyze_loop_operations): Use it when handling phis that
14624 are not in the loop header.
14626 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14628 PR tree-optimization/83914
14629 * tree-vect-loop.c (vectorizable_induction): Don't convert
14630 init_expr or apply the peeling adjustment for inductions
14631 that are nested within the vectorized loop.
14633 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14635 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14638 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14642 * function.h (gimplify_parameters): Add gimple_seq * argument.
14643 * function.c: Include gimple.h and options.h.
14644 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14645 for the added local temporaries if needed.
14646 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14647 if there are any parameter cleanups, wrap whole body into a
14648 try/finally with the cleanups.
14650 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14653 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14654 Use GET_MODE_CLASS for scalar floating point.
14656 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14660 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14661 Fix call of call_cgraph_insertion_hooks.
14663 2018-01-18 Martin Sebor <msebor@redhat.com>
14665 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14667 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14670 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14673 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14676 * common.opt: (-ffile-prefix-map): New option.
14677 * opts.c (common_handle_option): Defer it.
14678 * opts-global.c (handle_common_deferred_options): Handle it.
14679 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14680 * file-prefix-map.h: New file.
14681 (remap_debug_filename, add_debug_prefix_map): ...here.
14682 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14683 * final.c (debug_prefix_map, add_debug_prefix_map
14684 remap_debug_filename): Move to...
14685 * file-prefix-map.c: New file.
14686 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14687 generalize, get rid of alloca(), use strrchr() instead of strchr().
14688 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14689 Implement in terms of add_prefix_map().
14690 (remap_macro_filename, remap_debug_filename): Implement in term of
14692 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14693 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14694 * dbxout.c: Include file-prefix-map.h.
14695 * varasm.c: Likewise.
14696 * vmsdbgout.c: Likewise.
14697 * xcoffout.c: Likewise.
14698 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14699 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14700 * doc/invoke.texi (-ffile-prefix-map): Document.
14701 (-fdebug-prefix-map): Update description.
14703 2018-01-18 Martin Liska <mliska@suse.cz>
14705 * config/i386/i386.c (indirect_thunk_name): Document that also
14707 (output_indirect_thunk): Document why both instructions
14708 (pause and lfence) are generated.
14710 2018-01-18 Richard Biener <rguenther@suse.de>
14712 PR tree-optimization/83887
14713 * graphite-scop-detection.c
14714 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14715 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14716 (scop_detection::merge_sese): Re-implement with a flood-fill
14717 algorithm that properly finds a SESE region if it exists.
14719 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14722 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14723 pointer_diff optimizations use view_convert instead of convert.
14725 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14727 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14728 Generate different code for -mno-speculate-indirect-jumps.
14729 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14730 (*call_indirect_aix<mode>): Disable for
14731 -mno-speculate-indirect-jumps.
14732 (*call_indirect_aix<mode>_nospec): New define_insn.
14733 (*call_value_indirect_aix<mode>): Disable for
14734 -mno-speculate-indirect-jumps.
14735 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14736 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14737 -mno-speculate-indirect-jumps.
14738 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14740 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14742 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14743 long double type, set the flags for noting the default long double
14744 type, even if we don't pass or return a long double type.
14746 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14749 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14752 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14754 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14755 support for merge[hl].
14756 (fold_mergehl_helper): New helper function.
14757 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14758 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14759 (altivec_vmrglw_direct): Add xxmrglw insn.
14761 2018-01-17 Andrew Waterman <andrew@sifive.com>
14763 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14764 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14766 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14769 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14770 call the lto_location_cache before reading the
14771 DECL_SOURCE_LOCATION of the types.
14773 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14774 Richard Sandiford <richard.sandiford@linaro.org>
14776 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14777 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14778 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14779 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14780 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14782 * config/aarch64/constraints.md (aarch64_movti_operand):
14784 * config/aarch64/predicates.md (Uti): Add new constraint.
14786 2018-01-17 Carl Love <cel@us.ibm.com>
14788 * config/rs6000/vsx.md (define_expand xl_len_r,
14789 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14790 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14792 (define_expand, define_insn): Move the shift left from the
14793 define_insn to the define_expand for lxvl and stxvl instructions.
14794 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14795 and XL_LEN_R definitions to PURE.
14797 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14799 * config/i386/i386.c (indirect_thunk_name): Declare regno
14800 as unsigned int. Compare regno with INVALID_REGNUM.
14801 (output_indirect_thunk): Ditto.
14802 (output_indirect_thunk_function): Ditto.
14803 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14804 in the call to output_indirect_thunk_function.
14806 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14808 PR middle-end/83884
14809 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14810 rather than the size of inner_type to determine the stack slot size
14811 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14813 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14816 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14819 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14821 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14822 endian Linux systems to optionally enable multilibs for selecting
14823 the long double type if the user configured an explicit type.
14824 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14825 have no long double multilibs if not defined.
14826 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14827 warn if the user used -mabi={ieee,ibm}longdouble and we built
14828 multilibs for long double.
14829 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14830 appropriate multilib option.
14831 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14833 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14834 for building long double multilibs.
14835 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14837 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14839 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14842 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14844 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14847 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14850 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14853 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14855 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14856 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14858 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14860 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14861 different rtl trees depending on TARGET_64BIT.
14862 (rs6000_gen_lvx): Likewise.
14864 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14866 * config/visium/visium.md (nop): Tweak comment.
14867 (hazard_nop): Likewise.
14869 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14871 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14872 -mspeculate-indirect-jumps.
14873 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14874 for -mno-speculate-indirect-jumps.
14875 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14876 (*call_value_indirect_elfv2<mode>): Disable for
14877 -mno-speculate-indirect-jumps.
14878 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14879 (indirect_jump): Emit different RTL for
14880 -mno-speculate-indirect-jumps.
14881 (*indirect_jump<mode>): Disable for
14882 -mno-speculate-indirect-jumps.
14883 (*indirect_jump<mode>_nospec): New define_insn.
14884 (tablejump): Emit different RTL for
14885 -mno-speculate-indirect-jumps.
14886 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14887 (tablejumpsi_nospec): New define_expand.
14888 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14889 (tablejumpdi_nospec): New define_expand.
14890 (*tablejump<mode>_internal1): Disable for
14891 -mno-speculate-indirect-jumps.
14892 (*tablejump<mode>_internal1_nospec): New define_insn.
14893 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14896 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14898 * caller-save.c (insert_save): Drop unnecessary parameter. All
14901 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14902 Richard Biener <rguenth@suse.de>
14905 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14906 return early, inline manually is_gimple_sizepos. Make sure if we
14907 call gimplify_expr we don't end up with a gimple constant.
14908 * tree.c (variably_modified_type_p): Don't return true for
14909 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14910 * gimplify.h (is_gimple_sizepos): Remove.
14912 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14914 PR tree-optimization/83857
14915 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14916 vectorizable_live_operation for pure SLP statements.
14917 (vectorizable_live_operation): Handle PHIs.
14919 2018-01-16 Richard Biener <rguenther@suse.de>
14921 PR tree-optimization/83867
14922 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14923 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14925 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14928 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14929 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14930 If off is not INTEGER_CST, issue a may not be aligned warning
14931 rather than isn't aligned. Use isn%'t rather than isn't.
14932 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14934 <case MULT_EXPR>: Improve the case when bottom and one of the
14935 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14936 operand, in that case check if the other operand is multiple of
14937 bottom divided by the INTEGER_CST operand.
14939 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14942 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14943 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14944 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14945 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14946 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14947 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14948 (pa_function_arg_size): New function.
14950 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14952 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14953 in a separate statement.
14955 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14957 PR tree-optimization/83847
14958 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14959 group gathers and scatters.
14961 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14963 PR rtl-optimization/86620
14964 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14966 PR rtl-optimization/83213
14967 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14968 to last if both are JUMP_INSNs.
14970 PR tree-optimization/83843
14971 * gimple-ssa-store-merging.c
14972 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14973 store_immediate_info for bswap/nop orig_stores.
14975 2018-01-15 Andrew Waterman <andrew@sifive.com>
14977 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14979 <UDIV>: Increase cost if !TARGET_DIV.
14981 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14983 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14984 (define_attr "cr_logical_3op"): New.
14985 (cceq_ior_compare): Adjust.
14986 (cceq_ior_compare_complement): Adjust.
14987 (*cceq_rev_compare): Adjust.
14988 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14989 (is_cracked_insn): Adjust.
14990 (insn_must_be_first_in_group): Adjust.
14991 * config/rs6000/40x.md: Adjust.
14992 * config/rs6000/440.md: Adjust.
14993 * config/rs6000/476.md: Adjust.
14994 * config/rs6000/601.md: Adjust.
14995 * config/rs6000/603.md: Adjust.
14996 * config/rs6000/6xx.md: Adjust.
14997 * config/rs6000/7450.md: Adjust.
14998 * config/rs6000/7xx.md: Adjust.
14999 * config/rs6000/8540.md: Adjust.
15000 * config/rs6000/cell.md: Adjust.
15001 * config/rs6000/e300c2c3.md: Adjust.
15002 * config/rs6000/e500mc.md: Adjust.
15003 * config/rs6000/e500mc64.md: Adjust.
15004 * config/rs6000/e5500.md: Adjust.
15005 * config/rs6000/e6500.md: Adjust.
15006 * config/rs6000/mpc.md: Adjust.
15007 * config/rs6000/power4.md: Adjust.
15008 * config/rs6000/power5.md: Adjust.
15009 * config/rs6000/power6.md: Adjust.
15010 * config/rs6000/power7.md: Adjust.
15011 * config/rs6000/power8.md: Adjust.
15012 * config/rs6000/power9.md: Adjust.
15013 * config/rs6000/rs64.md: Adjust.
15014 * config/rs6000/titan.md: Adjust.
15016 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15018 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15019 ix86_indirect_branch_register logic.
15021 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15023 * config/i386/constraints.md (Bs): Update
15024 ix86_indirect_branch_register check. Don't check
15025 ix86_indirect_branch_register with GOT_memory_operand.
15027 * config/i386/predicates.md (GOT_memory_operand): Don't check
15028 ix86_indirect_branch_register here.
15029 (GOT32_symbol_operand): Likewise.
15031 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15033 * config/i386/predicates.md (constant_call_address_operand):
15034 Rewrite ix86_indirect_branch_register logic.
15035 (sibcall_insn_operand): Likewise.
15037 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15039 * config/i386/constraints.md (Bs): Replace
15040 ix86_indirect_branch_thunk_register with
15041 ix86_indirect_branch_register.
15043 * config/i386/i386.md (indirect_jump): Likewise.
15044 (tablejump): Likewise.
15045 (*sibcall_memory): Likewise.
15046 (*sibcall_value_memory): Likewise.
15047 Peepholes of indirect call and jump via memory: Likewise.
15048 * config/i386/i386.opt: Likewise.
15049 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15050 (GOT_memory_operand): Likewise.
15051 (call_insn_operand): Likewise.
15052 (sibcall_insn_operand): Likewise.
15053 (GOT32_symbol_operand): Likewise.
15055 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15057 PR middle-end/83837
15058 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15059 type rather than type addr's type points to.
15060 (expand_omp_atomic_mutex): Likewise.
15061 (expand_omp_atomic): Likewise.
15063 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15066 * config/i386/i386.c (output_indirect_thunk_function): Use
15067 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15068 for __x86_return_thunk.
15070 2018-01-15 Richard Biener <rguenther@suse.de>
15072 PR middle-end/83850
15073 * expmed.c (extract_bit_field_1): Fix typo.
15075 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15078 * config/arm/iterators.md (VF): New mode iterator.
15079 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15080 Remove integer-related logic from pattern.
15081 (neon_vabd<mode>_3): Likewise.
15083 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15085 PR middle-end/82694
15086 * common.opt (fstrict-overflow): No longer an alias.
15087 (fwrapv-pointer): New option.
15088 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15089 also for pointer types based on flag_wrapv_pointer.
15090 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15091 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15092 opts->x_flag_wrapv got set.
15093 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15094 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15095 POINTER_TYPE_OVERFLOW_UNDEFINED.
15096 * match.pd: Likewise in address comparison pattern.
15097 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15099 2018-01-15 Richard Biener <rguenther@suse.de>
15102 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15103 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15104 Reset type names to their identifier if their TYPE_DECL doesn't
15105 have linkage (and thus is used for ODR and devirt).
15106 (save_debug_info_for_decl): Remove.
15107 (save_debug_info_for_type): Likewise.
15108 (add_tree_to_fld_list): Adjust.
15109 * tree-pretty-print.c (dump_generic_node): Make dumping of
15110 type names more robust.
15112 2018-01-15 Richard Biener <rguenther@suse.de>
15114 * BASE-VER: Bump to 8.0.1.
15116 2018-01-14 Martin Sebor <msebor@redhat.com>
15119 * builtins.c (check_access): Avoid warning when the no-warning bit
15122 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15124 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15125 * ira-color (allocno_hard_regs_compare): Likewise.
15127 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15130 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15131 Use .pushsection/.popsection.
15133 2018-01-14 Martin Sebor <msebor@redhat.com>
15136 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15138 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15140 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15141 entry from extra_headers.
15142 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15143 extra_headers, make the list bitwise identical to the i?86-*-* one.
15145 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15147 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15148 -mcmodel=large with -mindirect-branch=thunk,
15149 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15150 -mfunction-return=thunk-extern.
15151 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15152 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15153 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15155 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15157 * config/i386/i386.c (print_reg): Print the name of the full
15158 integer register without '%'.
15159 (ix86_print_operand): Handle 'V'.
15160 * doc/extend.texi: Document 'V' modifier.
15162 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15164 * config/i386/constraints.md (Bs): Disallow memory operand for
15165 -mindirect-branch-register.
15167 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15168 (GOT_memory_operand): Likewise.
15169 (call_insn_operand): Likewise.
15170 (sibcall_insn_operand): Likewise.
15171 (GOT32_symbol_operand): Likewise.
15172 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15173 for -mindirect-branch-register.
15174 (tablejump): Likewise.
15175 (*sibcall_memory): Likewise.
15176 (*sibcall_value_memory): Likewise.
15177 Disallow peepholes of indirect call and jump via memory for
15178 -mindirect-branch-register.
15179 (*call_pop): Replace m with Bw.
15180 (*call_value_pop): Likewise.
15181 (*sibcall_pop_memory): Replace m with Bs.
15182 * config/i386/i386.opt (mindirect-branch-register): New option.
15183 * doc/invoke.texi: Document -mindirect-branch-register option.
15185 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15187 * config/i386/i386-protos.h (ix86_output_function_return): New.
15188 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15189 set function_return_type.
15190 (indirect_thunk_name): Add ret_p to indicate thunk for function
15192 (output_indirect_thunk_function): Pass false to
15193 indirect_thunk_name.
15194 (ix86_output_indirect_branch_via_reg): Likewise.
15195 (ix86_output_indirect_branch_via_push): Likewise.
15196 (output_indirect_thunk_function): Create alias for function
15197 return thunk if regno < 0.
15198 (ix86_output_function_return): New function.
15199 (ix86_handle_fndecl_attribute): Handle function_return.
15200 (ix86_attribute_table): Add function_return.
15201 * config/i386/i386.h (machine_function): Add
15202 function_return_type.
15203 * config/i386/i386.md (simple_return_internal): Use
15204 ix86_output_function_return.
15205 (simple_return_internal_long): Likewise.
15206 * config/i386/i386.opt (mfunction-return=): New option.
15207 (indirect_branch): Mention -mfunction-return=.
15208 * doc/extend.texi: Document function_return function attribute.
15209 * doc/invoke.texi: Document -mfunction-return= option.
15211 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15213 * config/i386/i386-opts.h (indirect_branch): New.
15214 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15215 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15216 with local indirect jump when converting indirect call and jump.
15217 (ix86_set_indirect_branch_type): New.
15218 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15219 (indirectlabelno): New.
15220 (indirect_thunk_needed): Likewise.
15221 (indirect_thunk_bnd_needed): Likewise.
15222 (indirect_thunks_used): Likewise.
15223 (indirect_thunks_bnd_used): Likewise.
15224 (INDIRECT_LABEL): Likewise.
15225 (indirect_thunk_name): Likewise.
15226 (output_indirect_thunk): Likewise.
15227 (output_indirect_thunk_function): Likewise.
15228 (ix86_output_indirect_branch_via_reg): Likewise.
15229 (ix86_output_indirect_branch_via_push): Likewise.
15230 (ix86_output_indirect_branch): Likewise.
15231 (ix86_output_indirect_jmp): Likewise.
15232 (ix86_code_end): Call output_indirect_thunk_function if needed.
15233 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15235 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15236 (ix86_attribute_table): Add indirect_branch.
15237 * config/i386/i386.h (machine_function): Add indirect_branch_type
15238 and has_local_indirect_jump.
15239 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15241 (tablejump): Likewise.
15242 (*indirect_jump): Use ix86_output_indirect_jmp.
15243 (*tablejump_1): Likewise.
15244 (simple_return_indirect_internal): Likewise.
15245 * config/i386/i386.opt (mindirect-branch=): New option.
15246 (indirect_branch): New.
15249 (thunk-inline): Likewise.
15250 (thunk-extern): Likewise.
15251 * doc/extend.texi: Document indirect_branch function attribute.
15252 * doc/invoke.texi: Document -mindirect-branch= option.
15254 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15257 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15259 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15261 * ipa-inline.c (want_inline_small_function_p): Return false if
15262 inlining has already failed with CIF_FINAL_ERROR.
15263 (update_caller_keys): Call want_inline_small_function_p before
15265 (update_callee_keys): Likewise.
15267 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15269 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15271 (rs6000_quadword_masked_address_p): Likewise.
15272 (quad_aligned_load_p): Likewise.
15273 (quad_aligned_store_p): Likewise.
15274 (const_load_sequence_p): Add comment to describe the outer-most loop.
15275 (mimic_memory_attributes_and_flags): New function.
15276 (rs6000_gen_stvx): Likewise.
15277 (replace_swapped_aligned_store): Likewise.
15278 (rs6000_gen_lvx): Likewise.
