1 2018-07-12 Richard Biener <rguenther@suse.de>
4 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5 Remove -mieee-fp handling.
7 2018-07-12 Richard Biener <rguenther@suse.de>
9 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10 left-over from last patch.
12 2018-07-12 Jakub Jelinek <jakub@redhat.com>
14 PR tree-optimization/86492
15 * gimple-ssa-store-merging.c
16 (imm_store_chain_info::coalesce_immediate_stores): Call
17 check_no_overlap even for the merge_overlapping case. Formatting fix.
19 2018-07-12 Richard Biener <rguenther@suse.de>
22 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
23 move possibly trapping operations into the conditional.
25 2018-07-12 Richard Biener <rguenther@suse.de>
27 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
28 (vn_lookup_simplify_result): Remove recursion limit applied
30 (vn_nary_build_or_lookup_1): Adjust.
31 (try_to_simplify): Likewise.
32 * gimple-match-head.c (gimple_resimplify1): Instead apply one
34 (gimple_resimplify2): Likewise.
35 (gimple_resimplify3): Likewise.
36 (gimple_resimplify4): Likewise.
38 2018-07-11 Jakub Jelinek <jakub@redhat.com>
40 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
41 Use __mmask64 type instead of __mmask8 for __M argument.
42 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
43 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
45 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
46 __mmask16 for __M argument.
47 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
48 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
49 to __mmask16 instead of __mmask8.
50 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
51 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
52 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
53 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
54 instead of __mmask16 for __U argument.
55 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
56 __mmask16 instead of __mmask8 for __U argument.
57 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
59 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
61 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
63 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
65 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
67 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
69 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
71 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
73 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
75 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
77 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
78 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
79 return type as well as __M argument type and all casts from __mmask8
81 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
82 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
83 return type as well as __M argument type and all casts from __mmask8
85 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
86 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
87 return type as well as __M argument type and all casts from __mmask8
89 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
90 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
91 return type as well as __M argument type and all casts from __mmask8
93 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
94 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
97 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
99 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
100 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
103 2018-07-11 Paul Koning <ni1d@arrl.net>
105 * doc/md.texi (define_subst): Document how multiple occurrences of
106 the same argument in the replacement pattern are handled.
108 2018-07-11 Paul Koning <ni1d@arrl.net>
110 * doc/extend.texi (Common Variable Attributes): Move "mode" into
112 (Common Type Attributes): Add "mode" attribute.
114 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
116 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
117 stream DECL_ORIGINAL_TYPE.
118 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
119 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
120 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
121 Do not walk original type.
122 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
124 (write_ts_decl_non_common_tree_pointers): Do not stream
126 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
127 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
129 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
131 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
132 thread twice from the same starting edge.
134 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
136 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
137 * gimple.c (gimple_call_nonnull_result_p): ...here...
138 (gimple_call_nonnull_arg): ...and here.
139 * gimple.h (gimple_call_nonnull_result_p): New.
140 (gimple_call_nonnull_arg): New.
142 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
144 * config/arm/arm-cpus.in: Move information from fpu field of each
145 cpu definition to the isa field.
146 * config/arm/parsecpu.awk (fpu): Delete match rule.
147 (gen_comm_data): Don't add bits from the CPU's FPU entry.
149 2018-07-11 Richard Biener <rguenther@suse.de>
152 * dwarf2out.c (init_sections_and_labels): Use
153 output_asm_line_debug_info consistently.
154 (dwarf2out_early_finish): Likewise.
155 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
158 2018-07-11 Richard Biener <rguenther@suse.de>
160 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
161 Rework father_bb setting in a way to avoid propagating constants
162 multiple times on a loop body.
164 2018-07-10 Mark Wielaard <mark@klomp.org>
167 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
169 2018-07-10 Richard Biener <rguenther@suse.de>
171 * hash-map.h (hash_map::iterator::operator*): Return
172 references to key and value.
174 2018-07-10 Jakub Jelinek <jakub@redhat.com>
177 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
178 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
179 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
180 (gimplify_omp_for): For composite loops, move outer
181 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
182 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
183 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
184 TREE_LIST for both the original class iterator and the "last" helper
185 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
186 loop, remember has_decl_expr from outer composite loops for the
187 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
189 2018-07-09 Martin Sebor <msebor@redhat.com>
193 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
194 accessing implicitly initialized array elements.
195 * expr.c (string_constant): Handle string initializers of
196 character arrays within aggregates.
197 * gimple-fold.c (fold_array_ctor_reference): Add argument.
198 Store element offset. As a special case, handle zero size.
199 (fold_nonarray_ctor_reference): Same.
200 (fold_ctor_reference): Add argument. Store subobject offset.
201 * gimple-fold.h (fold_ctor_reference): Add argument.
203 2018-07-09 Paul Koning <ni1d@arrl.net>
205 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
206 (pdp11_insn_cost): New function.
207 (pdp11_md_asm_adjust): New function.
208 (TARGET_INVALID_WITHIN_DOLOOP): Define.
209 (pdp11_rtx_costs): Update to match machine better.
210 (output_addr_const_pdp11): Correct format mismatch warnings.
211 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
212 * config/pdp11/pdp11.md: General change to add base_cost and/or
213 length attributes for use by new pdp11_insn_cost function.
214 (MIN_BRANCH): Correct definition.
216 (doloop_end): Use standard pattern name for looping pattern.
217 (doloop_end_nocc): New.
218 (movsf): Add another constraint alternative.
219 (zero_extendqihi2): Add constraint alternatives for not in place
221 (zero_extendhisi2): Remove.
222 (shift patterns): Add CC handling variants.
226 (define_peephole2): New peephole to recognize mov that sets CC for
229 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
231 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
232 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
235 2018-07-09 Jakub Jelinek <jakub@redhat.com>
238 * real.c (real_nextafter): Return true if result is denormal.