15279 (replace_swapped_aligned_load): Likewise.
15280 (replace_swapped_load_constant): Capitalize argument name in
15281 comment describing this function.
15282 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15283 and stores that access quad-word aligned addresses and replace
15284 with stvx or lvx instructions when appropriate.
15285 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15286 New function prototype.
15287 (rs6000_quadword_masked_address_p): Likewise.
15288 (rs6000_gen_lvx): Likewise.
15289 (rs6000_gen_stvx): Likewise.
15290 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15291 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15292 when memory address is aligned.
15293 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15294 this split to select lvx instruction when memory address is aligned.
15295 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15296 instruction when memory address is aligned.
15297 (*vsx_le_perm_load_v16qi): Likewise.
15298 (four unnamed splitters): Modify to select the stvx instruction
15299 when memory is aligned.
15301 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15303 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15304 which appears in the queue multiple times.
15306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15307 Alan Hayward <alan.hayward@arm.com>
15308 David Sherwood <david.sherwood@arm.com>
15310 * tree-vectorizer.h (vec_lower_bound): New structure.
15311 (_loop_vec_info): Add check_nonzero and lower_bounds.
15312 (LOOP_VINFO_CHECK_NONZERO): New macro.
15313 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15314 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15315 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15316 fields. Make seg_len the distance travelled, not including the
15318 (dr_direction_indicator): Declare.
15319 (dr_zero_step_indicator): Likewise.
15320 (dr_known_forward_stride_p): Likewise.
15321 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15323 (runtime_alias_check_p): Allow runtime alias checks with
15325 (operator ==): Compare access_size and align.
15326 (prune_runtime_alias_test_list): Rework for new distinction between
15327 the access_size and seg_len.
15328 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15330 (get_segment_min_max): New function.
15331 (create_intersect_range_checks): Use it.
15332 (dr_step_indicator): New function.
15333 (dr_direction_indicator): Likewise.
15334 (dr_zero_step_indicator): Likewise.
15335 (dr_known_forward_stride_p): Likewise.
15336 * tree-loop-distribution.c (data_ref_segment_size): Return
15337 DR_STEP * (niters - 1).
15338 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15340 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15341 (vect_preserves_scalar_order_p): New function, split out from...
15342 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15343 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15344 (vect_vfa_access_size): New function.
15345 (vect_vfa_align): Likewise.
15346 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15347 (dump_lower_bound): New function.
15348 (vect_check_lower_bound): Likewise.
15349 (vect_small_gap_p): Likewise.
15350 (vectorizable_with_step_bound_p): Likewise.
15351 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15352 depencies if the vectorization factor is 1. Convert the checks
15353 for nonzero steps into checks on the bounds of DR_STEP. Try using
15354 a bunds check for variable steps if the minimum required step is
15355 relatively small. Update calls to the dr_with_seg_len
15356 constructor and to vect_compile_time_alias.
15357 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15359 (vect_loop_versioning): Call it.
15360 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15362 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15365 Alan Hayward <alan.hayward@arm.com>
15366 David Sherwood <david.sherwood@arm.com>
15368 * doc/sourcebuild.texi (vect_scatter_store): Document.
15369 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15371 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15373 * genopinit.c (main): Add supports_vec_scatter_store and
15374 supports_vec_scatter_store_cached to target_optabs.
15375 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15376 IFN_MASK_SCATTER_STORE.
15377 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15379 * internal-fn.h (internal_store_fn_p): Declare.
15380 (internal_fn_stored_value_index): Likewise.
15381 * internal-fn.c (scatter_store_direct): New macro.
15382 (expand_scatter_store_optab_fn): New function.
15383 (direct_scatter_store_optab_supported_p): New macro.
15384 (internal_store_fn_p): New function.
15385 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15386 IFN_MASK_SCATTER_STORE.
15387 (internal_fn_mask_index): Likewise.
15388 (internal_fn_stored_value_index): New function.
15389 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15390 for scatter stores.
15391 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15392 * optabs-query.c (supports_vec_scatter_store_p): New function.
15393 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15394 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15395 true for scatter stores.
15396 (vect_gather_scatter_fn_p): Handle scatter stores too.
15397 (vect_check_gather_scatter): Consider using scatter stores if
15398 supports_vec_scatter_store_p.
15399 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15400 scatter stores too.
15401 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15402 internal_fn_stored_value_index.
15403 (check_load_store_masking): Handle scatter stores too.
15404 (vect_get_store_rhs): Make public.
15405 (vectorizable_call): Use internal_store_fn_p.
15406 (vectorizable_store): Handle scatter store internal functions.
15407 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15408 when deciding whether the end of the group has been reached.
15409 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15410 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15411 (mask_scatter_store<mode>): New insns.
15413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15414 Alan Hayward <alan.hayward@arm.com>
15415 David Sherwood <david.sherwood@arm.com>
15417 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15418 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15419 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15421 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15422 Use vect_truncate_gather_scatter_offset if we can't treat the
15423 operation as a normal gather load or scatter store.
15424 (get_group_load_store_type): Take the gather_scatter_info
15425 as argument. Try using a gather load or scatter store for
15426 single-element groups.
15427 (get_load_store_type): Update calls to get_group_load_store_type
15428 and vect_use_strided_gather_scatters_p.
15430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15431 Alan Hayward <alan.hayward@arm.com>
15432 David Sherwood <david.sherwood@arm.com>
15434 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15435 optional tree argument.
15436 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15438 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15439 but continue to use the current value as a fallback.
15440 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15441 to compare the updates.
15442 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15443 (get_load_store_type): Use it when handling a strided access.
15444 (vect_get_strided_load_store_ops): New function.
15445 (vect_get_data_ptr_increment): Likewise.
15446 (vectorizable_load): Handle strided gather loads. Always pass
15447 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15450 Alan Hayward <alan.hayward@arm.com>
15451 David Sherwood <david.sherwood@arm.com>
15453 * doc/md.texi (gather_load@var{m}): Document.
15454 (mask_gather_load@var{m}): Likewise.
15455 * genopinit.c (main): Add supports_vec_gather_load and
15456 supports_vec_gather_load_cached to target_optabs.
15457 * optabs-tree.c (init_tree_optimization_optabs): Use
15458 ggc_cleared_alloc to allocate target_optabs.
15459 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15460 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15462 * internal-fn.h (internal_load_fn_p): Declare.
15463 (internal_gather_scatter_fn_p): Likewise.
15464 (internal_fn_mask_index): Likewise.
15465 (internal_gather_scatter_fn_supported_p): Likewise.
15466 * internal-fn.c (gather_load_direct): New macro.
15467 (expand_gather_load_optab_fn): New function.
15468 (direct_gather_load_optab_supported_p): New macro.
15469 (direct_internal_fn_optab): New function.
15470 (internal_load_fn_p): Likewise.
15471 (internal_gather_scatter_fn_p): Likewise.
15472 (internal_fn_mask_index): Likewise.
15473 (internal_gather_scatter_fn_supported_p): Likewise.
15474 * optabs-query.c (supports_at_least_one_mode_p): New function.
15475 (supports_vec_gather_load_p): Likewise.
15476 * optabs-query.h (supports_vec_gather_load_p): Declare.
15477 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15478 and memory_type field.
15479 (NUM_PATTERNS): Bump to 15.
15480 * tree-vect-data-refs.c: Include internal-fn.h.
15481 (vect_gather_scatter_fn_p): New function.
15482 (vect_describe_gather_scatter_call): Likewise.
15483 (vect_check_gather_scatter): Try using internal functions for
15484 gather loads. Recognize existing calls to a gather load function.
15485 (vect_analyze_data_refs): Consider using gather loads if
15486 supports_vec_gather_load_p.
15487 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15488 (vect_get_gather_scatter_offset_type): Likewise.
15489 (vect_convert_mask_for_vectype): Likewise.
15490 (vect_add_conversion_to_patterm): Likewise.
15491 (vect_try_gather_scatter_pattern): Likewise.
15492 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15493 (vect_vect_recog_func_ptrs): Add it.
15494 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15495 internal_fn_mask_index and internal_gather_scatter_fn_p.
15496 (check_load_store_masking): Take the gather_scatter_info as an
15497 argument and handle gather loads.
15498 (vect_get_gather_scatter_ops): New function.
15499 (vectorizable_call): Check internal_load_fn_p.
15500 (vectorizable_load): Likewise. Handle gather load internal
15502 (vectorizable_store): Update call to check_load_store_masking.
15503 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15504 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15505 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15506 (aarch64_gather_scale_operand_d): New predicates.
15507 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15508 (mask_gather_load<mode>): New insns.
15510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15511 Alan Hayward <alan.hayward@arm.com>
15512 David Sherwood <david.sherwood@arm.com>
15514 * optabs.def (fold_left_plus_optab): New optab.
15515 * doc/md.texi (fold_left_plus_@var{m}): Document.
15516 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15517 * internal-fn.c (fold_left_direct): Define.
15518 (expand_fold_left_optab_fn): Likewise.
15519 (direct_fold_left_optab_supported_p): Likewise.
15520 * fold-const-call.c (fold_const_fold_left): New function.
15521 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15522 * tree-parloops.c (valid_reduction_p): New function.
15523 (gather_scalar_reductions): Use it.
15524 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15525 (vect_finish_replace_stmt): Declare.
15526 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15527 (needs_fold_left_reduction_p): New function, split out from...
15528 (vect_is_simple_reduction): ...here. Accept reductions that
15529 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15530 (vect_force_simple_reduction): Also store the reduction type in
15531 the assignment's STMT_VINFO_REDUC_TYPE.
15532 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15533 (merge_with_identity): New function.
15534 (vect_expand_fold_left): Likewise.
15535 (vectorize_fold_left_reduction): Likewise.
15536 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15537 scalar phi in place for it. Check for target support and reject
15538 cases that would reassociate the operation. Defer the transform
15539 phase to vectorize_fold_left_reduction.
15540 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15541 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15542 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15546 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15547 call to ifc_temp_var.
15549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15550 Alan Hayward <alan.hayward@arm.com>
15551 David Sherwood <david.sherwood@arm.com>
15553 * target.def (legitimize_address_displacement): Take the original
15554 offset as a poly_int.
15555 * targhooks.h (default_legitimize_address_displacement): Update
15557 * targhooks.c (default_legitimize_address_displacement): Likewise.
15558 * doc/tm.texi: Regenerate.
15559 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15560 as an argument, moving assert of ad->disp == ad->disp_term to...
15561 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15562 Try calling targetm.legitimize_address_displacement before expanding
15563 the address rather than afterwards, and adjust for the new interface.
15564 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15565 Match the new hook interface. Handle SVE addresses.
15566 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15567 new hook interface.
15569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15571 * Makefile.in (OBJS): Add early-remat.o.
15572 * target.def (select_early_remat_modes): New hook.
15573 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15574 * doc/tm.texi: Regenerate.
15575 * targhooks.h (default_select_early_remat_modes): Declare.
15576 * targhooks.c (default_select_early_remat_modes): New function.
15577 * timevar.def (TV_EARLY_REMAT): New timevar.
15578 * passes.def (pass_early_remat): New pass.
15579 * tree-pass.h (make_pass_early_remat): Declare.
15580 * early-remat.c: New file.
15581 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15583 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15586 Alan Hayward <alan.hayward@arm.com>
15587 David Sherwood <david.sherwood@arm.com>
15589 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15590 vfm1 with a bound_epilog parameter.
15591 (vect_do_peeling): Update calls accordingly, and move the prologue
15592 call earlier in the function. Treat the base bound_epilog as 0 for
15593 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15594 this base when peeling for gaps.
15595 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15596 with fully-masked loops.
15597 (vect_estimate_min_profitable_iters): Handle the single peeled
15598 iteration in that case.
15600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15601 Alan Hayward <alan.hayward@arm.com>
15602 David Sherwood <david.sherwood@arm.com>
15604 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15605 single-element interleaving even if the size is not a power of 2.
15606 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15607 accesses for single-element interleaving if the group size is
15610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15611 Alan Hayward <alan.hayward@arm.com>
15612 David Sherwood <david.sherwood@arm.com>
15614 * doc/md.texi (fold_extract_last_@var{m}): Document.
15615 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15616 * optabs.def (fold_extract_last_optab): New optab.
15617 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15618 * internal-fn.c (fold_extract_direct): New macro.
15619 (expand_fold_extract_optab_fn): Likewise.
15620 (direct_fold_extract_optab_supported_p): Likewise.
15621 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15622 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15623 EXTRACT_LAST_REDUCTION.
15624 (get_initial_def_for_reduction): Do not create an initial vector
15625 for EXTRACT_LAST_REDUCTION reductions.
15626 (vectorizable_reduction): Leave the scalar phi in place for
15627 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15628 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15629 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15630 transform phase to vectorizable_condition.
15631 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15633 (vect_finish_stmt_generation): ...here.
15634 (vect_finish_replace_stmt): New function.
15635 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15636 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15638 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15641 Alan Hayward <alan.hayward@arm.com>
15642 David Sherwood <david.sherwood@arm.com>
15644 * doc/md.texi (extract_last_@var{m}): Document.
15645 * optabs.def (extract_last_optab): New optab.
15646 * internal-fn.def (EXTRACT_LAST): New internal function.
15647 * internal-fn.c (cond_unary_direct): New macro.
15648 (expand_cond_unary_optab_fn): Likewise.
15649 (direct_cond_unary_optab_supported_p): Likewise.
15650 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15651 loops using EXTRACT_LAST.
15652 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15653 (extract_last_<mode>): ...this optab.
15654 (vec_extract<mode><Vel>): Update accordingly.
15656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15657 Alan Hayward <alan.hayward@arm.com>
15658 David Sherwood <david.sherwood@arm.com>
15660 * target.def (empty_mask_is_expensive): New hook.
15661 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15662 * doc/tm.texi: Regenerate.
15663 * targhooks.h (default_empty_mask_is_expensive): Declare.
15664 * targhooks.c (default_empty_mask_is_expensive): New function.
15665 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15666 if the target says that empty masks are expensive.
15667 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15669 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15672 Alan Hayward <alan.hayward@arm.com>
15673 David Sherwood <david.sherwood@arm.com>
15675 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15676 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15677 (vect_use_loop_mask_for_alignment_p): New function.
15678 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15679 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15680 niters_skip argument. Make sure that the first niters_skip elements
15681 of the first iteration are inactive.
15682 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15683 Update call to vect_set_loop_masks_directly.
15684 (get_misalign_in_elems): New function, split out from...
15685 (vect_gen_prolog_loop_niters): ...here.
15686 (vect_update_init_of_dr): Take a code argument that specifies whether
15687 the adjustment should be added or subtracted.
15688 (vect_update_init_of_drs): Likewise.
15689 (vect_prepare_for_masked_peels): New function.
15690 (vect_do_peeling): Skip prologue peeling if we're using a mask
15691 instead. Update call to vect_update_inits_of_drs.
15692 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15694 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15695 alignment. Do not include the number of peeled iterations in
15696 the minimum threshold in that case.
15697 (vectorizable_induction): Adjust the start value down by
15698 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15699 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15700 Take the number of skipped iterations into account when calculating
15702 * tree-vect-stmts.c (vect_gen_while_not): New function.
15704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15705 Alan Hayward <alan.hayward@arm.com>
15706 David Sherwood <david.sherwood@arm.com>
15708 * doc/sourcebuild.texi (vect_fully_masked): Document.
15709 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15710 default value to 0.
15711 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15713 (vect_analyze_loop_2): ...here. Don't check the vectorization
15714 factor against the number of loop iterations if the loop is
15717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15718 Alan Hayward <alan.hayward@arm.com>
15719 David Sherwood <david.sherwood@arm.com>
15721 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15722 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15723 (dump_groups): Update accordingly.
15724 (iv_use::mem_type): New member variable.
15725 (address_p): New function.
15726 (record_use): Add a mem_type argument and initialize the new
15728 (record_group_use): Add a mem_type argument. Use address_p.
15729 Remove obsolete null checks of base_object. Update call to record_use.
15730 (find_interesting_uses_op): Update call to record_group_use.
15731 (find_interesting_uses_cond): Likewise.
15732 (find_interesting_uses_address): Likewise.
15733 (get_mem_type_for_internal_fn): New function.
15734 (find_address_like_use): Likewise.
15735 (find_interesting_uses_stmt): Try find_address_like_use before
15736 calling find_interesting_uses_op.
15737 (addr_offset_valid_p): Use the iv mem_type field as the type
15738 of the addressed memory.
15739 (add_autoinc_candidates): Likewise.
15740 (get_address_cost): Likewise.
15741 (split_small_address_groups_p): Use address_p.
15742 (split_address_groups): Likewise.
15743 (add_iv_candidate_for_use): Likewise.
15744 (autoinc_possible_for_pair): Likewise.
15745 (rewrite_groups): Likewise.
15746 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15747 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15748 (get_alias_ptr_type_for_ptr_address): New function.
15749 (rewrite_use_address): Rewrite address uses in calls that were
15750 identified by find_address_like_use.
15752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15753 Alan Hayward <alan.hayward@arm.com>
15754 David Sherwood <david.sherwood@arm.com>
15756 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15758 * gimple-expr.h (is_gimple_addressable: Likewise.
15759 * gimple-expr.c (is_gimple_address): Likewise.
15760 * internal-fn.c (expand_call_mem_ref): New function.
15761 (expand_mask_load_optab_fn): Use it.
15762 (expand_mask_store_optab_fn): Likewise.
15764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15765 Alan Hayward <alan.hayward@arm.com>
15766 David Sherwood <david.sherwood@arm.com>
15768 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15769 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15770 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15771 (cond_umax@var{mode}): Document.
15772 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15773 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15774 (cond_umin_optab, cond_umax_optab): New optabs.
15775 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15776 (COND_IOR, COND_XOR): New internal functions.
15777 * internal-fn.h (get_conditional_internal_fn): Declare.
15778 * internal-fn.c (cond_binary_direct): New macro.
15779 (expand_cond_binary_optab_fn): Likewise.
15780 (direct_cond_binary_optab_supported_p): Likewise.