240 2018-07-09 Martin Liska <mliska@suse.cz>
242 * common.opt: Add back wrongly removed attribute.
244 2018-07-09 Richard Biener <rguenther@suse.de>
247 * dwarf2out.c (gen_block_die): For an early generated DIE
248 always output high/low PC attributes.
250 2018-07-09 Tom de Vries <tdevries@suse.de>
252 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
253 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
254 onto VAR_DECL with abstract origin.
256 2018-07-07 Jim Wilson <jimw@sifive.com>
258 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
260 2018-07-07 Tom de Vries <tdevries@suse.de>
262 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
264 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
266 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
268 (combine_bound): Use wide-int overflow calculation instead of
270 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
272 * fold-const.c (int_const_binop_2): Same.
273 (extract_muldiv_1): Same.
274 (fold_div_compare): Same.
275 (fold_abs_const): Same.
277 * poly-int.h (add): Same.
281 * predict.c (predict_iv_comparison): Same.
282 * profile-count.c (slow_safe_scale_64bit): Same.
283 * simplify-rtx.c (simplify_const_binary_operation): Same.
284 * tree-chrec.c (tree_fold_binomial): Same.
285 * tree-data-ref.c (split_constant_offset_1): Same.
286 * tree-if-conv.c (idx_within_array_bound): Same.
287 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
288 * tree-ssa-phiopt.c (minmax_replacement): Same.
289 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
290 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
291 * vr-values.c (vr_values::adjust_range_with_scev): Same.
292 * wide-int.cc (wi::add_large): Same.
293 (wi::mul_internal): Same.
294 (wi::sub_large): Same.
295 (wi::divmod_internal): Same.
296 * wide-int.h: Change overflow type to overflow_type for neg, add,
297 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
298 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
299 mul_internal, divmod_internal.
300 (overflow_type): New enum.
301 (accumulate_overflow): New.
303 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
305 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
306 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
308 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
310 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
311 argument is checked for zero before entering loop, avoid checking again.
313 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
315 * gimplify.h (generic_expr_could_trap_p): Set as global function.
316 * gimplify.h (generic_expr_could_trap_p): Likwise.
317 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
319 2018-07-06 Jakub Jelinek <jakub@redhat.com>
321 PR tree-optimization/86401
322 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
323 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
324 (fold_bit_and_mask): ... here. New helper function for match.pd.
325 * fold-const.h (fold_bit_and_mask): Declare.
326 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
328 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
331 * target.def (translate_mode_attribute): New hook.
332 * targhooks.h (default_translate_mode_attribute): Declare.
333 * targhooks.c (default_translate_mode_attribute): New function.
334 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
335 * doc/tm.texi: Regenerate.
336 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
337 (rs6000_translate_mode_attribute): New function.
339 2018-07-06 Paul Koning <ni1d@arrl.net>
341 * doc/md.texi (define_split): Document DONE and FAIL.
342 (define_peephole2): Ditto.
344 2018-07-05 Jeff Law <law@redhat.com>
346 PR tree-optimization/86010
347 * tree-ssa-dse.c (compute_trims): More aggressively trim at
348 both the head and tail of mem* and str* calls.
350 2018-07-05 Jim Wilson <jimw@sifive.com>
352 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
354 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
356 * config/aarch64/aarch64-simd.md: correct flags text for
357 MIN_EXPR replacement.
359 2018-07-05 James Clarke <jrtc27@jrtc27.com>
361 * configure: Regenerated.
363 2018-07-05 Carl Love <cel@us.ibm.com>
365 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
366 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
367 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
368 VSX_BUILTIN_DOUBLEL_V4SF.
370 2018-07-05 Martin Sebor <msebor@redhat.com>
373 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
374 than its domain to compute its the upper bound of a char array.
376 2018-07-05 Nathan Sidwell <nathan@acm.org>
378 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
379 * doc/cpp.texi: Update comment.
380 * doc/tm.texi: Rebuilt.
381 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
382 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
383 * doc/extend.texi (Backwards Compatibility): Clarify it is system
384 headers affected by extern "C".
385 * system.h: Poison NO_IMPLICIT_EXTERN_C.
386 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
387 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
388 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
389 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
390 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
391 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
392 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
393 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
394 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
395 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
396 config/sparc/sp64-elf.h, config/spu/spu.h,
397 config/stormy16/stormy16.h, config/v850/v850.h,
398 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
399 define NO_IMPLICIT_EXTERN_C.
400 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
402 2018-07-05 Tamar Christina <tamar.christina@arm.com>
405 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
406 instead of GET_MODE_SIZE when comparing Units.
408 2018-07-05 Tamar Christina <tamar.christina@arm.com>
411 * rtlanal.c (set_noop_p): Constrain on mode change,
412 include hard-reg-set.h
414 2018-07-05 Tamar Christina <tamar.christina@arm.com>
416 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
418 2018-07-05 Jakub Jelinek <jakub@redhat.com>
421 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
424 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
426 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
428 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
431 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
433 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
435 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
438 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
440 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
442 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
445 2018-07-04 Martin Liska <mliska@suse.cz>
447 * tree-switch-conversion.c: Define
448 max_ratio_for_speed and max_ratio_for_size constants.
450 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
451 Martin Liska <mliska@suse.cz>
456 * common.opt: Rename align options with 'str_' prefix.
457 * common/config/i386/i386-common.c (set_malign_value): New
459 (ix86_handle_option): Use it to set -falign-* options/
460 * config/aarch64/aarch64-protos.h (struct tune_params): Change
461 type from int to string.
462 * config/aarch64/aarch64.c: Update default values from int
464 * config/alpha/alpha.c (alpha_override_options_after_change):
466 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
467 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
468 max skip conditionally.
469 * config/i386/freebsd.h (SUBALIGN_LOG): New.