15781 (get_conditional_internal_fn): New function.
15782 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15783 Cope with reduction statements that are vectorized as calls rather
15785 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15786 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15787 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15788 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15789 (UNSPEC_COND_EOR): New unspecs.
15790 (optab): Add mappings for them.
15791 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15792 (sve_int_op, sve_fp_op): New int attributes.
15794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15795 Alan Hayward <alan.hayward@arm.com>
15796 David Sherwood <david.sherwood@arm.com>
15798 * optabs.def (while_ult_optab): New optab.
15799 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15800 * internal-fn.def (WHILE_ULT): New internal function.
15801 * internal-fn.h (direct_internal_fn_supported_p): New override
15802 that takes two types as argument.
15803 * internal-fn.c (while_direct): New macro.
15804 (expand_while_optab_fn): New function.
15805 (convert_optab_supported_p): Likewise.
15806 (direct_while_optab_supported_p): New macro.
15807 * wide-int.h (wi::udiv_ceil): New function.
15808 * tree-vectorizer.h (rgroup_masks): New structure.
15809 (vec_loop_masks): New typedef.
15810 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15811 and fully_masked_p.
15812 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15813 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15814 (vect_max_vf): New function.
15815 (slpeel_make_loop_iterate_ntimes): Delete.
15816 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15817 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15818 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15819 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15820 internal-fn.h, stor-layout.h and optabs-query.h.
15821 (vect_set_loop_mask): New function.
15822 (add_preheader_seq): Likewise.
15823 (add_header_seq): Likewise.
15824 (interleave_supported_p): Likewise.
15825 (vect_maybe_permute_loop_masks): Likewise.
15826 (vect_set_loop_masks_directly): Likewise.
15827 (vect_set_loop_condition_masked): Likewise.
15828 (vect_set_loop_condition_unmasked): New function, split out from
15829 slpeel_make_loop_iterate_ntimes.
15830 (slpeel_make_loop_iterate_ntimes): Rename to..
15831 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15832 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15833 (vect_do_peeling): Update call accordingly.
15834 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15836 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15837 mask_compare_type, can_fully_mask_p and fully_masked_p.
15838 (release_vec_loop_masks): New function.
15839 (_loop_vec_info): Use it to free the loop masks.
15840 (can_produce_all_loop_masks_p): New function.
15841 (vect_get_max_nscalars_per_iter): Likewise.
15842 (vect_verify_full_masking): Likewise.
15843 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15844 retries, and free the mask rgroups before retrying. Check loop-wide
15845 reasons for disallowing fully-masked loops. Make the final decision
15846 about whether use a fully-masked loop or not.
15847 (vect_estimate_min_profitable_iters): Do not assume that peeling
15848 for the number of iterations will be needed for fully-masked loops.
15849 (vectorizable_reduction): Disable fully-masked loops.
15850 (vectorizable_live_operation): Likewise.
15851 (vect_halve_mask_nunits): New function.
15852 (vect_double_mask_nunits): Likewise.
15853 (vect_record_loop_mask): Likewise.
15854 (vect_get_loop_mask): Likewise.
15855 (vect_transform_loop): Handle the case in which the final loop
15856 iteration might handle a partial vector. Call vect_set_loop_condition
15857 instead of slpeel_make_loop_iterate_ntimes.
15858 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15859 (check_load_store_masking): New function.
15860 (prepare_load_store_mask): Likewise.
15861 (vectorizable_store): Handle fully-masked loops.
15862 (vectorizable_load): Likewise.
15863 (supportable_widening_operation): Use vect_halve_mask_nunits for
15865 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15866 (vect_gen_while): New function.
15867 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15868 (aarch64_uqdec<mode>): New insn.
15870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15871 Alan Hayward <alan.hayward@arm.com>
15872 David Sherwood <david.sherwood@arm.com>
15874 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15875 (reduc_xor_scal_optab): New optabs.
15876 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15877 (reduc_xor_scal_@var{m}): Document.
15878 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15879 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15880 internal functions.
15881 * fold-const-call.c (fold_const_call): Handle them.
15882 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15883 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15884 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15885 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15886 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15887 (UNSPEC_XORV): New unspecs.
15888 (optab): Add entries for them.
15889 (BITWISEV): New int iterator.
15890 (bit_reduc_op): New int attributes.
15892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15893 Alan Hayward <alan.hayward@arm.com>
15894 David Sherwood <david.sherwood@arm.com>
15896 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15897 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15898 * optabs.def (vec_shl_insert_optab): New optab.
15899 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15900 (duplicate_and_interleave): Likewise.
15901 * tree-vect-loop.c: Include internal-fn.h.
15902 (neutral_op_for_slp_reduction): New function, split out from
15903 get_initial_defs_for_reduction.
15904 (get_initial_def_for_reduction): Handle option 2 for variable-length
15905 vectors by loading the neutral value into a vector and then shifting
15906 the initial value into element 0.
15907 (get_initial_defs_for_reduction): Replace the code argument with
15908 the neutral value calculated by neutral_op_for_slp_reduction.
15909 Use gimple_build_vector for constant-length vectors.
15910 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15911 but the first group_size elements have a neutral value.
15912 Use duplicate_and_interleave otherwise.
15913 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15914 Update call to get_initial_defs_for_reduction. Handle SLP
15915 reductions for variable-length vectors by creating one vector
15916 result for each scalar result, with the elements associated
15917 with other scalar results stubbed out with the neutral value.
15918 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15919 Require IFN_VEC_SHL_INSERT for double reductions on
15920 variable-length vectors, or SLP reductions that have
15921 a neutral value. Require can_duplicate_and_interleave_p
15922 support for variable-length unchained SLP reductions if there
15923 is no neutral value, such as for MIN/MAX reductions. Also require
15924 the number of vector elements to be a multiple of the number of
15925 SLP statements when doing variable-length unchained SLP reductions.
15926 Update call to vect_create_epilog_for_reduction.
15927 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15928 and remove initial values.
15929 (duplicate_and_interleave): Make public.
15930 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15931 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15934 Alan Hayward <alan.hayward@arm.com>
15935 David Sherwood <david.sherwood@arm.com>
15937 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15938 (can_duplicate_and_interleave_p): New function.
15939 (vect_get_and_check_slp_defs): Take the vector of statements
15940 rather than just the current one. Remove excess parentheses.
15941 Restriction rejectinon of vect_constant_def and vect_external_def
15942 for variable-length vectors to boolean types, or types for which
15943 can_duplicate_and_interleave_p is false.
15944 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15945 (duplicate_and_interleave): New function.
15946 (vect_get_constant_vectors): Use gimple_build_vector for
15947 constant-length vectors and suitable variable-length constant
15948 vectors. Use duplicate_and_interleave for other variable-length
15949 vectors. Don't defer the update when inserting new statements.
15951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15952 Alan Hayward <alan.hayward@arm.com>
15953 David Sherwood <david.sherwood@arm.com>
15955 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15956 min_profitable_iters doesn't go negative.
15958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15959 Alan Hayward <alan.hayward@arm.com>
15960 David Sherwood <david.sherwood@arm.com>
15962 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15963 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15964 * optabs.def (vec_mask_load_lanes_optab): New optab.
15965 (vec_mask_store_lanes_optab): Likewise.
15966 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15967 (MASK_STORE_LANES): Likewise.
15968 * internal-fn.c (mask_load_lanes_direct): New macro.
15969 (mask_store_lanes_direct): Likewise.
15970 (expand_mask_load_optab_fn): Handle masked operations.
15971 (expand_mask_load_lanes_optab_fn): New macro.
15972 (expand_mask_store_optab_fn): Handle masked operations.
15973 (expand_mask_store_lanes_optab_fn): New macro.
15974 (direct_mask_load_lanes_optab_supported_p): Likewise.
15975 (direct_mask_store_lanes_optab_supported_p): Likewise.
15976 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15978 (vect_load_lanes_supported): Likewise.
15979 * tree-vect-data-refs.c (strip_conversion): New function.
15980 (can_group_stmts_p): Likewise.
15981 (vect_analyze_data_ref_accesses): Use it instead of checking
15982 for a pair of assignments.
15983 (vect_store_lanes_supported): Take a masked_p parameter.
15984 (vect_load_lanes_supported): Likewise.
15985 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15986 vect_store_lanes_supported and vect_load_lanes_supported.
15987 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15988 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15989 parameter. Don't allow gaps for masked accesses.
15990 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15991 and vect_load_lanes_supported.
15992 (get_load_store_type): Take a masked_p parameter and update
15993 call to get_group_load_store_type.
15994 (vectorizable_store): Update call to get_load_store_type.
15995 Handle IFN_MASK_STORE_LANES.
15996 (vectorizable_load): Update call to get_load_store_type.
15997 Handle IFN_MASK_LOAD_LANES.
15999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16000 Alan Hayward <alan.hayward@arm.com>
16001 David Sherwood <david.sherwood@arm.com>
16003 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16005 * config/aarch64/aarch64-protos.h
16006 (aarch64_sve_struct_memory_operand_p): Declare.
16007 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16008 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16009 (VPRED, vpred): Handle SVE structure modes.
16010 * config/aarch64/constraints.md (Utx): New constraint.
16011 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16012 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16013 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16014 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16015 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16016 structure modes. Split into pieces after RA.
16017 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16018 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16020 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16021 SVE structure modes.
16022 (aarch64_classify_address): Likewise.
16023 (sizetochar): Move earlier in file.
16024 (aarch64_print_operand): Handle SVE register lists.
16025 (aarch64_array_mode): New function.
16026 (aarch64_sve_struct_memory_operand_p): Likewise.
16027 (TARGET_ARRAY_MODE): Redefine.
16029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16030 Alan Hayward <alan.hayward@arm.com>
16031 David Sherwood <david.sherwood@arm.com>
16033 * target.def (array_mode): New target hook.
16034 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16035 * doc/tm.texi: Regenerate.
16036 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16037 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16038 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16039 targetm.array_mode.
16040 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16044 Alan Hayward <alan.hayward@arm.com>
16045 David Sherwood <david.sherwood@arm.com>
16047 * fold-const.c (fold_binary_loc): Check the argument types
16048 rather than the result type when testing for a vector operation.
16050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16052 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16053 * doc/tm.texi: Regenerate.
16055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16056 Alan Hayward <alan.hayward@arm.com>
16057 David Sherwood <david.sherwood@arm.com>
16059 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16060 (sve): Document new AArch64 extension.
16061 * doc/md.texi (w): Extend the description of the AArch64
16062 constraint to include SVE vectors.
16063 (Upl, Upa): Document new AArch64 predicate constraints.
16064 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16066 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16067 (msve-vector-bits=): New option.
16068 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16069 SVE when these are disabled.
16070 (sve): New extension.
16071 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16072 modes. Adjust their number of units based on aarch64_sve_vg.
16073 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16074 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16075 aarch64_addr_query_type.
16076 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16077 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16078 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16079 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16080 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16081 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16082 (aarch64_simd_imm_zero_p): Delete.
16083 (aarch64_check_zero_based_sve_index_immediate): Declare.
16084 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16085 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16086 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16087 (aarch64_sve_float_mul_immediate_p): Likewise.
16088 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16089 rather than an rtx.
16090 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16091 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16092 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16093 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16094 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16095 (aarch64_regmode_natural_size): Likewise.
16096 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16097 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16099 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16100 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16101 for VG and the SVE predicate registers.
16102 (V_ALIASES): Add a "z"-prefixed alias.
16103 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16104 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16105 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16106 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16107 (REG_CLASS_NAMES): Add entries for them.
16108 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16109 and the predicate registers.
16110 (aarch64_sve_vg): Declare.
16111 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16112 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16113 (REGMODE_NATURAL_SIZE): Define.
16114 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16116 * config/aarch64/aarch64.c: Include cfgrtl.h.
16117 (simd_immediate_info): Add a constructor for series vectors,
16118 and an associated step field.
16119 (aarch64_sve_vg): New variable.
16120 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16121 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16122 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16123 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16124 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16125 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16126 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16127 (aarch64_get_mask_mode): New functions.
16128 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16129 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16130 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16131 predicate modes and predicate registers. Explicitly restrict
16132 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16133 to store a vector mode if it is recognized by
16134 aarch64_classify_vector_mode.
16135 (aarch64_regmode_natural_size): New function.
16136 (aarch64_hard_regno_caller_save_mode): Return the original mode
16138 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16139 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16140 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16141 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16143 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16144 does not overlap dest if the function is frame-related. Handle
16146 (aarch64_split_add_offset): New function.
16147 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16148 them aarch64_add_offset.
16149 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16150 and update call to aarch64_sub_sp.
16151 (aarch64_add_cfa_expression): New function.
16152 (aarch64_expand_prologue): Pass extra temporary registers to the
16153 functions above. Handle the case in which we need to emit new
16154 DW_CFA_expressions for registers that were originally saved
16155 relative to the stack pointer, but now have to be expressed
16156 relative to the frame pointer.
16157 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16159 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16160 IP0 and IP1 values for SVE frames.
16161 (aarch64_expand_vec_series): New function.
16162 (aarch64_expand_sve_widened_duplicate): Likewise.
16163 (aarch64_expand_sve_const_vector): Likewise.
16164 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16165 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16166 into the register, rather than emitting a SET directly.
16167 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16168 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16169 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16170 (offset_9bit_signed_scaled_p): New functions.
16171 (aarch64_replicate_bitmask_imm): New function.
16172 (aarch64_bitmask_imm): Use it.
16173 (aarch64_cannot_force_const_mem): Reject expressions involving
16174 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16175 (aarch64_classify_index): Handle SVE indices, by requiring
16176 a plain register index with a scale that matches the element size.
16177 (aarch64_classify_address): Handle SVE addresses. Assert that
16178 the mode of the address is VOIDmode or an integer mode.
16179 Update call to aarch64_classify_symbol.
16180 (aarch64_classify_symbolic_expression): Update call to
16181 aarch64_classify_symbol.
16182 (aarch64_const_vec_all_in_range_p): New function.
16183 (aarch64_print_vector_float_operand): Likewise.
16184 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16185 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16186 and the FP immediates 1.0 and 0.5.
16187 (aarch64_print_address_internal): Handle SVE addresses.
16188 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16189 (aarch64_regno_regclass): Handle predicate registers.
16190 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16192 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16193 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16194 (aarch64_convert_sve_vector_bits): New function.
16195 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16196 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16197 rather than an rtx.
16198 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16199 Handle SVE vector and predicate modes. Accept VL-based constants
16200 that need only one temporary register, and VL offsets that require
16201 no temporary registers.
16202 (aarch64_conditional_register_usage): Mark the predicate registers
16203 as fixed if SVE isn't available.
16204 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16205 Return true for SVE vector and predicate modes.
16206 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16207 rather than an unsigned int. Handle SVE modes.
16208 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16210 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16212 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16213 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16214 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16215 (aarch64_sve_float_mul_immediate_p): New functions.
16216 (aarch64_sve_valid_immediate): New function.
16217 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16218 Explicitly reject structure modes. Check for INDEX constants.
16219 Handle PTRUE and PFALSE constants.
16220 (aarch64_check_zero_based_sve_index_immediate): New function.
16221 (aarch64_simd_imm_zero_p): Delete.
16222 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16223 vector modes. Accept constants in the range of CNT[BHWD].
16224 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16225 ask for an Advanced SIMD mode.
16226 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16227 (aarch64_simd_vector_alignment): Handle SVE predicates.
16228 (aarch64_vectorize_preferred_vector_alignment): New function.
16229 (aarch64_simd_vector_alignment_reachable): Use it instead of
16231 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16232 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16234 (MAX_VECT_LEN): Delete.
16235 (expand_vec_perm_d): Add a vec_flags field.
16236 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16237 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16238 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16240 (aarch64_evpc_rev): Rename to...
16241 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16242 (aarch64_evpc_rev_global): New function.
16243 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16244 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16246 (aarch64_evpc_sve_tbl): New function.
16247 (aarch64_expand_vec_perm_const_1): Update after rename of
16248 aarch64_evpc_rev. Handle SVE permutes too, trying
16249 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16250 than aarch64_evpc_tbl.
16251 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16252 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16253 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16254 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16255 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16256 (aarch64_expand_sve_vcond): New functions.
16257 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16258 of aarch64_vector_mode_p.
16259 (aarch64_dwarf_poly_indeterminate_value): New function.
16260 (aarch64_compute_pressure_classes): Likewise.
16261 (aarch64_can_change_mode_class): Likewise.
16262 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16263 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16264 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16265 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16266 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16267 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16268 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16269 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16271 (Dn, Dl, Dr): Accept const as well as const_vector.
16272 (Dz): Likewise. Compare against CONST0_RTX.
16273 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16274 of "vector" where appropriate.
16275 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16276 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16277 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16278 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16279 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16280 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16281 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16282 (v_int_equiv): Extend to SVE modes.
16283 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16285 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16286 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16287 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16288 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16289 (SVE_COND_FP_CMP): New int iterators.
16290 (perm_hilo): Handle the new unpack unspecs.
16291 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16293 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16294 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16295 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16296 (aarch64_equality_operator, aarch64_constant_vector_operand)
16297 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16298 (aarch64_sve_nonimmediate_operand): Likewise.
16299 (aarch64_sve_general_operand): Likewise.
16300 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16301 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16302 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16303 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16304 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16305 (aarch64_sve_float_arith_immediate): Likewise.
16306 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16307 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16308 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16309 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16310 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16311 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16312 (aarch64_sve_float_arith_operand): Likewise.
16313 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16314 (aarch64_sve_float_mul_operand): Likewise.
16315 (aarch64_sve_vec_perm_operand): Likewise.
16316 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16317 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16318 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16319 as well as const_vector.
16320 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16321 in file. Use CONST0_RTX and CONSTM1_RTX.
16322 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16323 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16324 Use aarch64_simd_imm_zero.
16325 * config/aarch64/aarch64-sve.md: New file.
16326 * config/aarch64/aarch64.md: Include it.
16327 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16328 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16329 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16330 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16331 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16332 (sve): New attribute.