470 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
471 max skip conditionally.
472 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
473 max skip conditionally.
474 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
475 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
476 max skip conditionally.
477 * config/i386/i386.c (struct ptt): Change type from int to
479 (ix86_default_align): Set default values.
480 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
481 max skip conditionally.
482 * config/i386/iamcu.h (SUBALIGN_LOG): New.
483 (ASM_OUTPUT_MAX_SKIP_ALIGN):
484 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
485 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
486 max skip conditionally.
487 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
488 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
489 * config/i386/x86-64.h (SUBALIGN_LOG): New.
490 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
491 max skip conditionally.
492 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
493 * config/ia64/ia64.c (ia64_option_override): Set default values
494 for alignment options.
495 * config/m68k/m68k.c: Handle new str_align_* options.
496 * config/mips/mips.c (mips_set_compression_mode): Change
498 (mips_option_override): Set default values for options.
499 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
501 * config/rs6000/rs6000.c (rs6000_option_override_internal):
503 * config/rx/rx.c (rx_option_override): Likewise.
504 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
505 (LABEL_ALIGN): Use align_labels_log.
506 (LOOP_ALIGN): Use align_loops_align.
507 * config/s390/s390.c (s390_asm_output_function_label): Use new
509 * config/sh/sh.c (sh_override_options_after_change):
510 Change type of constants.
511 * config/spu/spu.c (spu_sched_init): Likewise.
512 * config/sparc/sparc.c (sparc_option_override): Set default
514 * config/visium/visium.c (visium_option_override): Likewise.
515 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
516 emit p2align format with last argument if it's not needed.
517 * doc/invoke.texi: Document extended format of -falign-*.
518 * final.c: Use align_labels alignment.
519 * flags.h (struct target_flag_state): Change type to use
521 (struct align_flags_tuple): New.
522 (struct align_flags): Likewise.
523 (align_loops_log): Redefine macro to use new types.
524 (align_loops_max_skip): Redefine macro to use new types.
525 (align_jumps_log): Redefine macro to use new types.
526 (align_jumps_max_skip): Redefine macro to use new types.
527 (align_labels_log): Redefine macro to use new types.
528 (align_labels_max_skip): Redefine macro to use new types.
529 (align_functions_log): Redefine macro to use new types.
530 (align_loops): Redefine macro to use new types.
531 (align_jumps): Redefine macro to use new types.
532 (align_labels): Redefine macro to use new types.
533 (align_functions): Redefine macro to use new types.
534 (align_functions_max_skip): Redefine macro to use new types.
535 (align_loops_value): New macro.
536 (align_jumps_value): New macro.
537 (align_labels_value): New macro.
538 (align_functions_value): New macro.
539 * function.c (invoke_set_current_function_hook): Propagate
540 alignment values from flags to global variables default in
542 * ipa-icf.c (sem_function::equals_wpa): Use
543 cl_optimization_option_eq instead of memcmp.
544 * lto-streamer.h (cl_optimization_stream_out): Support streaming
546 (cl_optimization_stream_in): Likewise.
547 * optc-save-gen.awk: Support strings in cl_optimization.
548 * opth-gen.awk: Likewise.
549 * opts.c (finish_options): Remove error checking of invalid
551 (MAX_CODE_ALIGN): Remove.
552 (MAX_CODE_ALIGN_VALUE): Likewise.
553 (parse_and_check_align_values): New function.
554 (check_alignment_argument): Likewise.
555 (common_handle_option): Use check_alignment_argument.
556 * opts.h (parse_and_check_align_values): Declare.
557 * toplev.c (init_alignments): Remove.
558 (read_log_maxskip): New.
559 (parse_N_M): Likewise.
560 (parse_alignment_opts): Likewise.
561 (backend_init_target): Remove usage of init_alignments.
562 * toplev.h (parse_alignment_opts): Declare.
563 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
565 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
566 * tree.c (cl_option_hasher::equal): New.
567 * varasm.c: Use new global macros.
569 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
571 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
572 Use a simpler align directive also if MAXSKIP = ALIGN-1.
573 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
574 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
575 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
576 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
577 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
578 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
579 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
580 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
581 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
582 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
583 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
585 2018-07-04 Martin Liska <mliska@suse.cz>
586 Jonathan Wakely <jwakely@redhat.com>
588 * coverage.c: Use correct type.
589 * doc/invoke.texi: Language correction.
591 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
594 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
595 ENDBRANCH for non-tail call which may return via indirect branch.
596 * doc/extend.texi: Document indirect_return attribute.
598 2018-07-03 Martin Sebor <msebor@redhat.com>
600 PR tree-optimization/86274
601 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
603 (format_floating): Correct handling of infinities and NaNs.
605 2018-07-03 Martin Sebor <msebor@redhat.com>
607 * print-tree.c (print_real_cst): New function.
608 (print_node_brief): Call it.
611 2018-07-03 Jeff Law <law@redhat.com>
613 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
614 into a single pattern.
616 * config/h8300/h8300.md (ors code_iterator): New.
617 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
618 a single pattern and single splitter.
619 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
620 (iorqi3_1, xorqi3_1): Likewise.
621 (iorqi3, xorqi3 expanders): Similarly.
623 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
624 (movmd_internal) into a single pattern using the P mode iterator.
625 (movmd splitters): Similarly.
626 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
627 (movsd splitters): Similarly.
629 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
630 ADDB, ADDW and ADDL into a single ADD attribute which selects the
631 right table based on the size of the operand.
632 * config/h8300/h8300.md (length_table): Corresponding changes. All
633 references to "addb", "addw" and "addl" changed to "add".
634 (btst patterns): Merge two variants into a single pattern.
635 (tstqi, tsthi): Likewise.
636 (addhi3_incdec, addsi3_incdec): Likewise.