16333 (enabled): Disable instructions with the sve attribute unless
16335 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16336 aarch64_expand_mov_immediate.
16337 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16338 CNT[BHSD] immediates.
16339 (movti): Split CONST_POLY_INT moves into two halves.
16340 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16341 Split additions that need a temporary here if the destination
16342 is the stack pointer.
16343 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16344 (*add<mode>3_poly_1): New instruction.
16345 (set_clobber_cc): New expander.
16347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16349 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16350 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16351 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16352 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16353 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16354 Change innermode from fixed_mode_size to machine_mode.
16355 (simplify_subreg): Update call accordingly. Handle a constant-sized
16356 subreg of a variable-length CONST_VECTOR.
16358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16359 Alan Hayward <alan.hayward@arm.com>
16360 David Sherwood <david.sherwood@arm.com>
16362 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16363 (add_offset_to_base): New function, split out from...
16364 (create_mem_ref): ...here. When handling a scale other than 1,
16365 check first whether the address is valid without the offset.
16366 Add it into the base if so, leaving the index and scale as-is.
16368 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16371 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16372 fold_for_warn before checking if arg2 is INTEGER_CST.
16374 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16376 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16377 (store_multiple_operation): Delete.
16378 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16379 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16380 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16381 guarded by TARGET_STRING.
16382 (rs6000_output_load_multiple): Delete.
16383 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16384 OPTION_MASK_STRING / TARGET_STRING handling.
16385 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16386 (const rs6000_opt_masks) <"string">: Change mask to 0.
16387 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16388 (MASK_STRING): Delete.
16389 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16391 (load_multiple): Delete.
16398 (store_multiple): Delete.
16405 (movmemsi_8reg): Delete.
16406 (corresponding unnamed define_insn): Delete.
16407 (movmemsi_6reg): Delete.
16408 (corresponding unnamed define_insn): Delete.
16409 (movmemsi_4reg): Delete.
16410 (corresponding unnamed define_insn): Delete.
16411 (movmemsi_2reg): Delete.
16412 (corresponding unnamed define_insn): Delete.
16413 (movmemsi_1reg): Delete.
16414 (corresponding unnamed define_insn): Delete.
16415 * config/rs6000/rs6000.opt (mno-string): New.
16416 (mstring): Replace by deprecation warning stub.
16417 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16419 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16421 * regrename.c (regrename_do_replace): If replacing the same
16422 reg multiple times, try to reuse last created gen_raw_REG.
16425 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16426 main to workaround a bug in GDB.
16428 2018-01-12 Tom de Vries <tom@codesourcery.com>
16431 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16433 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16435 PR rtl-optimization/80481
16436 * ira-color.c (get_cap_member): New function.
16437 (allocnos_conflict_by_live_ranges_p): Use it.
16438 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16439 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16441 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16444 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16445 (*saddl_se_1): Ditto.
16446 (*ssubsi_1): Ditto.
16447 (*ssubl_se_1): Ditto.
16449 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16451 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16452 rather than wi::to_widest for DR_INITs.
16453 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16454 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16455 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16457 (vect_analyze_group_access_1): Note that here.
16459 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16461 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16462 polynomial type sizes.
16464 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16466 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16467 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16468 (gimple_add_tmp_var): Likewise.
16470 2018-01-12 Martin Liska <mliska@suse.cz>
16472 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16473 (gimple_alloc_sizes): Likewise.
16474 (dump_gimple_statistics): Use PRIu64 in printf format.
16475 * gimple.h: Change uint64_t to int.
16477 2018-01-12 Martin Liska <mliska@suse.cz>
16479 * tree-core.h: Use uint64_t instead of int.
16480 * tree.c (tree_node_counts): Likewise.
16481 (tree_node_sizes): Likewise.
16482 (dump_tree_statistics): Use PRIu64 in printf format.
16484 2018-01-12 Martin Liska <mliska@suse.cz>
16486 * Makefile.in: As qsort_chk is implemented in vec.c, add
16487 vec.o to linkage of gencfn-macros.
16488 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16489 passing the info to record_node_allocation_statistics.
16490 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16492 * ggc-common.c (struct ggc_usage): Add operator== and use
16493 it in operator< and compare function.
16494 * mem-stats.h (struct mem_usage): Likewise.
16495 * vec.c (struct vec_usage): Remove operator< and compare
16496 function. Can be simply inherited.
16498 2018-01-12 Martin Jambor <mjambor@suse.cz>
16501 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16502 * tree-ssa-math-opts.c: Include domwalk.h.
16503 (convert_mult_to_fma_1): New function.
16504 (fma_transformation_info): New type.
16505 (fma_deferring_state): Likewise.
16506 (cancel_fma_deferring): New function.
16507 (result_of_phi): Likewise.
16508 (last_fma_candidate_feeds_initial_phi): Likewise.
16509 (convert_mult_to_fma): Added deferring logic, split actual
16510 transformation to convert_mult_to_fma_1.
16511 (math_opts_dom_walker): New type.
16512 (math_opts_dom_walker::after_dom_children): New method, body moved
16513 here from pass_optimize_widening_mul::execute, added deferring logic
16515 (pass_optimize_widening_mul::execute): Moved most of code to
16516 math_opts_dom_walker::after_dom_children.
16517 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16518 * config/i386/i386.c (ix86_option_override_internal): Added
16519 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16521 2018-01-12 Richard Biener <rguenther@suse.de>
16524 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16525 inline instance vars.
16527 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16530 * config/rx/rx.c (rx_is_restricted_memory_address):
16531 Handle SUBREG case.
16533 2018-01-12 Richard Biener <rguenther@suse.de>
16535 PR tree-optimization/80846
16536 * target.def (split_reduction): New target hook.
16537 * targhooks.c (default_split_reduction): New function.
16538 * targhooks.h (default_split_reduction): Declare.
16539 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16540 target requests first reduce vectors by combining low and high
16542 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16543 (get_vectype_for_scalar_type_and_size): Export.
16544 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16545 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16546 * doc/tm.texi: Regenerate.
16547 * config/i386/i386.c (ix86_split_reduction): Implement
16548 TARGET_VECTORIZE_SPLIT_REDUCTION.
16550 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16553 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16554 in PIC mode except for TARGET_VXWORKS_RTP.
16555 * config/sparc/sparc.c: Include cfgrtl.h.
16556 (TARGET_INIT_PIC_REG): Define.
16557 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16558 (sparc_pic_register_p): New predicate.
16559 (sparc_legitimate_address_p): Use it.
16560 (sparc_legitimize_pic_address): Likewise.
16561 (sparc_delegitimize_address): Likewise.
16562 (sparc_mode_dependent_address_p): Likewise.
16563 (gen_load_pcrel_sym): Remove 4th parameter.
16564 (load_got_register): Adjust call to above. Remove obsolete stuff.
16565 (sparc_expand_prologue): Do not call load_got_register here.
16566 (sparc_flat_expand_prologue): Likewise.
16567 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16568 (sparc_use_pseudo_pic_reg): New function.
16569 (sparc_init_pic_reg): Likewise.
16570 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16571 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16573 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16575 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16576 Add item for branch_cost.
16578 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16580 PR rtl-optimization/83565
16581 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16582 not extend the result to a larger mode for rotate operations.
16583 (num_sign_bit_copies1): Likewise.
16585 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16588 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16590 Use values-Xc.o for -pedantic.
16591 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16593 2018-01-12 Martin Liska <mliska@suse.cz>
16596 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16598 (possible_polymorphic_call_targets): Use it.
16599 (ipa_devirt): Likewise.
16601 2018-01-12 Martin Liska <mliska@suse.cz>
16603 * profile-count.h (enum profile_quality): Use 0 as invalid
16604 enum value of profile_quality.
16606 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16608 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16609 -mext-string options.
16611 2018-01-12 Richard Biener <rguenther@suse.de>
16613 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16614 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16615 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16617 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16619 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16621 * configure.ac (--with-long-double-format): Add support for the
16622 configuration option to change the default long double format on
16624 * config.gcc (powerpc*-linux*-*): Likewise.
16625 * configure: Regenerate.
16626 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16627 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16628 used without modification.
16630 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16632 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16633 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16634 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16635 MISC_BUILTIN_SPEC_BARRIER.
16636 (rs6000_init_builtins): Likewise.
16637 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16639 (speculation_barrier): New define_insn.
16640 * doc/extend.texi: Document __builtin_speculation_barrier.
16642 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16645 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16646 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16647 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16649 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16650 integral modes instead of "ss" and "sd".
16651 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16652 vectors with 32-bit and 64-bit elements.
16653 (vecdupssescalarmodesuffix): New mode attribute.
16654 (vec_dup<mode>): Use it.
16656 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16659 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16660 frame if argument is passed on stack.
16662 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16665 * ree.c (combine_reaching_defs): Optimize also
16666 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16667 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16669 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16671 PR middle-end/83189
16672 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16674 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16676 PR middle-end/83718
16677 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16678 after they are computed.
16680 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16682 PR tree-optimization/83695
16683 * gimple-loop-linterchange.cc
16684 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16685 reset cached scev information after interchange.
16686 (pass_linterchange::execute): Remove call to scev_reset_htab.
16688 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16690 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16691 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16692 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16693 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16694 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16695 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16696 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16697 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16698 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16699 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16700 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16701 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16702 (V_lane_reg): Likewise.
16703 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16705 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16706 (vfmal_lane_low<mode>_intrinsic,
16707 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16708 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16709 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16710 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16711 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16712 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16714 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16716 * config/arm/arm-cpus.in (fp16fml): New feature.
16717 (ALL_SIMD): Add fp16fml.
16718 (armv8.2-a): Add fp16fml as an option.
16719 (armv8.3-a): Likewise.
16720 (armv8.4-a): Add fp16fml as part of fp16.
16721 * config/arm/arm.h (TARGET_FP16FML): Define.
16722 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16724 * config/arm/arm-modes.def (V2HF): Define.
16725 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16726 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16727 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16728 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16729 vfmsl_low, vfmsl_high): New set of builtins.
16730 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16731 (vfml_op): New code attribute.
16732 (VFMLHALVES): New int iterator.
16733 (VFML, VFMLSEL): New mode attributes.
16734 (V_reg): Define mapping for V2HF.
16735 (V_hi, V_lo): New mode attributes.
16736 (VF_constraint): Likewise.
16737 (vfml_half, vfml_half_selector): New int attributes.
16738 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16740 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16741 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16743 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16744 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16745 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16746 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16748 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16749 Document new effective target and option set.
16751 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16753 * config/arm/arm-cpus.in (armv8_4): New feature.
16754 (ARMv8_4a): New fgroup.
16755 (armv8.4-a): New arch.
16756 * config/arm/arm-tables.opt: Regenerate.
16757 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16758 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16759 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16760 Add matching rules for -march=armv8.4-a and extensions.
16761 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16763 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16766 * config/rx/rx.md (BW): New mode attribute.
16767 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16769 2018-01-11 Richard Biener <rguenther@suse.de>
16771 PR tree-optimization/83435
16772 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16773 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16774 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16776 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16777 Alan Hayward <alan.hayward@arm.com>
16778 David Sherwood <david.sherwood@arm.com>
16780 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16782 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16783 (aarch64_print_address_internal): Use it to check for a zero offset.
16785 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16786 Alan Hayward <alan.hayward@arm.com>
16787 David Sherwood <david.sherwood@arm.com>
16789 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16790 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16791 Return a poly_int64 rather than a HOST_WIDE_INT.
16792 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16793 rather than a HOST_WIDE_INT.
16794 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16795 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16796 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16797 final_offset from HOST_WIDE_INT to poly_int64.
16798 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16799 to_constant when getting the number of units in an Advanced SIMD
16801 (aarch64_builtin_vectorized_function): Check for a constant number
16803 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16805 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16806 attribute instead of GET_MODE_NUNITS.
16807 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16808 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16809 GET_MODE_SIZE for fixed-size registers.
16810 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16811 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16812 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16813 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16814 (aarch64_print_operand, aarch64_print_address_internal)
16815 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16816 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16817 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16818 Handle polynomial GET_MODE_SIZE.
16819 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16820 wider than SImode without modification.
16821 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16822 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16823 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16824 passing and returning SVE modes.
16825 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16826 rather than GEN_INT.
16827 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16828 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16829 (aarch64_allocate_and_probe_stack_space): Likewise.
16830 (aarch64_layout_frame): Cope with polynomial offsets.
16831 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16832 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16833 polynomial offsets.
16834 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16835 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16836 poly_int64 rather than a HOST_WIDE_INT.
16837 (aarch64_get_separate_components, aarch64_process_components)
16838 (aarch64_expand_prologue, aarch64_expand_epilogue)
16839 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16840 (aarch64_anchor_offset): New function, split out from...
16841 (aarch64_legitimize_address): ...here.
16842 (aarch64_builtin_vectorization_cost): Handle polynomial
16843 TYPE_VECTOR_SUBPARTS.
16844 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16846 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16847 number of elements from the PARALLEL rather than the mode.
16848 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16849 rather than GET_MODE_BITSIZE.
16850 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16851 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16852 (aarch64_expand_vec_perm_const_1): Handle polynomial
16853 d->perm.length () and d->perm elements.
16854 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16855 Apply to_constant to d->perm elements.
16856 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16857 polynomial CONST_VECTOR_NUNITS.
16858 (aarch64_move_pointer): Take amount as a poly_int64 rather
16860 (aarch64_progress_pointer): Avoid temporary variable.
16861 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16862 the mode attribute instead of GET_MODE.
16864 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16865 Alan Hayward <alan.hayward@arm.com>
16866 David Sherwood <david.sherwood@arm.com>
16868 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16869 x exists before using it.
16870 (aarch64_add_constant_internal): Rename to...
16871 (aarch64_add_offset_1): ...this. Replace regnum with separate
16872 src and dest rtxes. Handle the case in which they're different,
16873 including when the offset is zero. Replace scratchreg with an rtx.
16874 Use 2 additions if there is no spare register into which we can
16875 move a 16-bit constant.
16876 (aarch64_add_constant): Delete.
16877 (aarch64_add_offset): Replace reg with separate src and dest
16878 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16879 Use aarch64_add_offset_1.
16880 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16881 an rtx rather than an int. Take the delta as a poly_int64
16882 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16883 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16884 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16885 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16886 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16887 and aarch64_add_sp.
16888 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16889 aarch64_add_constant.
16891 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16893 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16894 Use scalar_float_mode.
16896 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16898 * config/aarch64/aarch64-simd.md
16899 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16900 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16901 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16902 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16903 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16904 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16905 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16906 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16907 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16908 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16910 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16913 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16914 targ_options->x_arm_arch_string is non NULL.
16916 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16918 * config/aarch64/aarch64.h
16919 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16921 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16924 * expmed.c (emit_store_flag_force): Swap if const op0
16925 and change VOIDmode to mode of op0.
16927 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16929 PR rtl-optimization/83761
16930 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16931 than bytes to mode_for_size.
16933 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16935 PR middle-end/83189
16936 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16937 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16940 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16942 PR middle-end/83575
16943 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16944 when in layout mode.
16945 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16946 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16949 2018-01-10 Michael Collison <michael.collison@arm.com>
16951 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16952 * config/aarch64/aarch64-option-extension.def: Add
16953 AARCH64_OPT_EXTENSION of 'fp16fml'.
16954 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16955 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16956 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16957 * config/aarch64/constraints.md (Ui7): New constraint.
16958 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16959 (VFMLA_SEL_W): Ditto.
16962 (VFMLA16_LOW): New int iterator.
16963 (VFMLA16_HIGH): Ditto.
16964 (UNSPEC_FMLAL): New unspec.
16965 (UNSPEC_FMLSL): Ditto.
16966 (UNSPEC_FMLAL2): Ditto.
16967 (UNSPEC_FMLSL2): Ditto.
16968 (f16mac): New code attribute.
16969 * config/aarch64/aarch64-simd-builtins.def
16970 (aarch64_fmlal_lowv2sf): Ditto.
16971 (aarch64_fmlsl_lowv2sf): Ditto.
16972 (aarch64_fmlalq_lowv4sf): Ditto.
16973 (aarch64_fmlslq_lowv4sf): Ditto.
16974 (aarch64_fmlal_highv2sf): Ditto.
16975 (aarch64_fmlsl_highv2sf): Ditto.
16976 (aarch64_fmlalq_highv4sf): Ditto.
16977 (aarch64_fmlslq_highv4sf): Ditto.
16978 (aarch64_fmlal_lane_lowv2sf): Ditto.
16979 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16980 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16981 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16982 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16983 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16984 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16985 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16986 (aarch64_fmlal_lane_highv2sf): Ditto.
16987 (aarch64_fmlsl_lane_highv2sf): Ditto.
16988 (aarch64_fmlal_laneq_highv2sf): Ditto.
16989 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16990 (aarch64_fmlalq_lane_highv4sf): Ditto.
16991 (aarch64_fmlsl_lane_highv4sf): Ditto.
16992 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16993 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16994 * config/aarch64/aarch64-simd.md:
16995 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16996 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16997 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16998 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16999 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17000 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17001 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17002 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17003 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17004 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17005 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17006 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17007 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17008 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17009 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17010 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17011 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17012 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17013 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17014 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17015 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17016 (vfmlsl_low_u32): Ditto.
17017 (vfmlalq_low_u32): Ditto.
17018 (vfmlslq_low_u32): Ditto.
17019 (vfmlal_high_u32): Ditto.
17020 (vfmlsl_high_u32): Ditto.
17021 (vfmlalq_high_u32): Ditto.
17022 (vfmlslq_high_u32): Ditto.
17023 (vfmlal_lane_low_u32): Ditto.
17024 (vfmlsl_lane_low_u32): Ditto.
17025 (vfmlal_laneq_low_u32): Ditto.
17026 (vfmlsl_laneq_low_u32): Ditto.
17027 (vfmlalq_lane_low_u32): Ditto.
17028 (vfmlslq_lane_low_u32): Ditto.
17029 (vfmlalq_laneq_low_u32): Ditto.
17030 (vfmlslq_laneq_low_u32): Ditto.