637 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
638 (mulhi3, mulsi3): Likewise.
639 (udivhi3, udivsi3): Likewise.
640 (divhi3, divsi3): Likewise.
641 (andorqi3, andorhi3, andorsi3): Likewise.
643 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
646 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
647 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
649 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
651 PR tree-optimization/85694
652 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
653 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
655 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
656 (<u>avg<mode>3_ceil): New patterns.
658 2018-07-03 David Malcolm <dmalcolm@redhat.com>
660 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
661 scan-tree-dump directive.
662 * gcc.dg/vect/slp-perm-2.c: Likewise.
663 * gcc.dg/vect/slp-perm-3.c: Likewise.
664 * gcc.dg/vect/slp-perm-5.c: Likewise.
665 * gcc.dg/vect/slp-perm-6.c: Likewise.
666 * gcc.dg/vect/slp-perm-7.c: Likewise.
667 * gcc.dg/vect/slp-perm-8.c: Likewise.
669 2018-07-03 Marek Polacek <polacek@redhat.com>
672 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
674 2018-07-03 Richard Biener <rguenther@suse.de>
677 * tree-ssa-structalias.c (find_func_clobbers): Properly
678 handle indirect calls.
680 2018-07-03 Jeff Law <law@redhat.com>
682 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
683 (shifts): New code iterator.
684 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
685 expander. Fix HImode handling on H8/SX.
686 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
687 (subqi3, subhi3, subsi3 expanders): Likewise.
688 (andqi3, andhi3, andsi3 expanders): Likewise.
689 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
690 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
691 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
692 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
693 (zero_extendqihi2, zero_extendqisi2): Likewise.
694 (extendqihi2, extendqisi2): Likewise.
695 (rotlqi3, rotlhi3, rotlsi3): Likewise.
696 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
697 (rotlqi3_1, rotlhi3_1): Likewise.
698 (logicalhi3_sn, logicalsi3_sn): Likewise.
699 (logicalhi3, logicalsi3): Likewise.
701 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
703 * tree-vect-patterns.c (vect_recog_rotate_pattern)
704 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
705 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
706 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
707 type to append_pattern_def_seq instead of creating a stmt_vec_info
709 (build_mask_conversion): Likewise. Remove vinfo argument.
710 (vect_add_conversion_to_patterm): Likewise, renaming to...
711 (vect_add_conversion_to_pattern): ...this.
712 (vect_recog_mask_conversion_pattern): Update call to
713 build_mask_conversion. Pass the vector type to
714 append_pattern_def_seq here too.
715 (vect_recog_gather_scatter_pattern): Update call to
716 vect_add_conversion_to_pattern.
718 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
720 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
721 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
722 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
723 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
724 STMT_VINFO_PATTERN_DEF_SEQ to null here.
725 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
726 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
727 append_pattern_def_seq instead of new_pattern_def_seq.
728 (vect_recog_divmod_pattern): Do both of the above.
729 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
732 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
734 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
735 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
736 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
737 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
738 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
739 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
740 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
741 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
742 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
743 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
744 parameter with a single stmt_vec_info.
745 (vect_recog_func_ptr): Likewise.
746 (vect_recog_gather_scatter_pattern): Likewise, folding in...
747 (vect_try_gather_scatter_pattern): ...this.
748 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
749 the stmt_vec_info of the statement to be matched. Don't clear
750 STMT_VINFO_RELATED_STMT.
751 (vect_pattern_recog): Update call accordingly.
753 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
755 PR tree-optimization/85694
756 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
757 (uavgM3_ceil): Document new optabs.
758 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
759 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
761 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
762 (savg_ceil_optab): New optabs.
763 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
764 (vect_vect_recog_func_ptrs): Add it.
765 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
766 constant directly from the associated lhs.
768 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
770 * tree-vect-patterns.c (vect_split_statement): New function.
771 (vect_convert_input): Use it to try to split an existing cast.
773 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
775 * poly-int.h (print_hex): New function.
776 * dumpfile.h (dump_dec, dump_hex): Declare.
777 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
778 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
779 min_input_precision, operation_precision and operation_sign.
780 * tree-vect-patterns.c (vect_get_range_info): New function.
781 (vect_same_loop_or_bb_p, vect_single_imm_use)
782 (vect_operation_fits_smaller_type): Delete.
783 (vect_look_through_possible_promotion): Add an optional
784 single_use_p parameter.
785 (vect_recog_over_widening_pattern): Rewrite to use new
786 stmt_vec_info infomration. Handle one operation at a time.
787 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
788 (vect_truncatable_operation_p, vect_set_operation_type)
789 (vect_set_min_input_precision): New functions.
790 (vect_determine_min_output_precision_1): Likewise.
791 (vect_determine_min_output_precision): Likewise.
792 (vect_determine_precisions_from_range): Likewise.
793 (vect_determine_precisions_from_users): Likewise.
794 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
795 (vect_vect_recog_func_ptrs): Put over_widening first.
797 (vect_pattern_recog): Call vect_determine_precisions.
799 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
801 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
802 statements that have been replaced by further pattern statements.
803 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
805 2018-07-03 Richard Biener <rguenther@suse.de>
807 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
808 always set *dt. Dump vectype in vectype overload.
809 * dumpfile.h (dump_gimple_expr): New function.
810 (dump_gimple_expr_loc): Likewise.
811 * dumpfile.c (dump_gimple_expr): New function.
812 (dump_gimple_expr_loc): Likewise.
814 2018-07-02 Jeff Law <law@redhat.com>
816 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
817 the H8/300, H8/300H and H8/S variants into a single pattern.
818 (movhi_h8300, movqi_h8300hs): Similarly.
819 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
820 (QHI mode iterator): New.
822 * config/h8300/h8300.md: Remove trailing whitespace.
824 2018-07-02 Jim Wilson <jimw@sifive.com>
826 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
827 instead of emit_insn for interrupt returns.
828 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
829 (riscv_sret, riscv_uret): Likewise.
831 2018-07-02 David Malcolm <dmalcolm@redhat.com>
833 * pretty-print.c (selftest::test_pp_format): Move save and restore
834 of quotes to class auto_fix_quotes, and add an instance.
835 * selftest.c: Include "intl.h".
836 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
837 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
838 * selftest.h (selftest::auto_fix_quotes): New class.
840 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
842 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
843 (aarch64_sve_prepare_conditional_op): Remove.
844 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
845 Allow aarch64_simd_reg_or_zero as select operand; remove
846 the aarch64_sve_prepare_conditional_op call.
847 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
848 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
849 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
850 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
851 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
852 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
853 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
854 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
855 and a splitters to match all of the *_any patterns.
856 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
858 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
859 (SVE_COND_FP_BINARY_REV): Remove.
860 (sve_int_op_rev, sve_fp_op_rev): New.
861 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
862 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
863 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
864 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
865 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
866 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
867 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
868 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
869 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
871 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
872 Remove match_dup 1 from the inner unspec.
873 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
875 * config/aarch64/aarch64.md (movprfx): New attr.
876 (length): Default movprfx to 8.
877 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
878 (*madd<SVE_I>, *msub<SVE_I): Likewise.
879 (*<su>mul<SVE_I>3_highpart): Likewise.
880 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
881 (*v<ASHIFT><SVE_I>3): Likewise.
882 (*<su><MAXMIN><SVE_I>3): Likewise.
883 (*<su><MAXMIN><SVE_F>3): Likewise.
884 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
885 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
886 (*div<SVE_F>4): Likewise.
888 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
890 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
893 2018-07-02 Richard Biener <rguenther@suse.de>
895 PR tree-optimization/86363
896 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
897 memset argument refers to a non-variable address.
899 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
901 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
902 {PLUS,MINUS}_EXPR code to...
903 (adjust_symbolic_bound): ...here,
904 (combine_bound): ...here,
905 (set_value_range_with_overflow): ...and here.
907 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
909 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
911 (extract_range_from_abs_expr): ...here.
913 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
915 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
916 -fno-omit-frame-pointer when not optimizing.
918 2018-07-02 Martin Liska <mliska@suse.cz>
921 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
922 (propagate_nothrow): Likewise.
924 2018-07-02 Martin Liska <mliska@suse.cz>
927 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
929 2018-07-02 David Malcolm <dmalcolm@redhat.com>
931 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
932 function in r262149, changing "loc" param from source_location to
933 const dump_location_t &.
934 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
935 declaration, as above.
937 2018-07-01 Paul Koning <ni1d@arrl.net>
939 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
940 -munit-asm, -mgnu-asm, -mdec-asm.
941 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
942 (pdp11_output_labelref): New.
943 (pdp11_output_def): New.
944 (pdp11_output_addr_vec_elt): New.
945 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
946 %# and %@ format codes.
947 (pdp11_option_override): New.
948 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
949 (pdp11_output_ident): New.
950 (pdp11_asm_named_section): New.
951 (pdp11_asm_init_sections): New.
952 (pdp11_file_start): New.
953 (pdp11_file_end): New.
954 (output_ascii): Use .ascii/.asciz for -mdec-asm.
955 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
956 %o, like %c but octal.
957 (pdp11_option_override): New.
958 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
960 (DATA_SECTION_ASM_OP): Ditto.
961 (READONLY_DATA_SECTION_ASM_OP): New.
962 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
963 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
964 (ASM_OUTPUT_LABELREF): Ditto.
965 (ASM_OUTPUT_DEF): Ditto.
966 (ASM_OUTPUT_EXTERNAL): New.
967 (ASM_OUTPUT_SOURCE_FILENAME): New.
968 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
969 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
970 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
971 %# and %@ format codes.
972 * config/pdp11/pdp11.opt (mgnu-asm): New.
973 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
974 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
975 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
977 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
979 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
980 dereferencing path[] beyond its length.
982 (debug_all_paths): New.
983 (rewire_first_differing_edge): New.
984 (adjust_paths_after_duplication): New.
985 (duplicate_thread_path): Call adjust_paths_after_duplication.
987 (thread_through_all_blocks): Add new argument to
988 duplicate_thread_path.
990 2018-06-30 Jim Wilson <jimw@sifive.com>
992 * config/riscv/predicates.md (p2m1_shift_operand): New.
993 (high_mask_shift_operand): New.
994 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
995 pattern using p2m1_shift_operand.
996 (lshsi3_zero_extend_3+2): New combiner pattern using
997 high_mask_shift_operand.
999 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1001 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1003 (vect_recog_rotate_pattern): ...here.
1004 (vect_convert_input): Try to insert casts of invariants in the
1006 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1007 preheader to be empty.
1009 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1011 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1012 vector type. If given, install it in the new statement's
1014 (vect_element_precision): New function.
1015 (vect_unpromoted_value): New struct.
1016 (vect_unpromoted_value::vect_unpromoted_value): New function.
1017 (vect_unpromoted_value::set_op): Likewise.
1018 (vect_look_through_possible_promotion): Likewise.
1019 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1020 (vect_widened_op_tree, vect_convert_input): Likewise.
1021 (vect_convert_inputs, vect_convert_output): Likewise.
1022 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1023 to handle the optional cast of the multiplication result and
1024 vect_widened_op_tree to detect the widened multiplication itself.
1025 Do not require the input and output of promotion casts to have
1026 the same sign, but base the signedness of the operation on the
1027 input rather than the result. If the pattern includes two
1028 promotions, check that those promotions have the same sign.
1029 Do not restrict the MULT_EXPR handling to a double-width result;
1030 handle quadruple-width results and wider. Use vect_convert_inputs
1031 to convert the inputs to the common type.