17031 (vfmlal_lane_high_u32): Ditto.
17032 (vfmlsl_lane_high_u32): Ditto.
17033 (vfmlal_laneq_high_u32): Ditto.
17034 (vfmlsl_laneq_high_u32): Ditto.
17035 (vfmlalq_lane_high_u32): Ditto.
17036 (vfmlslq_lane_high_u32): Ditto.
17037 (vfmlalq_laneq_high_u32): Ditto.
17038 (vfmlslq_laneq_high_u32): Ditto.
17039 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17040 (AARCH64_FL_FOR_ARCH8_4): New.
17041 (AARCH64_ISA_F16FML): New ISA flag.
17042 (TARGET_F16FML): New feature flag for fp16fml.
17043 (doc/invoke.texi): Document new fp16fml option.
17045 2018-01-10 Michael Collison <michael.collison@arm.com>
17047 * config/aarch64/aarch64-builtins.c:
17048 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17049 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17050 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17051 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17052 (AARCH64_ISA_SHA3): New ISA flag.
17053 (TARGET_SHA3): New feature flag for sha3.
17054 * config/aarch64/iterators.md (sha512_op): New int attribute.
17055 (CRYPTO_SHA512): New int iterator.
17056 (UNSPEC_SHA512H): New unspec.
17057 (UNSPEC_SHA512H2): Ditto.
17058 (UNSPEC_SHA512SU0): Ditto.
17059 (UNSPEC_SHA512SU1): Ditto.
17060 * config/aarch64/aarch64-simd-builtins.def
17061 (aarch64_crypto_sha512hqv2di): New builtin.
17062 (aarch64_crypto_sha512h2qv2di): Ditto.
17063 (aarch64_crypto_sha512su0qv2di): Ditto.
17064 (aarch64_crypto_sha512su1qv2di): Ditto.
17065 (aarch64_eor3qv8hi): Ditto.
17066 (aarch64_rax1qv2di): Ditto.
17067 (aarch64_xarqv2di): Ditto.
17068 (aarch64_bcaxqv8hi): Ditto.
17069 * config/aarch64/aarch64-simd.md:
17070 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17071 (aarch64_crypto_sha512su0qv2di): Ditto.
17072 (aarch64_crypto_sha512su1qv2di): Ditto.
17073 (aarch64_eor3qv8hi): Ditto.
17074 (aarch64_rax1qv2di): Ditto.
17075 (aarch64_xarqv2di): Ditto.
17076 (aarch64_bcaxqv8hi): Ditto.
17077 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17078 (vsha512h2q_u64): Ditto.
17079 (vsha512su0q_u64): Ditto.
17080 (vsha512su1q_u64): Ditto.
17081 (veor3q_u16): Ditto.
17082 (vrax1q_u64): Ditto.
17083 (vxarq_u64): Ditto.
17084 (vbcaxq_u16): Ditto.
17085 * config/arm/types.md (crypto_sha512): New type attribute.
17086 (crypto_sha3): Ditto.
17087 (doc/invoke.texi): Document new sha3 option.
17089 2018-01-10 Michael Collison <michael.collison@arm.com>
17091 * config/aarch64/aarch64-builtins.c:
17092 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17093 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17094 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17095 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17096 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17097 (AARCH64_ISA_SM4): New ISA flag.
17098 (TARGET_SM4): New feature flag for sm4.
17099 * config/aarch64/aarch64-simd-builtins.def
17100 (aarch64_sm3ss1qv4si): Ditto.
17101 (aarch64_sm3tt1aq4si): Ditto.
17102 (aarch64_sm3tt1bq4si): Ditto.
17103 (aarch64_sm3tt2aq4si): Ditto.
17104 (aarch64_sm3tt2bq4si): Ditto.
17105 (aarch64_sm3partw1qv4si): Ditto.
17106 (aarch64_sm3partw2qv4si): Ditto.
17107 (aarch64_sm4eqv4si): Ditto.
17108 (aarch64_sm4ekeyqv4si): Ditto.
17109 * config/aarch64/aarch64-simd.md:
17110 (aarch64_sm3ss1qv4si): Ditto.
17111 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17112 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17113 (aarch64_sm4eqv4si): Ditto.
17114 (aarch64_sm4ekeyqv4si): Ditto.
17115 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17116 (sm3part_op): Ditto.
17117 (CRYPTO_SM3TT): Ditto.
17118 (CRYPTO_SM3PART): Ditto.
17119 (UNSPEC_SM3SS1): New unspec.
17120 (UNSPEC_SM3TT1A): Ditto.
17121 (UNSPEC_SM3TT1B): Ditto.
17122 (UNSPEC_SM3TT2A): Ditto.
17123 (UNSPEC_SM3TT2B): Ditto.
17124 (UNSPEC_SM3PARTW1): Ditto.
17125 (UNSPEC_SM3PARTW2): Ditto.
17126 (UNSPEC_SM4E): Ditto.
17127 (UNSPEC_SM4EKEY): Ditto.
17128 * config/aarch64/constraints.md (Ui2): New constraint.
17129 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17130 * config/arm/types.md (crypto_sm3): New type attribute.
17131 (crypto_sm4): Ditto.
17132 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17133 (vsm3tt1aq_u32): Ditto.
17134 (vsm3tt1bq_u32): Ditto.
17135 (vsm3tt2aq_u32): Ditto.
17136 (vsm3tt2bq_u32): Ditto.
17137 (vsm3partw1q_u32): Ditto.
17138 (vsm3partw2q_u32): Ditto.
17139 (vsm4eq_u32): Ditto.
17140 (vsm4ekeyq_u32): Ditto.
17141 (doc/invoke.texi): Document new sm4 option.
17143 2018-01-10 Michael Collison <michael.collison@arm.com>
17145 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17146 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17147 (AARCH64_FL_FOR_ARCH8_4): New.
17148 (AARCH64_FL_V8_4): New flag.
17149 (doc/invoke.texi): Document new armv8.4-a option.
17151 2018-01-10 Michael Collison <michael.collison@arm.com>
17153 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17154 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17155 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17156 * config/aarch64/aarch64-option-extension.def: Add
17157 AARCH64_OPT_EXTENSION of 'sha2'.
17158 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17159 (crypto): Disable sha2 and aes if crypto disabled.
17160 (crypto): Enable aes and sha2 if enabled.
17161 (simd): Disable sha2 and aes if simd disabled.
17162 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17164 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17165 (TARGET_SHA2): New feature flag for sha2.
17166 (TARGET_AES): New feature flag for aes.
17167 * config/aarch64/aarch64-simd.md:
17168 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17169 conditional on TARGET_AES.
17170 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17171 (aarch64_crypto_sha1hsi): Make pattern conditional
17173 (aarch64_crypto_sha1hv4si): Ditto.
17174 (aarch64_be_crypto_sha1hv4si): Ditto.
17175 (aarch64_crypto_sha1su1v4si): Ditto.
17176 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17177 (aarch64_crypto_sha1su0v4si): Ditto.
17178 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17179 (aarch64_crypto_sha256su0v4si): Ditto.
17180 (aarch64_crypto_sha256su1v4si): Ditto.
17181 (doc/invoke.texi): Document new aes and sha2 options.
17183 2018-01-10 Martin Sebor <msebor@redhat.com>
17185 PR tree-optimization/83781
17186 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17189 2018-01-11 Martin Sebor <msebor@gmail.com>
17190 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17192 PR tree-optimization/83501
17193 PR tree-optimization/81703
17195 * tree-ssa-strlen.c (get_string_cst): Rename...
17196 (get_string_len): ...to this. Handle global constants.
17197 (handle_char_store): Adjust.
17199 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17200 Jim Wilson <jimw@sifive.com>
17202 * config/riscv/riscv-protos.h (riscv_output_return): New.
17203 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17204 (riscv_attribute_table, riscv_output_return),
17205 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17206 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17207 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17209 (riscv_expand_prologue): Add early return for naked function.
17210 (riscv_expand_epilogue): Likewise.
17211 (riscv_function_ok_for_sibcall): Return false for naked function.
17212 (riscv_set_current_function): New.
17213 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17214 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17215 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17216 * doc/extend.texi (RISC-V Function Attributes): New.
17218 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17220 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17221 check for 128-bit long double before checking TCmode.
17222 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17223 128-bit long doubles before checking TFmode or TCmode.
17224 (FLOAT128_IBM_P): Likewise.
17226 2018-01-10 Martin Sebor <msebor@redhat.com>
17228 PR tree-optimization/83671
17229 * builtins.c (c_strlen): Unconditionally return zero for the empty
17231 Use -Warray-bounds for warnings.
17232 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17233 for non-constant array indices with COMPONENT_REF, arrays of
17234 arrays, and pointers to arrays.
17235 (gimple_fold_builtin_strlen): Determine and set length range for
17236 non-constant character arrays.
17238 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17240 PR middle-end/81897
17241 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17244 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17246 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17248 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17251 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17252 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17253 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17254 indexed_or_indirect_operand predicate.
17255 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17256 (*vsx_le_perm_load_v8hi): Likewise.
17257 (*vsx_le_perm_load_v16qi): Likewise.
17258 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17259 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17260 (*vsx_le_perm_store_v8hi): Likewise.
17261 (*vsx_le_perm_store_v16qi): Likewise.
17262 (eight unnamed splitters): Likewise.
17264 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17266 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17267 * config/rs6000/emmintrin.h: Likewise.
17268 * config/rs6000/mmintrin.h: Likewise.
17269 * config/rs6000/xmmintrin.h: Likewise.
17271 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17274 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17276 * tree.c (tree_nop_conversion): Return true for location wrapper
17278 (maybe_wrap_with_location): New function.
17279 (selftest::check_strip_nops): New function.
17280 (selftest::test_location_wrappers): New function.
17281 (selftest::tree_c_tests): Call it.
17282 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17283 (maybe_wrap_with_location): New decl.
17284 (EXPR_LOCATION_WRAPPER_P): New macro.
17285 (location_wrapper_p): New inline function.
17286 (tree_strip_any_location_wrapper): New inline function.
17288 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17291 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17292 stack_realign_offset for the largest alignment of stack slot
17294 (ix86_find_max_used_stack_alignment): New function.
17295 (ix86_finalize_stack_frame_flags): Use it. Set
17296 max_used_stack_alignment if we don't realign stack.
17297 * config/i386/i386.h (machine_function): Add
17298 max_used_stack_alignment.
17300 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17302 * config/arm/arm.opt (-mbranch-cost): New option.
17303 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17306 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17309 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17310 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17312 2018-01-10 Richard Biener <rguenther@suse.de>
17315 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17316 early out so it also covers the case where we have a non-NULL
17319 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17321 PR tree-optimization/83753
17322 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17323 for non-strided grouped accesses if the number of elements is 1.
17325 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17328 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17329 * i386.h (TARGET_USE_GATHER): Define.
17330 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17332 2018-01-10 Martin Liska <mliska@suse.cz>
17335 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17336 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17338 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17339 CLEANUP_NO_PARTITIONING is not set.
17341 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17343 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17344 for vectors, as a partial revert of r254296.
17345 * rtl.h (const_vec_p): Delete.
17346 (const_vec_duplicate_p): Don't test for vector CONSTs.
17347 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17348 * expmed.c (make_tree): Likewise.
17351 * common.md (E, F): Use CONSTANT_P instead of checking for
17353 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17354 checking for CONST_VECTOR.
17356 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17358 PR middle-end/83575
17359 * predict.c (force_edge_cold): Handle in more sane way edges
17360 with no prediction.
17362 2018-01-09 Carl Love <cel@us.ibm.com>
17364 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17366 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17367 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17368 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17369 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17370 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17371 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17372 * config/rs6000/rs6000-protos.h: Add extern defition for
17373 rs6000_generate_float2_double_code.
17374 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17376 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17377 (float2_v2df): Add define_expand.
17379 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17382 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17383 op_mode in the force_to_mode call.
17385 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17387 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17388 instead of checking each element individually.
17389 (aarch64_evpc_uzp): Likewise.
17390 (aarch64_evpc_zip): Likewise.
17391 (aarch64_evpc_ext): Likewise.
17392 (aarch64_evpc_rev): Likewise.
17393 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17394 instead of checking each element individually. Return true without
17396 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17397 whether all selected elements come from the same input, instead of
17398 checking each element individually. Remove calls to gen_rtx_REG,
17399 start_sequence and end_sequence and instead assert that no rtl is
17402 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17404 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17405 order of HIGH and CONST checks.
17407 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17409 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17410 if the destination isn't an SSA_NAME.
17412 2018-01-09 Richard Biener <rguenther@suse.de>
17414 PR tree-optimization/83668
17415 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17417 (canonicalize_loop_form): ... here, renamed from ...
17418 (canonicalize_loop_closed_ssa_form): ... this and amended to
17419 swap successor edges for loop exit blocks to make us use
17420 the RPO order we need for initial schedule generation.
17422 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17424 PR tree-optimization/64811
17425 * match.pd: When optimizing comparisons with Inf, avoid
17426 introducing or losing exceptions from comparisons with NaN.
17428 2018-01-09 Martin Liska <mliska@suse.cz>
17431 * asan.c (shadow_mem_size): Add gcc_assert.
17433 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17435 Don't save registers in main().
17438 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17439 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17440 * config/avr/avr.c (avr_set_current_function): Don't error if
17441 naked, OS_task or OS_main are specified at the same time.
17442 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17444 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17446 * common/config/avr/avr-common.c (avr_option_optimization_table):
17447 Switch on -mmain-is-OS_task for optimizing compilations.
17449 2018-01-09 Richard Biener <rguenther@suse.de>
17451 PR tree-optimization/83572
17452 * graphite.c: Include cfganal.h.
17453 (graphite_transform_loops): Connect infinite loops to exit
17454 and remove fake edges at the end.
17456 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17458 * ipa-inline.c (edge_badness): Revert accidental checkin.
17460 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17463 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17464 symbols; not inline clones.
17466 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17469 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17470 hard registers. Formatting fixes.
17472 PR preprocessor/83722
17473 * gcc.c (try_generate_repro): Pass
17474 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17475 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17478 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17479 Kito Cheng <kito.cheng@gmail.com>
17481 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17482 (riscv_leaf_function_p): Delete.
17483 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17485 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17487 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17489 (do_ifelse): New function.
17490 (do_isel): New function.
17491 (do_sub3): New function.
17492 (do_add3): New function.
17493 (do_load_mask_compare): New function.
17494 (do_overlap_load_compare): New function.
17495 (expand_compare_loop): New function.
17496 (expand_block_compare): Call expand_compare_loop() when appropriate.
17497 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17498 option description.
17499 (-mblock-compare-inline-loop-limit): New option.
17501 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17504 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17505 Reverse order of second and third operands in first alternative.
17506 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17507 of first and second elements in UNSPEC_VPERMR vector.
17508 (altivec_expand_vec_perm_le): Likewise.
17510 2018-01-08 Jeff Law <law@redhat.com>
17512 PR rtl-optimizatin/81308
17513 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17514 (process_switch): If group_case_labels makes a change, then set
17516 (pass_convert_switch::execute): If a switch is converted, then
17517 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17519 PR rtl-optimization/81308
17520 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17523 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17525 PR target/83663 - Revert r255946
17526 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17527 generation for cases where splatting a value is not useful.
17528 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17529 across a vec_duplicate and a paradoxical subreg forming a vector
17530 mode to a vec_concat.
17532 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17534 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17535 -march=armv8.3-a variants.
17536 * config/arm/t-multilib: Likewise.
17537 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17539 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17541 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17543 (cceq_ior_compare_complement): Give it a name so I can use it, and
17544 change boolean_or_operator predicate to boolean_operator so it can
17545 be used to generate a crand.
17546 (eqne): New code iterator.
17547 (bd/bd_neg): New code_attrs.
17548 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17549 a single define_insn.
17550 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17551 decrement (bdnzt/bdnzf/bdzt/bdzf).
17552 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17553 with the new names of the branch decrement patterns, and added the
17554 names of the branch decrement conditional patterns.
17556 2018-01-08 Richard Biener <rguenther@suse.de>
17558 PR tree-optimization/83563
17559 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17562 2018-01-08 Richard Biener <rguenther@suse.de>
17564 PR middle-end/83713
17565 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17567 2018-01-08 Richard Biener <rguenther@suse.de>
17569 PR tree-optimization/83685
17570 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17571 references to abnormals.
17573 2018-01-08 Richard Biener <rguenther@suse.de>
17576 * dwarf2out.c (output_indirect_strings): Handle empty
17577 skeleton_debug_str_hash.
17578 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17580 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17582 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17583 (emit_store_direct): Likewise.
17584 (arc_trampoline_adjust_address): Likewise.
17585 (arc_asm_trampoline_template): New function.
17586 (arc_initialize_trampoline): Use asm_trampoline_template.
17587 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17588 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17589 * config/arc/arc.md (flush_icache): Delete pattern.
17591 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17593 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17594 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17597 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17600 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17601 by not USED_FOR_TARGET.
17602 (make_pass_resolve_sw_modes): Likewise.
17604 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17606 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17609 2018-01-08 Richard Biener <rguenther@suse.de>
17611 PR middle-end/83580
17612 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17614 2018-01-08 Richard Biener <rguenther@suse.de>
17616 PR middle-end/83517
17617 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17619 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17621 PR middle-end/81897
17622 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17623 basic blocks with a small number of successors.
17624 (convert_control_dep_chain_into_preds): Improve handling of
17626 (dump_predicates): Split apart into...
17627 (dump_pred_chain): ...here...
17628 (dump_pred_info): ...and here.
17629 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17630 (can_chain_union_be_invalidated_p): Improve check for invalidation
17632 (uninit_uses_cannot_happen): Avoid unnecessary if
17633 convert_control_dep_chain_into_preds yielded nothing.
17635 2018-01-06 Martin Sebor <msebor@redhat.com>
17637 PR tree-optimization/83640
17638 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17639 subtracting negative offset from size.
17640 (builtin_access::overlap): Adjust offset bounds of the access to fall
17641 within the size of the object if possible.