1032 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1033 to handle the optional cast of the ABS result. Also allow a sign
1034 change or a sign extension between the ABS and MINUS.
1035 Use vect_widened_op_tree to detect the widened subtraction and use
1036 vect_convert_inputs to convert the inputs to the common type.
1037 (vect_handle_widen_op_by_const): Delete.
1038 (vect_recog_widen_op_pattern): New function.
1039 (vect_recog_widen_mult_pattern): Use it.
1040 (vect_recog_widen_shift_pattern): Likewise.
1041 (vect_recog_widen_sum_pattern): Use
1042 vect_look_through_possible_promotion to handle the promoted
1045 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1047 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1048 the containing gimple_seq *.
1049 * gimple-iterator.h (gsi_for_stmt): Declare it.
1050 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1051 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1052 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1053 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1054 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1056 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1058 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1059 statement being replaced is part of an existing pattern
1060 definition sequence, inserting the new pattern statements before
1062 (vect_pattern_recog_1): Don't return a bool. If the statement
1063 is already part of a pattern, instead apply pattern matching
1064 to the pattern definition statements. Don't clear the
1065 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1066 (vect_pattern_recog): Don't break after the first match;
1067 continue processing the pattern definition statements instead.
1068 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1070 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1072 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1073 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1074 (vect_recog_widen_sum_pattern): Use it.
1076 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1078 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1079 phi is not a pattern statement and has not been replaced by
1080 a pattern statement.
1081 * tree-vect-patterns.c (type_conversion_p): Don't check
1082 STMT_VINFO_IN_PATTERN_P.
1083 (vect_recog_vector_vector_shift_pattern): Likewise.
1084 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1085 the pattern statement rather than the original statement; check
1086 directly for a WIDEN_MULT_EXPR here.
1087 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1088 vect_is_simple_use to return the pattern statement rather
1089 than the original statement; use is_pattern_stmt_p to check
1090 for such a pattern statement.
1091 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1092 to return the pattern statement rather than the original statement;
1093 don't do the same transformation here.
1094 (vect_is_simple_use): If the defining statement has been replaced
1095 by a pattern statement, return the pattern statement instead.
1096 Remove the corresponding (local) transformation from the vectype
1099 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1101 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1102 end and default to null.
1103 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1104 (vectorizable_reduction): Update calls accordingly, dropping the
1105 gimple ** argument if the passed-back statement isn't needed.
1106 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1107 (vect_recog_rotate_pattern): Likewise.
1108 (vect_recog_mask_conversion_pattern): Likewise.
1109 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1110 (vect_mask_constant_operand_p): Likewise.
1111 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1112 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1113 (get_group_load_store_type, get_load_store_type): Likewise.
1114 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1115 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1116 (vectorizable_conversion, vectorizable_assignment): Likewise.
1117 (vectorizable_shift, vectorizable_operation): Likewise.
1118 (vectorizable_store, vect_is_simple_cond): Likewise.
1119 (vectorizable_condition, vectorizable_comparison): Likewise.
1120 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1121 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1122 and move it to the end. Cope with null def_stmt_outs.
1124 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1126 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1128 2018-06-29 Jeff Law <law@redhat.com>
1130 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1131 displacements for TARGET_V850E2V3 and newer.
1132 (TARGET_LRA_P): Remove. Defaults to LRA now.
1133 * config/v850/v850.md (sign23byte_load): Remove.
1134 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1135 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1137 2018-06-29 Martin Liska <mliska@suse.cz>
1140 * coverage.c (coverage_init): Mangle full path name.
1141 * doc/invoke.texi: Document the change.
1142 * gcov-io.c (mangle_path): New.
1143 * gcov-io.h (mangle_path): Likewise.
1144 * gcov.c (mangle_name): Use mangle_path for path mangling.
1146 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1148 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1149 if starting source register is not even.
1151 2018-06-29 Martin Liska <mliska@suse.cz>
1153 PR tree-optimization/86263
1154 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1155 Make edge redirection.
1157 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1159 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1160 (dump_scope_depth): New variable.
1161 (get_dump_scope_depth): New function.
1162 (dump_begin_scope): New function.
1163 (dump_end_scope): New function.
1164 * dumpfile.h (get_dump_scope_depth): New declaration.
1165 (dump_begin_scope): New declaration.
1166 (dump_end_scope): New declaration.
1167 (class auto_dump_scope): New class.
1168 (AUTO_DUMP_SCOPE): New macro.
1169 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1172 2018-06-29 Richard Biener <rguenther@suse.de>
1174 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1175 compute_all_dependences succeeds.
1176 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1177 exceed --param loop-max-datarefs-for-datadeps.
1179 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1181 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1183 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1186 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1187 alternative 0 in preferred_for_speed attribute.
1189 2018-06-28 Paul Koning <ni1d@arrl.net>
1191 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1192 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1193 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1194 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1195 for shift insn patterns.
1197 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1199 * cgraph.c (cgraph_node::get_body): Replace assignments to
1200 "dump_file" with calls to set_dump_file.
1201 * dumpfile.c (alt_dump_file): Make static, and group with...
1202 (alt_flags): ...this definition.
1203 (dumps_are_enabled): New variable.
1204 (refresh_dumps_are_enabled): New function.
1205 (set_dump_file): New function.
1206 (set_alt_dump_file): New function.
1207 (gcc::dump_manager::dump_start): Replace assignments to
1208 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1210 (gcc::dump_manager::dump_finish): Likewise.
1211 * dumpfile.h (alt_dump_file): Delete decl.
1212 (dumps_are_enabled): New variable decl.
1213 (set_dump_file): New function decl.
1214 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1216 * tree-nested.c (lower_nested_functions): Replace assignments to
1217 "dump_file" with calls to set_dump_file.
1219 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1221 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1222 goto_locus of each outgoing edge of each basic block.
1224 2018-06-28 Richard Biener <rguenther@suse.de>
1226 * dwarf2out.c (decl_scope_table): Remove.
1227 (push_decl_scope): Likewise.
1228 (pop_decl_scope): Likewise.
1229 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1230 (gen_struct_or_union_type_die): Likewise.
1231 (gen_tagged_type_die): Likewise.
1232 (dwarf2out_init): Do not initialize decl_scope_table.
1233 (dwarf2out_c_finalize): Do not free it.
1235 2018-06-28 Richard Biener <rguenther@suse.de>
1237 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1238 deciding whether to not re-use a DIE.
1240 2018-06-28 Richard Biener <rguenther@suse.de>
1242 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1243 DW_AT_abstract_origin attribute.
1245 2018-06-28 Martin Liska <mliska@suse.cz>
1247 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1248 Use newly introduced constants.
1249 * tree-switch-conversion.h (struct jump_table_cluster):
1250 Define max_ratio_for_size and max_ratio_for_speed.
1252 2018-06-28 Martin Liska <mliska@suse.cz>
1254 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1255 Add new checking assert to catch invalid state.
1256 (jump_table_cluster::can_be_handled): Handle single case
1258 (jump_table_cluster::is_beneficial): Bail out for such case.
1259 (bit_test_cluster::find_bit_tests):
1260 Add new checking assert to catch invalid state.
1261 (bit_test_cluster::can_be_handled): Handle single case
1263 (bit_test_cluster::is_beneficial): Bail out for such case.
1264 (switch_decision_tree::analyze_switch_statement):
1267 2018-06-28 Martin Liska <mliska@suse.cz>
1269 * common.opt: Introduce -completion option.
1270 * gcc.c (driver_handle_option): Handle it.
1271 (driver::main): Print completions if completion
1273 * opt-suggestions.c (option_proposer::get_completions):
1275 (option_proposer::suggest_completion): Likewise.
1276 (option_proposer::find_param_completions): Likewise.
1277 (verify_autocompletions): Likewise.
1278 (test_completion_valid_options): Likewise.
1279 (test_completion_valid_params): Likewise.
1280 (in_completion_p): Likewise.
1281 (empty_completion_p): Likewise.
1282 (test_completion_partial_match): Likewise.
1283 (test_completion_garbage): Likewise.
1284 (opt_proposer_c_tests): Likewise.
1285 * opt-suggestions.h: Declare new functions.
1286 * opts.c (common_handle_option): Handle OPT__completion_.
1287 * selftest-run-tests.c (selftest::run_tests): Add
1288 opt_proposer_c_tests.
1289 * selftest.c (assert_str_startswith): New.
1290 * selftest.h (assert_str_startswith): Likewise.
1291 (opt_proposer_c_tests): New.
1292 (ASSERT_STR_STARTSWITH): Likewise.
1294 2018-06-28 Martin Liska <mliska@suse.cz>
1296 * Makefile.in: Add opt-suggestions.o.
1297 * gcc-main.c: Include opt-suggestions.h.
1298 * gcc.c (driver::driver): Likewise.
1299 (driver::~driver): Remove m_option_suggestions.
1300 (driver::build_option_suggestions): Moved to option_proposer.
1301 (driver::suggest_option): Likewise.
1302 (driver::handle_unrecognized_options): Use option_proposer.
1303 * gcc.h (class driver): Add new memver m_option_proposer.
1304 * opt-suggestions.c: New file.
1305 * opt-suggestions.h: New file.
1307 2018-06-28 Martin Liska <mliska@suse.cz>
1309 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1310 (auto_string_vec::~auto_string_vec): Likewise.
1312 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1314 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1315 prevent_decl_creation_for_types fields up and add reset_location field.
1316 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1317 statement if id->reset_location is true.
1318 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1319 id->reset_location is true.
1320 (copy_phis_for_bb): Force input_location on the arguments if
1321 id->reset_location is true.
1322 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1323 is set on the function to be inlined.
1325 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1327 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1329 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1331 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1332 registers for Pmode.
1333 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1334 hard registers for the clobbered pseudo.
1336 2018-06-27 Paul Koning <ni1d@arrl.net>
1338 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1339 mutually exclusive options.
1340 * config/pdp11/constraints.md (h): New constraint.
1341 (O): Update definition to match shift code generation.
1342 (D): New constraint.
1343 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1345 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1347 (output_jump): Change arguments.
1348 (pdp11_fixed_cc_regs): New function.
1349 (pdp11_cc_mode): Ditto.
1350 (pdp11_expand_shift): Ditto.
1351 (pdp11_assemble_shift): Ditto.
1352 (pdp11_small_shift): Ditto.
1353 (pdp11_branch_cost): Remove.
1354 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1356 (pdp11_register_move_cost): Update for CC registers.
1357 (pdp11_rtx_costs): Add case for LSHIFTRT.
1358 (pdp11_output_jump): Add CCNZ mode conditional branches.
1359 (notice_update_cc_on_set): Remove.
1360 (pdp11_cc_mode): New function.
1361 (simple_memory_operand): Correct pre/post decrement case.
1362 (no_side_effect_operand): New function.
1363 (pdp11_regno_reg_class): Add CC_REGS class.
1364 (pdp11_fixed_cc_regs): New function.
1365 (pdp11_small_shift): New function.
1366 (pdp11_expand_shift): New function to expand shift insns.
1367 (pdp11_assemble_shift): New function to output shifts.
1368 (pdp11_branch_cost): Remove.
1369 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1370 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1371 (WCHAR_TYPE): Ditto.
1372 (PTRDIFF_TYPE): Ditto.
1373 (ADJUST_INSN_LENGTH): New macro.
1374 (FIXED_REGISTERS): Add CC registers.