17643 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17645 PR rtl-optimization/83699
17646 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17647 extract_bit_field_as_subreg to cases in which the extracted
17648 value is also a vector.
17650 * lra-constraints.c (process_alt_operands): Test for the equivalence
17651 substitutions when detecting a possible reload cycle.
17653 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17656 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17657 by default if flag_selective_schedling{,2}. Formatting fixes.
17659 PR rtl-optimization/83682
17660 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17661 if it has non-VECTOR_MODE element mode.
17662 (vec_duplicate_p): Likewise.
17664 PR middle-end/83694
17665 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17666 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17668 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17671 * config/i386/i386-builtin.def
17672 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17673 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17674 Require also OPTION_MASK_ISA_AVX512F in addition to
17675 OPTION_MASK_ISA_GFNI.
17676 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17677 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17678 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17679 to OPTION_MASK_ISA_GFNI.
17680 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17681 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17682 OPTION_MASK_ISA_AVX512BW.
17683 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17684 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17685 addition to OPTION_MASK_ISA_GFNI.
17686 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17687 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17688 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17689 to OPTION_MASK_ISA_GFNI.
17690 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17691 a requirement for all ISAs rather than any of them with a few
17693 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17695 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17696 bitmasks to be enabled with 3 exceptions, instead of requiring any
17697 enabled ISA with lots of exceptions.
17698 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17699 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17700 Change avx512bw in isa attribute to avx512f.
17701 * config/i386/sgxintrin.h: Add license boilerplate.
17702 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17703 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17704 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17705 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17707 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17708 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17709 temporarily sse2 rather than sse if not enabled already.
17712 * config/i386/sse.md (VI248_VLBW): Rename to ...
17713 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17714 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17715 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17716 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17717 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17718 mode iterator instead of VI248_VLBW.
17720 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17722 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17723 (record_modified): Skip clobbers; add debug output.
17724 (param_change_prob): Use sreal frequencies.
17726 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17728 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17729 punt for user-aligned variables.
17731 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17733 * tree-chrec.c (chrec_contains_symbols): Return true for
17736 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17739 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17740 of (x|y) == x for BICS pattern.
17742 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17744 PR tree-optimization/83605
17745 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17746 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17749 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17751 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17752 * config/epiphany/rtems.h: New file.
17754 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17755 Uros Bizjak <ubizjak@gmail.com>
17758 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17759 QIreg_operand instead of register_operand predicate.
17760 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17761 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17762 comments instead of -fmitigate[-_]rop.
17764 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17767 * cgraphunit.c (symbol_table::compile): Switch to text_section
17768 before calling assembly_start debug hook.
17769 * run-rtl-passes.c (run_rtl_passes): Likewise.
17772 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17774 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17775 range_int_cst_p rather than !symbolic_range_p before calling
17776 extract_range_from_multiplicative_op_1.
17778 2018-01-04 Jeff Law <law@redhat.com>
17780 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17781 redundant test in assertion.
17783 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17785 * doc/rtl.texi: Document machine_mode wrapper classes.
17787 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17789 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17790 using tree_to_uhwi.
17792 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17794 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17795 the VEC_PERM_EXPR fold to fail.
17797 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17800 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17801 to switched_sections.
17803 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17806 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17807 test for d.testing.
17809 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17812 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17813 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17815 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17818 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17819 is BLKmode and bitpos not zero or mode change is needed.
17821 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17824 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17827 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17830 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17831 instead of MULT rtx. Update all corresponding splitters.
17832 (*saddl_se): Ditto.
17833 (*ssub<modesuffix>): Ditto.
17834 (*ssubl_se): Ditto.
17835 (*cmp_sadd_di): Update split patterns.
17836 (*cmp_sadd_si): Ditto.
17837 (*cmp_sadd_sidi): Ditto.
17838 (*cmp_ssub_di): Ditto.
17839 (*cmp_ssub_si): Ditto.
17840 (*cmp_ssub_sidi): Ditto.
17841 * config/alpha/predicates.md (const23_operand): New predicate.
17842 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17843 Look for ASHIFT, not MULT inner operand.
17844 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17846 2018-01-04 Martin Liska <mliska@suse.cz>
17848 PR gcov-profile/83669
17849 * gcov.c (output_intermediate_file): Add version to intermediate
17851 * doc/gcov.texi: Document new field 'version' in intermediate
17852 file format. Fix location of '-k' option of gcov command.
17854 2018-01-04 Martin Liska <mliska@suse.cz>
17857 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17859 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17861 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17863 2018-01-03 Martin Sebor <msebor@redhat.com>
17865 PR tree-optimization/83655
17866 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17867 checking calls with invalid arguments.
17869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17871 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17872 (vectorizable_mask_load_store): Delete.
17873 (vectorizable_call): Return false for masked loads and stores.
17874 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17875 instead of gimple_assign_rhs1.
17876 (vectorizable_load): Handle IFN_MASK_LOAD.
17877 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17881 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17883 (vectorizable_mask_load_store): ...here.
17884 (vectorizable_load): ...and here.
17886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17888 * tree-vect-stmts.c (vect_build_all_ones_mask)
17889 (vect_build_zero_merge_argument): New functions, split out from...
17890 (vectorizable_load): ...here.
17892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17894 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17896 (vectorizable_mask_load_store): ...here.
17897 (vectorizable_store): ...and here.
17899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17901 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17903 (vectorizable_mask_load_store): ...here.
17905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17907 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17908 (vect_model_store_cost): Take a vec_load_store_type instead of a
17910 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17911 (vect_model_store_cost): Take a vec_load_store_type instead of a
17913 (vectorizable_mask_load_store): Update accordingly.
17914 (vectorizable_store): Likewise.
17915 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17919 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17920 IFN_MASK_LOAD calls here rather than...
17921 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17924 Alan Hayward <alan.hayward@arm.com>
17925 David Sherwood <david.sherwood@arm.com>
17927 * expmed.c (extract_bit_field_1): For vector extracts,
17928 fall back to extract_bit_field_as_subreg if vec_extract
17931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17932 Alan Hayward <alan.hayward@arm.com>
17933 David Sherwood <david.sherwood@arm.com>
17935 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17936 they are variable or constant sized.
17937 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17938 slots for constant-sized data.
17940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17941 Alan Hayward <alan.hayward@arm.com>
17942 David Sherwood <david.sherwood@arm.com>
17944 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17945 handling COND_EXPRs with boolean comparisons, try to find a better
17946 basis for the mask type than the boolean itself.
17948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17950 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17951 is calculated and how it can be overridden.
17952 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17953 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17955 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17959 Alan Hayward <alan.hayward@arm.com>
17960 David Sherwood <david.sherwood@arm.com>
17962 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17963 Remove the mode argument.
17964 (aarch64_simd_valid_immediate): Remove the mode and inverse
17966 * config/aarch64/iterators.md (bitsize): New iterator.
17967 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17968 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17969 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17970 aarch64_simd_valid_immediate.
17971 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17972 (aarch64_reg_or_bic_imm): Likewise.
17973 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17974 with an insn_type enum and msl with a modifier_type enum.
17975 Replace element_width with a scalar_mode. Change the shift
17976 to unsigned int. Add constructors for scalar_float_mode and
17977 scalar_int_mode elements.
17978 (aarch64_vect_float_const_representable_p): Delete.
17979 (aarch64_can_const_movi_rtx_p)
17980 (aarch64_simd_scalar_immediate_valid_for_move)
17981 (aarch64_simd_make_constant): Update call to
17982 aarch64_simd_valid_immediate.
17983 (aarch64_advsimd_valid_immediate_hs): New function.
17984 (aarch64_advsimd_valid_immediate): Likewise.
17985 (aarch64_simd_valid_immediate): Remove mode and inverse
17986 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17987 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17988 and aarch64_float_const_representable_p on the result.
17989 (aarch64_output_simd_mov_immediate): Remove mode argument.
17990 Update call to aarch64_simd_valid_immediate and use of
17991 simd_immediate_info.
17992 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18000 (mode_nunits): Likewise CONST_MODE_NUNITS.
18001 * machmode.def (ADJUST_NUNITS): Document.
18002 * genmodes.c (mode_data::need_nunits_adj): New field.
18003 (blank_mode): Update accordingly.
18004 (adj_nunits): New variable.
18005 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18007 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18008 listed in adj_nunits.
18009 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18010 listed in adj_nunits. Don't emit case statements for such modes.
18011 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18012 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18013 nothing if adj_nunits is nonnull.
18014 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18015 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18016 (emit_mode_fbit): Update use of print_maybe_const_decl.
18017 (emit_move_size): Likewise. Treat the array as non-const
18019 (emit_mode_adjustments): Handle adj_nunits.
18021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18023 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18024 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18025 (VECTOR_MODES): Use it.
18026 (make_vector_modes): Take the prefix as an argument.
18028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18029 Alan Hayward <alan.hayward@arm.com>
18030 David Sherwood <david.sherwood@arm.com>
18032 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18033 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18034 for MODE_VECTOR_BOOL.
18035 * machmode.def (VECTOR_BOOL_MODE): Document.
18036 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18037 (make_vector_bool_mode): New function.
18038 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18040 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18041 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18043 * stor-layout.c (int_mode_for_mode): Likewise.
18044 * tree.c (build_vector_type_for_mode): Likewise.
18045 * varasm.c (output_constant_pool_2): Likewise.
18046 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18047 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18048 for MODE_VECTOR_BOOL.
18049 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18050 of mode class checks.
18051 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18052 instead of a list of mode class checks.
18053 (expand_vector_scalar_condition): Likewise.
18054 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18057 Alan Hayward <alan.hayward@arm.com>
18058 David Sherwood <david.sherwood@arm.com>
18060 * machmode.h (mode_size): Change from unsigned short to
18062 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18063 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18064 or if measurement_type is not polynomial.
18065 (fixed_size_mode::includes_p): Check for constant-sized modes.
18066 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18067 return a poly_uint16 rather than an unsigned short.
18068 (emit_mode_size): Change the type of mode_size from unsigned short
18069 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18070 (emit_mode_adjustments): Cope with polynomial vector sizes.
18071 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18073 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18075 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18076 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18077 * caller-save.c (setup_save_areas): Likewise.
18078 (replace_reg_with_saved_mem): Likewise.
18079 * calls.c (emit_library_call_value_1): Likewise.
18080 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18081 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18082 (gen_lowpart_for_combine): Likewise.
18083 * convert.c (convert_to_integer_1): Likewise.
18084 * cse.c (equiv_constant, cse_insn): Likewise.
18085 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18086 (cselib_subst_to_values): Likewise.
18087 * dce.c (word_dce_process_block): Likewise.
18088 * df-problems.c (df_word_lr_mark_ref): Likewise.
18089 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18090 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18091 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18092 (rtl_for_decl_location): Likewise.
18093 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18094 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18095 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18096 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18097 (expand_expr_real_1): Likewise.
18098 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18099 (pad_below): Likewise.
18100 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18101 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18102 * ira.c (get_subreg_tracking_sizes): Likewise.
18103 * ira-build.c (ira_create_allocno_objects): Likewise.
18104 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18105 (ira_sort_regnos_for_alter_reg): Likewise.
18106 * ira-costs.c (record_operand_costs): Likewise.
18107 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18108 (resolve_simple_move): Likewise.
18109 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18110 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18111 (lra_constraints): Likewise.
18112 (CONST_POOL_OK_P): Reject variable-sized modes.
18113 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18114 (add_pseudo_to_slot, lra_spill): Likewise.
18115 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18116 * optabs-query.c (get_best_extraction_insn): Likewise.
18117 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18118 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18119 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18120 * recog.c (offsettable_address_addr_space_p): Likewise.
18121 * regcprop.c (maybe_mode_change): Likewise.
18122 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18123 * regrename.c (build_def_use): Likewise.
18124 * regstat.c (dump_reg_info): Likewise.
18125 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18126 (find_reloads, find_reloads_subreg_address): Likewise.
18127 * reload1.c (eliminate_regs_1): Likewise.
18128 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18129 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18130 (simplify_binary_operation_1, simplify_subreg): Likewise.
18131 * targhooks.c (default_function_arg_padding): Likewise.
18132 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18133 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18134 (verify_gimple_assign_ternary): Likewise.
18135 * tree-inline.c (estimate_move_cost): Likewise.
18136 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18137 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18138 (get_address_cost_ainc): Likewise.
18139 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18140 (vect_supportable_dr_alignment): Likewise.
18141 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18142 (vectorizable_reduction): Likewise.
18143 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18144 (vectorizable_operation, vectorizable_load): Likewise.
18145 * tree.c (build_same_sized_truth_vector_type): Likewise.
18146 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18147 * var-tracking.c (emit_note_insn_var_location): Likewise.
18148 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18149 (ADDR_VEC_ALIGN): Likewise.
18151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18152 Alan Hayward <alan.hayward@arm.com>
18153 David Sherwood <david.sherwood@arm.com>
18155 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18157 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18158 or if measurement_type is polynomial.
18159 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18160 * combine.c (make_extraction): Likewise.
18161 * dse.c (find_shift_sequence): Likewise.
18162 * dwarf2out.c (mem_loc_descriptor): Likewise.
18163 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18164 (extract_bit_field, extract_low_bits): Likewise.
18165 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18166 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18167 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18168 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18169 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18170 * reload.c (find_reloads): Likewise.
18171 * reload1.c (alter_reg): Likewise.
18172 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18173 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18174 * tree-if-conv.c (predicate_mem_writes): Likewise.
18175 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18176 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18177 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18178 * valtrack.c (dead_debug_insert_temp): Likewise.
18179 * varasm.c (mergeable_constant_section): Likewise.
18180 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18183 Alan Hayward <alan.hayward@arm.com>
18184 David Sherwood <david.sherwood@arm.com>
18186 * expr.c (expand_assignment): Cope with polynomial mode sizes
18187 when assigning to a CONCAT.
18189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18190 Alan Hayward <alan.hayward@arm.com>
18191 David Sherwood <david.sherwood@arm.com>
18193 * machmode.h (mode_precision): Change from unsigned short to
18195 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18197 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18198 or if measurement_type is not polynomial.
18199 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18200 in which the mode is already known to be a scalar_int_mode.
18201 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18202 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18204 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18205 for GET_MODE_PRECISION.
18206 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18207 for GET_MODE_PRECISION.
18208 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18210 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18211 (expand_field_assignment, make_extraction): Likewise.
18212 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18213 (get_last_value): Likewise.
18214 * convert.c (convert_to_integer_1): Likewise.
18215 * cse.c (cse_insn): Likewise.
18216 * expr.c (expand_expr_real_1): Likewise.
18217 * lra-constraints.c (simplify_operand_subreg): Likewise.
18218 * optabs-query.c (can_atomic_load_p): Likewise.
18219 * optabs.c (expand_atomic_load): Likewise.
18220 (expand_atomic_store): Likewise.
18221 * ree.c (combine_reaching_defs): Likewise.
18222 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18223 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18224 * tree.h (type_has_mode_precision_p): Likewise.
18225 * ubsan.c (instrument_si_overflow): Likewise.
18227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18228 Alan Hayward <alan.hayward@arm.com>
18229 David Sherwood <david.sherwood@arm.com>
18231 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18232 polynomial numbers of units.
18233 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18234 (valid_vector_subparts_p): New function.
18235 (build_vector_type): Remove temporary shim and take the number
18236 of units as a poly_uint64 rather than an int.
18237 (build_opaque_vector_type): Take the number of units as a
18238 poly_uint64 rather than an int.
18239 * tree.c (build_vector_from_ctor): Handle polynomial
18240 TYPE_VECTOR_SUBPARTS.
18241 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18242 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18243 (build_vector_from_val): If the number of units is variable,
18244 use build_vec_duplicate_cst for constant operands and
18245 VEC_DUPLICATE_EXPR otherwise.
18246 (make_vector_type): Remove temporary is_constant ().
18247 (build_vector_type, build_opaque_vector_type): Take the number of
18248 units as a poly_uint64 rather than an int.
18249 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18251 * cfgexpand.c (expand_debug_expr): Likewise.
18252 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18253 (store_constructor, expand_expr_real_1): Likewise.
18254 (const_scalar_mask_from_tree): Likewise.
18255 * fold-const-call.c (fold_const_reduction): Likewise.
18256 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18257 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18258 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18259 (fold_relational_const): Likewise.
18260 (native_interpret_vector): Likewise. Change the size from an
18261 int to an unsigned int.
18262 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18263 TYPE_VECTOR_SUBPARTS.
18264 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18265 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18266 duplicating a non-constant operand into a variable-length vector.
18267 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18268 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18269 * ipa-icf.c (sem_variable::equals): Likewise.
18270 * match.pd: Likewise.
18271 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18272 * print-tree.c (print_node): Likewise.
18273 * stor-layout.c (layout_type): Likewise.
18274 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18275 * tree-cfg.c (verify_gimple_comparison): Likewise.
18276 (verify_gimple_assign_binary): Likewise.
18277 (verify_gimple_assign_ternary): Likewise.
18278 (verify_gimple_assign_single): Likewise.
18279 * tree-pretty-print.c (dump_generic_node): Likewise.
18280 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18281 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18282 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18283 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18284 (vect_shift_permute_load_chain): Likewise.
18285 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18286 (expand_vector_condition, optimize_vector_constructor): Likewise.
18287 (lower_vec_perm, get_compute_type): Likewise.
18288 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18289 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18290 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18291 (vect_recog_mask_conversion_pattern): Likewise.
18292 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18293 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18294 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18295 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18296 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18297 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18298 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18299 (supportable_widening_operation): Likewise.
18300 (supportable_narrowing_operation): Likewise.
18301 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18303 * varasm.c (output_constant): Likewise.
18305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18306 Alan Hayward <alan.hayward@arm.com>
18307 David Sherwood <david.sherwood@arm.com>
18309 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18310 so that both the length == 3 and length != 3 cases set up their
18311 own permute vectors. Add comments explaining why we know the
18312 number of elements is constant.
18313 (vect_permute_load_chain): Likewise.
18315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18316 Alan Hayward <alan.hayward@arm.com>
18317 David Sherwood <david.sherwood@arm.com>
18319 * machmode.h (mode_nunits): Change from unsigned char to
18321 (ONLY_FIXED_SIZE_MODES): New macro.
18322 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18323 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18324 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18326 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18327 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18328 or if measurement_type is not polynomial.
18329 * genmodes.c (ZERO_COEFFS): New macro.
18330 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18332 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18333 Use ZERO_COEFFS when emitting initializers.
18334 * data-streamer.h (bp_pack_poly_value): New function.
18335 (bp_unpack_poly_value): Likewise.
18336 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18337 for GET_MODE_NUNITS.
18338 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18339 for GET_MODE_NUNITS.
18340 * tree.c (make_vector_type): Remove temporary shim and make
18341 the real function take the number of units as a poly_uint64
18342 rather than an int.
18343 (build_vector_type_for_mode): Handle polynomial nunits.
18344 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18345 * emit-rtl.c (const_vec_series_p_1): Likewise.
18346 (gen_rtx_CONST_VECTOR): Likewise.
18347 * fold-const.c (test_vec_duplicate_folding): Likewise.
18348 * genrecog.c (validate_pattern): Likewise.
18349 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18350 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18351 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18352 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18353 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18354 * rtlanal.c (subreg_get_info): Likewise.
18355 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18356 (vect_grouped_load_supported): Likewise.
18357 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18358 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18359 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18360 (simplify_const_unary_operation, simplify_binary_operation_1)
18361 (simplify_const_binary_operation, simplify_ternary_operation)
18362 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18363 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18364 instead of CONST_VECTOR_NUNITS.
18365 * varasm.c (output_constant_pool_2): Likewise.
18366 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18367 explicit-encoded elements in the XVEC for variable-length vectors.
18369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18371 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18374 Alan Hayward <alan.hayward@arm.com>
18375 David Sherwood <david.sherwood@arm.com>
18377 * coretypes.h (fixed_size_mode): Declare.
18378 (fixed_size_mode_pod): New typedef.
18379 * builtins.h (target_builtins::x_apply_args_mode)
18380 (target_builtins::x_apply_result_mode): Change type to
18381 fixed_size_mode_pod.
18382 * builtins.c (apply_args_size, apply_result_size, result_vector)
18383 (expand_builtin_apply_args_1, expand_builtin_apply)
18384 (expand_builtin_return): Update accordingly.
18386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18388 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18389 * cselib.c (cselib_hash_rtx): Likewise.
18390 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18391 CONST_VECTOR encoding.
18393 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18394 Jeff Law <law@redhat.com>
18397 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18398 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18399 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18400 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18403 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18404 explicitly probe *sp in a noreturn function if there were any callee
18405 register saves or frame pointer is needed.
18407 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18410 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18411 BLKmode for ternary, binary or unary expressions.
18414 * var-tracking.c (delete_vta_debug_insn): New inline function.
18415 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18416 insns from get_insns () to NULL instead of each bb separately.
18417 Use delete_vta_debug_insn. No longer static.
18418 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18419 delete_vta_debug_insns callers.
18420 * rtl.h (delete_vta_debug_insns): Declare.
18421 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18422 instead of variable_tracking_main.
18424 2018-01-03 Martin Sebor <msebor@redhat.com>
18426 PR tree-optimization/83603
18427 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18428 arguments past the endof the argument list in functions declared
18429 without a prototype.
18430 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18431 Avoid checking when arguments are null.
18433 2018-01-03 Martin Sebor <msebor@redhat.com>
18436 * doc/extend.texi (attribute const): Fix a typo.
18437 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18438 issuing -Wsuggest-attribute for void functions.
18440 2018-01-03 Martin Sebor <msebor@redhat.com>
18442 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18443 offset_int::from instead of wide_int::to_shwi.
18444 (maybe_diag_overlap): Remove assertion.
18445 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18446 * gimple-ssa-sprintf.c (format_directive): Same.
18447 (parse_directive): Same.
18448 (sprintf_dom_walker::compute_format_length): Same.
18449 (try_substitute_return_value): Same.
18451 2018-01-03 Jeff Law <law@redhat.com>
18453 PR middle-end/83654
18454 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18455 non-constant residual for zero at runtime and avoid probing in
18456 that case. Reorganize code for trailing problem to mirror handling
18459 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18461 PR tree-optimization/83501
18462 * tree-ssa-strlen.c (get_string_cst): New.
18463 (handle_char_store): Call get_string_cst.
18465 2018-01-03 Martin Liska <mliska@suse.cz>
18467 PR tree-optimization/83593
18468 * tree-ssa-strlen.c: Include tree-cfg.h.
18469 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18470 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18471 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18473 (strlen_dom_walker::before_dom_children): Call
18474 gimple_purge_dead_eh_edges. Dump tranformation with details
18476 (strlen_dom_walker::before_dom_children): Update call by adding
18477 new argument cleanup_eh.
18478 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18480 2018-01-03 Martin Liska <mliska@suse.cz>
18483 * cif-code.def (VARIADIC_THUNK): New enum value.
18484 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18487 2018-01-03 Jan Beulich <jbeulich@suse.com>
18489 * sse.md (mov<mode>_internal): Tighten condition for when to use
18490 vmovdqu<ssescalarsize> for TI and OI modes.
18492 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18494 Update copyright years.
18496 2018-01-03 Martin Liska <mliska@suse.cz>
18499 * ipa-visibility.c (function_and_variable_visibility): Skip
18500 functions with noipa attribure.
18502 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18504 * gcc.c (process_command): Update copyright notice dates.
18505 * gcov-dump.c (print_version): Ditto.
18506 * gcov.c (print_version): Ditto.
18507 * gcov-tool.c (print_version): Ditto.
18508 * gengtype.c (create_file): Ditto.
18509 * doc/cpp.texi: Bump @copying's copyright year.
18510 * doc/cppinternals.texi: Ditto.
18511 * doc/gcc.texi: Ditto.
18512 * doc/gccint.texi: Ditto.
18513 * doc/gcov.texi: Ditto.
18514 * doc/install.texi: Ditto.
18515 * doc/invoke.texi: Ditto.
18517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18519 * vector-builder.h (vector_builder::m_full_nelts): Change from
18520 unsigned int to poly_uint64.
18521 (vector_builder::full_nelts): Update prototype accordingly.
18522 (vector_builder::new_vector): Likewise.
18523 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18524 (vector_builder::operator ==): Likewise.
18525 (vector_builder::finalize): Likewise.
18526 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18527 Take the number of elements as a poly_uint64 rather than an
18529 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18530 from unsigned int to poly_uint64.
18531 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18532 (vec_perm_indices::new_vector): Likewise.
18533 (vec_perm_indices::length): Likewise.
18534 (vec_perm_indices::nelts_per_input): Likewise.
18535 (vec_perm_indices::input_nelts): Likewise.
18536 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18537 number of elements per input as a poly_uint64 rather than an
18538 unsigned int. Use the original encoding for variable-length
18539 vectors, rather than clamping each individual element.
18540 For the second and subsequent elements in each pattern,
18541 clamp the step and base before clamping their sum.
18542 (vec_perm_indices::series_p): Handle polynomial element counts.
18543 (vec_perm_indices::all_in_range_p): Likewise.
18544 (vec_perm_indices_to_tree): Likewise.
18545 (vec_perm_indices_to_rtx): Likewise.
18546 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18547 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18548 (tree_vector_builder::new_binary_operation): Handle polynomial
18549 element counts. Return false if we need to know the number
18550 of elements at compile time.
18551 * fold-const.c (fold_vec_perm): Punt if the number of elements
18552 isn't known at compile time.
18554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18556 * vec-perm-indices.h (vec_perm_builder): Change element type
18557 from HOST_WIDE_INT to poly_int64.
18558 (vec_perm_indices::element_type): Update accordingly.
18559 (vec_perm_indices::clamp): Handle polynomial element_types.
18560 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18561 (vec_perm_indices::all_in_range_p): Likewise.
18562 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18564 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18565 polynomial vec_perm_indices element types.
18566 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18567 * fold-const.c (fold_vec_perm): Likewise.
18568 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18569 * tree-vect-generic.c (lower_vec_perm): Likewise.
18570 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18571 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18572 element type to HOST_WIDE_INT.
18574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18575 Alan Hayward <alan.hayward@arm.com>
18576 David Sherwood <david.sherwood@arm.com>
18578 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18579 rather than an int. Use plus_constant.
18580 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18581 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18584 Alan Hayward <alan.hayward@arm.com>
18585 David Sherwood <david.sherwood@arm.com>
18587 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18588 a HOST_WIDE_INT to a poly_int64.
18590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18591 Alan Hayward <alan.hayward@arm.com>
18592 David Sherwood <david.sherwood@arm.com>
18594 * calls.c (load_register_parameters): Cope with polynomial
18595 mode sizes. Require a constant size for BLKmode parameters
18596 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18597 forces a parameter to be padded at the lsb end in order to
18598 fill a complete number of words, require the parameter size
18599 to be ordered wrt UNITS_PER_WORD.
18601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18602 Alan Hayward <alan.hayward@arm.com>
18603 David Sherwood <david.sherwood@arm.com>
18605 * reload1.c (spill_stack_slot_width): Change element type
18606 from unsigned int to poly_uint64_pod.
18607 (alter_reg): Treat mode sizes as polynomial.
18609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18610 Alan Hayward <alan.hayward@arm.com>
18611 David Sherwood <david.sherwood@arm.com>
18613 * reload.c (complex_word_subreg_p): New function.
18614 (reload_inner_reg_of_subreg, push_reload): Use it.
18616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18617 Alan Hayward <alan.hayward@arm.com>
18618 David Sherwood <david.sherwood@arm.com>
18620 * lra-constraints.c (process_alt_operands): Reject matched
18621 operands whose sizes aren't ordered.
18622 (match_reload): Refer to this check here.
18624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18625 Alan Hayward <alan.hayward@arm.com>
18626 David Sherwood <david.sherwood@arm.com>
18628 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18629 that the mode size is in the set {1, 2, 4, 8, 16}.
18631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18632 Alan Hayward <alan.hayward@arm.com>
18633 David Sherwood <david.sherwood@arm.com>
18635 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18636 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18643 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18644 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18645 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18646 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18647 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18648 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18649 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18650 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18651 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18653 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18654 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18655 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18656 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18657 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18658 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18659 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18660 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18661 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18662 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18664 * expr.c (emit_move_resolve_push): Treat the input and result
18665 of PUSH_ROUNDING as a poly_int64.
18666 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18667 (emit_push_insn): Likewise.
18668 * lra-eliminations.c (mark_not_eliminable): Likewise.
18669 * recog.c (push_operand): Likewise.
18670 * reload1.c (elimination_effects): Likewise.
18671 * rtlanal.c (nonzero_bits1): Likewise.
18672 * calls.c (store_one_arg): Likewise. Require the padding to be
18673 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18680 Use plus_constant instead of gen_rtx_PLUS.
18682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18683 Alan Hayward <alan.hayward@arm.com>
18684 David Sherwood <david.sherwood@arm.com>
18686 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18687 rather than an int.
18689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18690 Alan Hayward <alan.hayward@arm.com>
18691 David Sherwood <david.sherwood@arm.com>
18693 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18694 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18695 via stack temporaries. Treat the mode size as polynomial too.
18697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18698 Alan Hayward <alan.hayward@arm.com>
18699 David Sherwood <david.sherwood@arm.com>
18701 * expr.c (expand_expr_real_2): When handling conversions involving
18702 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18703 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18704 as a poly_uint64 too.
18706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18707 Alan Hayward <alan.hayward@arm.com>
18708 David Sherwood <david.sherwood@arm.com>
18710 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
18717 REGMODE_NATURAL_SIZE.
18718 * expmed.c (store_bit_field_1): Likewise.
18719 * expr.c (store_constructor): Likewise.
18720 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18721 and polynomial REGMODE_NATURAL_SIZE.
18722 (gen_lowpart_common): Likewise.
18723 * reginfo.c (record_subregs_of_mode): Likewise.
18724 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18731 numbers of elements.
18733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18734 Alan Hayward <alan.hayward@arm.com>
18735 David Sherwood <david.sherwood@arm.com>
18737 * match.pd: Cope with polynomial numbers of vector elements.
18739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18740 Alan Hayward <alan.hayward@arm.com>
18741 David Sherwood <david.sherwood@arm.com>
18743 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18744 in a POINTER_PLUS_EXPR.
18746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18747 Alan Hayward <alan.hayward@arm.com>
18748 David Sherwood <david.sherwood@arm.com>
18750 * omp-simd-clone.c (simd_clone_subparts): New function.
18751 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18752 (ipa_simd_modify_function_body): Likewise.
18754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18755 Alan Hayward <alan.hayward@arm.com>
18756 David Sherwood <david.sherwood@arm.com>
18758 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18759 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18760 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18761 (expand_vector_condition, vector_element): Likewise.
18762 (subparts_gt): New function.
18763 (get_compute_type): Use subparts_gt.
18764 (count_type_subparts): Delete.
18765 (expand_vector_operations_1): Use subparts_gt instead of
18766 count_type_subparts.
18768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18769 Alan Hayward <alan.hayward@arm.com>
18770 David Sherwood <david.sherwood@arm.com>
18772 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18773 (vect_compile_time_alias): ...this new function. Do the calculation
18774 on poly_ints rather than trees.
18775 (vect_prune_runtime_alias_test_list): Update call accordingly.
18777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18778 Alan Hayward <alan.hayward@arm.com>
18779 David Sherwood <david.sherwood@arm.com>
18781 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18783 (vect_schedule_slp_instance): Likewise.
18785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18786 Alan Hayward <alan.hayward@arm.com>
18787 David Sherwood <david.sherwood@arm.com>
18789 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18790 constant and extern definitions for variable-length vectors.
18791 (vect_get_constant_vectors): Note that the number of units
18792 is known to be constant.
18794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18795 Alan Hayward <alan.hayward@arm.com>
18796 David Sherwood <david.sherwood@arm.com>
18798 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18799 of units as polynomial. Choose between WIDE and NARROW based
18802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18803 Alan Hayward <alan.hayward@arm.com>
18804 David Sherwood <david.sherwood@arm.com>
18806 * tree-vect-stmts.c (simd_clone_subparts): New function.
18807 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18810 Alan Hayward <alan.hayward@arm.com>
18811 David Sherwood <david.sherwood@arm.com>
18813 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18814 vectors as polynomial. Use build_index_vector for
18815 IFN_GOMP_SIMD_LANE.
18817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18818 Alan Hayward <alan.hayward@arm.com>
18819 David Sherwood <david.sherwood@arm.com>
18821 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18822 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18823 for variable-length vectors.
18824 (vectorizable_mask_load_store): Treat the number of units as
18825 polynomial, asserting that it is constant if the condition has
18826 already been enforced.
18827 (vectorizable_store, vectorizable_load): Likewise.
18829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18830 Alan Hayward <alan.hayward@arm.com>
18831 David Sherwood <david.sherwood@arm.com>
18833 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18834 of units as polynomial. Punt if we can't tell at compile time
18835 which vector contains the final result.
18837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18838 Alan Hayward <alan.hayward@arm.com>
18839 David Sherwood <david.sherwood@arm.com>
18841 * tree-vect-loop.c (vectorizable_induction): Treat the number
18842 of units as polynomial. Punt on SLP inductions. Use an integer
18843 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18844 cast of such a series for variable-length floating-point
18847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18848 Alan Hayward <alan.hayward@arm.com>
18849 David Sherwood <david.sherwood@arm.com>
18851 * tree.h (build_index_vector): Declare.
18852 * tree.c (build_index_vector): New function.
18853 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18854 of units as polynomial, forcibly converting it to a constant if
18855 vectorizable_reduction has already enforced the condition.
18856 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18857 to create a {1,2,3,...} vector.
18858 (vectorizable_reduction): Treat the number of units as polynomial.
18859 Choose vectype_in based on the largest scalar element size rather
18860 than the smallest number of units. Enforce the restrictions
18863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18864 Alan Hayward <alan.hayward@arm.com>
18865 David Sherwood <david.sherwood@arm.com>
18867 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18868 number of units as polynomial.
18870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18871 Alan Hayward <alan.hayward@arm.com>
18872 David Sherwood <david.sherwood@arm.com>
18874 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18875 * target.def (autovectorize_vector_sizes): Return the vector sizes
18876 by pointer, using vector_sizes rather than a bitmask.
18877 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18878 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18879 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18881 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18882 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18883 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18884 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18885 * omp-general.c (omp_max_vf): Likewise.
18886 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18887 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18888 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18889 * tree-vect-slp.c (vect_slp_bb): Likewise.
18890 * doc/tm.texi: Regenerate.
18891 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18893 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18894 the vector size as a poly_uint64 rather than an unsigned int.
18895 (current_vector_size): Change from an unsigned int to a poly_uint64.
18896 (get_vectype_for_scalar_type): Update accordingly.
18897 * tree.h (build_truth_vector_type): Take the size and number of
18898 units as a poly_uint64 rather than an unsigned int.
18899 (build_vector_type): Add a temporary overload that takes
18900 the number of units as a poly_uint64 rather than an unsigned int.
18901 * tree.c (make_vector_type): Likewise.
18902 (build_truth_vector_type): Take the number of units as a poly_uint64
18903 rather than an unsigned int.
18905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18906 Alan Hayward <alan.hayward@arm.com>
18907 David Sherwood <david.sherwood@arm.com>
18909 * target.def (get_mask_mode): Take the number of units and length
18910 as poly_uint64s rather than unsigned ints.
18911 * targhooks.h (default_get_mask_mode): Update accordingly.
18912 * targhooks.c (default_get_mask_mode): Likewise.
18913 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18914 * doc/tm.texi: Regenerate.
18916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18917 Alan Hayward <alan.hayward@arm.com>
18918 David Sherwood <david.sherwood@arm.com>
18920 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18921 * omp-general.c (omp_max_vf): Likewise.
18922 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18923 (expand_omp_simd): Handle polynomial safelen.
18924 * omp-low.c (omplow_simd_context): Add a default constructor.
18925 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18926 (lower_rec_simd_input_clauses): Update accordingly.
18927 (lower_rec_input_clauses): Likewise.
18929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18930 Alan Hayward <alan.hayward@arm.com>
18931 David Sherwood <david.sherwood@arm.com>
18933 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18934 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18935 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18936 (vect_analyze_slp_cost): Likewise.
18937 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18938 (vect_model_load_cost): Likewise.
18940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18941 Alan Hayward <alan.hayward@arm.com>
18942 David Sherwood <david.sherwood@arm.com>
18944 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18945 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18946 from an unsigned int * to a poly_uint64_pod *.
18947 (calculate_unrolling_factor): New function.
18948 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18951 Alan Hayward <alan.hayward@arm.com>
18952 David Sherwood <david.sherwood@arm.com>
18954 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18955 from an unsigned int to a poly_uint64.
18956 (_loop_vec_info::slp_unrolling_factor): Likewise.
18957 (_loop_vec_info::vectorization_factor): Change from an int
18959 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18960 (vect_get_num_vectors): New function.
18961 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18962 (vect_get_num_copies): Use vect_get_num_vectors.
18963 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18964 to an unsigned int *.
18965 (vect_analyze_data_refs): Change min_vf from an int * to a
18967 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18968 than an unsigned HOST_WIDE_INT.
18969 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18970 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18971 to an unsigned int *.
18972 (vect_analyze_data_ref_dependences): Likewise.
18973 (vect_compute_data_ref_alignment): Handle polynomial vf.
18974 (vect_enhance_data_refs_alignment): Likewise.
18975 (vect_prune_runtime_alias_test_list): Likewise.
18976 (vect_shift_permute_load_chain): Likewise.
18977 (vect_supportable_dr_alignment): Likewise.
18978 (dependence_distance_ge_vf): Take the vectorization factor as a
18979 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18980 (vect_analyze_data_refs): Change min_vf from an int * to a
18982 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18983 vfm1 as a poly_uint64 rather than an int. Make the same change
18984 for the returned bound_scalar.
18985 (vect_gen_vector_loop_niters): Handle polynomial vf.
18986 (vect_do_peeling): Likewise. Update call to
18987 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18988 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18990 * tree-vect-loop.c (vect_determine_vectorization_factor)
18991 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18992 (vect_get_known_peeling_cost): Likewise.
18993 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18994 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18995 (vect_transform_loop): Likewise. Use the lowest possible VF when
18996 updating the upper bounds of the loop.
18997 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18998 rather than an int.
18999 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19000 polynomial unroll factors.
19001 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19002 (vect_make_slp_decision): Likewise.
19003 (vect_supported_load_permutation_p): Likewise, and polynomial
19005 (vect_analyze_slp_cost): Handle polynomial vf.
19006 (vect_slp_analyze_node_operations): Likewise.
19007 (vect_slp_analyze_bb_1): Likewise.
19008 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19009 than an unsigned HOST_WIDE_INT.
19010 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19011 (vectorizable_load): Handle polynomial vf.
19012 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19014 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19017 Alan Hayward <alan.hayward@arm.com>
19018 David Sherwood <david.sherwood@arm.com>
19020 * match.pd: Handle bit operations involving three constants
19021 and try to fold one pair.
19023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19025 * tree-vect-loop-manip.c: Include gimple-fold.h.
19026 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19027 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19028 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19029 Add a path that uses a step of VF instead of 1, but disable it
19031 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19032 and niters_no_overflow parameters. Update calls to
19033 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19034 Create a new SSA name if the latter choses to use a ste other
19035 than zero, and return it via niters_vector_mult_vf_var.
19036 * tree-vect-loop.c (vect_transform_loop): Update calls to
19037 vect_do_peeling, vect_gen_vector_loop_niters and
19038 slpeel_make_loop_iterate_ntimes.
19039 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19040 (vect_gen_vector_loop_niters): Update declarations after above changes.
19042 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19044 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19045 128-bit round to integer instructions.
19046 (ceil<mode>2): Likewise.
19047 (btrunc<mode>2): Likewise.
19048 (round<mode>2): Likewise.
19050 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19052 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19053 unaligned VSX load/store on P8/P9.
19054 (expand_block_clear): Allow the use of unaligned VSX
19055 load/store on P8/P9.
19057 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19059 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19061 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19062 swap associated with both a load and a store.
19064 2018-01-02 Andrew Waterman <andrew@sifive.com>
19066 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19067 * config/riscv/riscv.md (clear_cache): Use it.
19069 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19071 * web.c: Remove out-of-date comment.
19073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19075 * expr.c (fixup_args_size_notes): Check that any existing
19076 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19077 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19078 (emit_single_push_insn): ...here.
19080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19082 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19083 (const_vector_encoded_nelts): New function.
19084 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19085 (const_vector_int_elt, const_vector_elt): Declare.
19086 * emit-rtl.c (const_vector_int_elt_1): New function.
19087 (const_vector_elt): Likewise.
19088 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19089 of CONST_VECTOR_ELT.
19091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19093 * expr.c: Include rtx-vector-builder.h.
19094 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19095 directly on the tree encoding.
19096 (const_vector_from_tree): Likewise.
19097 * optabs.c: Include rtx-vector-builder.h.
19098 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19099 sequence of "u" values.
19100 * vec-perm-indices.c: Include rtx-vector-builder.h.
19101 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19102 directly on the vec_perm_indices encoding.
19104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19106 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19107 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19108 * rtx-vector-builder.h: New file.
19109 * rtx-vector-builder.c: Likewise.
19110 * rtl.h (rtx_def::u2): Add a const_vector field.
19111 (CONST_VECTOR_NPATTERNS): New macro.
19112 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19113 (CONST_VECTOR_DUPLICATE_P): Likewise.
19114 (CONST_VECTOR_STEPPED_P): Likewise.
19115 (CONST_VECTOR_ENCODED_ELT): Likewise.
19116 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19117 (unwrap_const_vec_duplicate): Likewise.
19118 (const_vec_series_p): Check for a non-duplicated vector encoding.
19119 Say that the function only returns true for integer vectors.
19120 * emit-rtl.c: Include rtx-vector-builder.h.
19121 (gen_const_vec_duplicate_1): Delete.
19122 (gen_const_vector): Call gen_const_vec_duplicate instead of
19123 gen_const_vec_duplicate_1.
19124 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19125 (gen_const_vec_duplicate): Use rtx_vector_builder.
19126 (gen_const_vec_series): Likewise.
19127 (gen_rtx_CONST_VECTOR): Likewise.
19128 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19129 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19130 Build a new vector rather than modifying a CONST_VECTOR in-place.
19131 (handle_special_swappables): Update call accordingly.
19132 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19133 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19134 Build a new vector rather than modifying a CONST_VECTOR in-place.
19135 (handle_special_swappables): Update call accordingly.
19137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19139 * simplify-rtx.c (simplify_const_binary_operation): Use
19140 CONST_VECTOR_ELT instead of XVECEXP.
19142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19144 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19145 the selector elements to be different from the data elements
19146 if the selector is a VECTOR_CST.
19147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19148 ssizetype for the selector.
19150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19152 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19153 before testing each element individually.
19154 * tree-vect-generic.c (lower_vec_perm): Likewise.
19156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19158 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19159 * selftest-run-tests.c (selftest::run_tests): Call it.
19160 * vector-builder.h (vector_builder::operator ==): New function.
19161 (vector_builder::operator !=): Likewise.
19162 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19163 (vec_perm_indices::all_from_input_p): New function.
19164 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19165 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19166 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19167 instead of reading the VECTOR_CST directly. Detect whether both
19168 vector inputs are the same before constructing the vec_perm_indices,
19169 and update the number of inputs argument accordingly. Use the
19170 utility functions added above. Only construct sel2 if we need to.
19172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19174 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19175 the broadcast of the low byte.
19176 (expand_mult_highpart): Use an explicit encoding for the permutes.
19177 * optabs-query.c (can_mult_highpart_p): Likewise.
19178 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19179 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19180 (vectorizable_bswap): Likewise.
19181 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19182 explicit encoding for the power-of-2 permutes.
19183 (vect_permute_store_chain): Likewise.
19184 (vect_grouped_load_supported): Likewise.
19185 (vect_permute_load_chain): Likewise.
19187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19189 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19190 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19191 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19192 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19193 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19194 (vect_gen_perm_mask_any): Likewise.
19196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19198 * int-vector-builder.h: New file.
19199 * vec-perm-indices.h: Include int-vector-builder.h.
19200 (vec_perm_indices): Redefine as an int_vector_builder.
19201 (auto_vec_perm_indices): Delete.
19202 (vec_perm_builder): Redefine as a stand-alone class.
19203 (vec_perm_indices::vec_perm_indices): New function.
19204 (vec_perm_indices::clamp): Likewise.
19205 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19206 (vec_perm_indices::new_vector): New function.
19207 (vec_perm_indices::new_expanded_vector): Update for new
19208 vec_perm_indices class.
19209 (vec_perm_indices::rotate_inputs): New function.
19210 (vec_perm_indices::all_in_range_p): Operate directly on the
19211 encoded form, without computing elided elements.
19212 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19213 encoding. Update for new vec_perm_indices class.
19214 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19215 the given vec_perm_builder.
19216 (expand_vec_perm_var): Update vec_perm_builder constructor.
19217 (expand_mult_highpart): Use vec_perm_builder instead of
19218 auto_vec_perm_indices.
19219 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19220 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19221 or double series encoding as appropriate.
19222 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19223 vec_perm_indices instead of auto_vec_perm_indices.
19224 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19225 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19226 (vect_permute_store_chain): Likewise.
19227 (vect_grouped_load_supported): Likewise.
19228 (vect_permute_load_chain): Likewise.
19229 (vect_shift_permute_load_chain): Likewise.
19230 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19231 (vect_transform_slp_perm_load): Likewise.
19232 (vect_schedule_slp_instance): Likewise.
19233 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19234 (vectorizable_mask_load_store): Likewise.
19235 (vectorizable_bswap): Likewise.
19236 (vectorizable_store): Likewise.
19237 (vectorizable_load): Likewise.
19238 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19239 vec_perm_indices instead of auto_vec_perm_indices. Use
19240 tree_to_vec_perm_builder to read the vector from a tree.
19241 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19242 vec_perm_builder instead of a vec_perm_indices.
19243 (have_whole_vector_shift): Use vec_perm_builder and
19244 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19245 truncation to calc_vec_perm_mask_for_shift.
19246 (vect_create_epilog_for_reduction): Likewise.
19247 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19248 from auto_vec_perm_indices to vec_perm_indices.
19249 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19250 instead of changing individual elements.
19251 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19252 the vector in d.perm.
19253 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19254 from auto_vec_perm_indices to vec_perm_indices.
19255 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19256 instead of changing individual elements.
19257 (arm_vectorize_vec_perm_const): Use new_vector to install
19258 the vector in d.perm.
19259 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19260 Update vec_perm_builder constructor.
19261 (rs6000_expand_interleave): Likewise.
19262 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19263 (rs6000_expand_interleave): Likewise.
19265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19267 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19268 to qimode could truncate the indices.
19269 * optabs.c (expand_vec_perm_var): Likewise.
19271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19273 * Makefile.in (OBJS): Add vec-perm-indices.o.
19274 * vec-perm-indices.h: New file.
19275 * vec-perm-indices.c: Likewise.
19276 * target.h (vec_perm_indices): Replace with a forward class
19278 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19279 * optabs.h: Include vec-perm-indices.h.
19280 (expand_vec_perm): Delete.
19281 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19282 (expand_vec_perm_const): Declare.
19283 * target.def (vec_perm_const_ok): Replace with...
19284 (vec_perm_const): ...this new hook.
19285 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19286 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19287 * doc/tm.texi: Regenerate.
19288 * optabs.def (vec_perm_const): Delete.
19289 * doc/md.texi (vec_perm_const): Likewise.
19290 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19291 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19292 expand_vec_perm for constant permutation vectors. Assert that
19293 the mode of variable permutation vectors is the integer equivalent
19294 of the mode that is being permuted.
19295 * optabs-query.h (selector_fits_mode_p): Declare.
19296 * optabs-query.c: Include vec-perm-indices.h.
19297 (selector_fits_mode_p): New function.
19298 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19299 is defined, instead of checking whether the vec_perm_const_optab
19300 exists. Use targetm.vectorize.vec_perm_const instead of
19301 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19302 fit in the vector mode before using a variable permute.
19303 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19304 vec_perm_indices instead of an rtx.
19305 (expand_vec_perm): Replace with...
19306 (expand_vec_perm_const): ...this new function. Take the selector
19307 as a vec_perm_indices rather than an rtx. Also take the mode of
19308 the selector. Update call to shift_amt_for_vec_perm_mask.
19309 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19310 Use vec_perm_indices::new_expanded_vector to expand the original
19311 selector into bytes. Check whether the indices fit in the vector
19312 mode before using a variable permute.
19313 (expand_vec_perm_var): Make global.
19314 (expand_mult_highpart): Use expand_vec_perm_const.
19315 * fold-const.c: Includes vec-perm-indices.h.
19316 * tree-ssa-forwprop.c: Likewise.
19317 * tree-vect-data-refs.c: Likewise.
19318 * tree-vect-generic.c: Likewise.
19319 * tree-vect-loop.c: Likewise.
19320 * tree-vect-slp.c: Likewise.
19321 * tree-vect-stmts.c: Likewise.
19322 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19324 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19325 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19326 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19327 (aarch64_vectorize_vec_perm_const): ...this new function.
19328 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19329 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19330 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19331 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19332 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19333 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19334 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19336 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19337 check for NEON modes.
19338 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19339 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19340 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19341 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19343 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19344 the old VEC_PERM_CONST conditions.
19345 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19346 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19347 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19348 (ia64_vectorize_vec_perm_const_ok): Merge into...
19349 (ia64_vectorize_vec_perm_const): ...this new function.
19350 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19351 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19352 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19353 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19354 * config/mips/mips.c (mips_expand_vec_perm_const)
19355 (mips_vectorize_vec_perm_const_ok): Merge into...
19356 (mips_vectorize_vec_perm_const): ...this new function.
19357 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19358 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19359 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19360 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19361 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19362 (rs6000_expand_vec_perm_const): Delete.
19363 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19365 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19366 (altivec_expand_vec_perm_const_le): Take each operand individually.
19367 Operate on constant selectors rather than rtxes.
19368 (altivec_expand_vec_perm_const): Likewise. Update call to
19369 altivec_expand_vec_perm_const_le.
19370 (rs6000_expand_vec_perm_const): Delete.
19371 (rs6000_vectorize_vec_perm_const_ok): Delete.
19372 (rs6000_vectorize_vec_perm_const): New function.
19373 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19374 an element count and rtx array.
19375 (rs6000_expand_extract_even): Update call accordingly.
19376 (rs6000_expand_interleave): Likewise.
19377 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19378 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19379 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19380 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19381 (rs6000_expand_vec_perm_const): Delete.
19382 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19383 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19384 (altivec_expand_vec_perm_const_le): Take each operand individually.
19385 Operate on constant selectors rather than rtxes.
19386 (altivec_expand_vec_perm_const): Likewise. Update call to
19387 altivec_expand_vec_perm_const_le.
19388 (rs6000_expand_vec_perm_const): Delete.
19389 (rs6000_vectorize_vec_perm_const_ok): Delete.
19390 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19391 reference to the SPE evmerge intructions.
19392 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19393 an element count and rtx array.
19394 (rs6000_expand_extract_even): Update call accordingly.
19395 (rs6000_expand_interleave): Likewise.
19396 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19397 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19401 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19403 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19404 vector mode and that that mode matches the mode of the data
19406 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19407 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19408 directly using expand_vec_perm_1 when forcing selectors into
19410 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19414 * optabs-query.h (can_vec_perm_p): Delete.
19415 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19416 * optabs-query.c (can_vec_perm_p): Split into...
19417 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19418 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19419 particular selector is valid.
19420 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19421 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19422 (vect_grouped_load_supported): Likewise.
19423 (vect_shift_permute_load_chain): Likewise.
19424 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19425 (vect_transform_slp_perm_load): Likewise.
19426 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19427 (vectorizable_bswap): Likewise.
19428 (vect_gen_perm_mask_checked): Likewise.
19429 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19430 implementations of variable permutation vectors into account
19431 when deciding which selector to use.
19432 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19433 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19434 with a false third argument.
19435 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19436 to test whether the constant selector is valid and can_vec_perm_var_p
19437 to test whether a variable selector is valid.
19439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19441 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19442 * optabs-query.c (can_vec_perm_p): Likewise.
19443 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19444 instead of vec_perm_indices.
19445 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19446 (vect_gen_perm_mask_checked): Likewise,
19447 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19448 (vect_gen_perm_mask_checked): Likewise,
19450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19452 * optabs-query.h (qimode_for_vec_perm): Declare.
19453 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19454 (qimode_for_vec_perm): ...this new function.
19455 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19457 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19459 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19460 does not have a conditional at the top.
19462 2018-01-02 Richard Biener <rguenther@suse.de>
19464 * ipa-inline.c (big_speedup_p): Fix expression.
19466 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19469 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19472 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19476 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19477 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19478 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19479 cond_taken_branch_cost 3->4.
19481 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19483 PR tree-optimization/83581
19484 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19485 TODO_cleanup_cfg if any changes have been made.
19487 PR middle-end/83608
19488 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19489 convert_modes if target mode has the right side, but different mode
19492 PR middle-end/83609
19493 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19494 last argument when extracting from CONCAT. If either from_real or
19495 from_imag is NULL, use expansion through memory. If result is not
19496 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19497 the parts directly to inner mode, if even that fails, use expansion
19500 PR middle-end/83623
19501 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19502 check for bswap in mode rather than HImode and use that in expand_unop
19505 Copyright (C) 2018 Free Software Foundation, Inc.
19507 Copying and distribution of this file, with or without modification,
19508 are permitted in any medium without royalty provided the copyright
19509 notice and this notice are preserved.