1375 (CALL_USED_REGISTERS): Ditto.
1377 (REG_CLASS_NAMES): Ditto.
1378 (REG_CLASS_CONTENTS): Ditto.
1379 (SELECT_CC_MODE): Use new function.
1380 (TARGET_FLAGS_REGNUM): New macro.
1381 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1382 (cc0_reg_rtx): Remove.
1383 (CC_STATUS_MDEP): Remove.
1384 (CC_STATUS_MDEFP_INIT): Remove.
1385 (CC_IN_FPU): Remove.
1386 (NOTICE_UPDATE_CC): Remove.
1387 (REGISTER_NAMES): Add CC registers.
1388 (BRANCH_COST): Change to constant 1.
1389 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1391 * config/pdp11/pdp11.opt (mbcopy): Remove.
1392 (mbcopy-builtin): Remove.
1393 (mbranch-cheap): Remove.
1394 (mbranch-expensive): Remove.
1395 * config/pdp11/predicates.md (expand_shift_operand): Update to
1396 match shift code generation.
1397 (ccnz_operator): New predicate.
1398 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1399 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1400 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1402 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1403 description of O constraint.
1405 2018-06-27 Jeff Law <law@redhat.com>
1406 Austin Law <austinklaw@gmail.com>
1408 * config/v850/v850.md (addsi3_set_flags): New pattern.
1409 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1410 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1411 (zero_extendhisi2_v850_set_flags): Likewise.
1412 (zero_extendqisi2_v850_set_flags): Likewise.
1413 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1414 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1415 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1417 * config/v850/v850-protos.h (notice_update_cc): Remove.
1418 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1419 (v850_print_operand): Handle 'D' and "d".
1420 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1421 Add handling of arithmetic/logical operations compared against zero.
1422 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1423 Do not look at v850_compare_op, instead get mode from last argument.
1424 (v850_gen_compare): Remove
1425 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1426 after reload for prologue insns.
1427 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1429 (construct_save_jarl): Likewise.
1430 (TARGET_FLAGS_REGNUM): Define.
1431 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1432 (NOTICE_UPDATE_CC): Remove.
1433 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1434 than cc0. Conditionalize on reload_completed.
1435 (cmpsi_insn, setfcc_insn): Likewise.
1436 (tst1 splitter): Turn into define_and_split which sets the flags
1438 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1439 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1440 (cstoresf4, cstoredf4): Clobber the flags.
1441 (cmpsi, cmpsf, cmpdf): Remove expanders.
1442 (setf_insn): Remove pattern.
1443 (addsi3): Turn into define_and_split which clobbers the flags after
1444 reload and a suitable pattern (addsi3_clobber_flags) for use after
1446 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1447 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1448 (ashrsi3, ashrsi3_v850e2): Likewise.
1449 (bins): Clobber the flags.
1450 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1451 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1452 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1453 (call_value_internal_short, call_value_internal_long): Likewise.
1454 (callt_save_interrupt, callt_return_interrupt): Likewise.
1455 (save_interrupt, return_interrupt): Likewise.
1456 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1457 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1458 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1459 (All FP comparisons): Only allow after reload has completed.
1461 (divh, divhu): Tweak output template.
1462 (branch_z_normal, branch_z_invert): Remove
1463 (branch_nz_normal, branch_nz_invert): Likewise.
1464 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1466 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1467 * config/v850/v850.c (notice_update_cc): Remove.
1468 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1469 (CC_NO_CARRY): Likewise.
1470 (NOTICE_UPDATE_CC): Define to nothing.
1471 * config/v850/v850.md: Remove block comment on cc0 handling
1472 Remove "cc" attribute from all patterns. Remove cc_status handling
1473 from all patterns. Minor formatting fixes.
1475 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1477 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1478 (cortex-a76.cortex-a55): Likewise.
1479 * config/aarch64/aarch64-tune.md: Regenerate.
1480 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1481 cortex-a76.cortex-a55.
1483 2018-06-27 Jeff Law <law@redhat.com>
1485 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1486 (MULTILIB_DIRNAMES): Similarly.
1488 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1490 * gimple.h (gimple_return_retbnd): Delete.
1491 (gimple_return_set_retbnd): Likewise.
1492 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1493 gimple_return_set_retbnd.
1494 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1495 gimple_return_retbnd and adjust.
1496 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1497 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1498 Explicitly return NULL in a couple more cases. Move assertion
1499 on debug statements and remove unreachable code.
1500 (reset_debug_binding): Do not test id->retbnd.
1501 (expand_call_inline): Do not set it.
1503 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1505 * configure.ac: Add --disable-gcov option.
1506 * configure: Regenerate.
1507 * Makefile.in: Honour @enable_gcov@.
1508 * doc/install.texi: Document --disable-gcov.
1510 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1512 * config/arm/arm-cpus.in (cortex-a76): New entry.
1513 (cortex-a76.cortex-a55): Likewise.
1514 * config/arm/arm-tables.opt: Regenerate.
1515 * config/arm/arm-tune.md: Likewise.
1516 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1517 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1518 cortex-a76.cortex-a55.
1520 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1523 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1525 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1527 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1529 (EPILOGUE_USES): Likewise.
1531 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1533 * tree-inline.c (remap_location): New function extracted from...
1534 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1535 (copy_phis_for_bb): ...here. Call remap_location.
1536 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1538 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1540 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1541 unaligned vsx for 16B memset.
1543 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1546 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1547 ieee128_float_type_node to long_double_type_node unless
1548 TARGET_LONG_DOUBLE_128 is set.
1550 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1552 * cfgloop.c (get_loop_location): Convert return type from
1553 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1554 by implicit construction from rtx_insn *, and using
1555 dump_user_location_t::from_function_decl for the fallback case.
1556 * cfgloop.h (get_loop_location): Convert return type from
1557 location_t to dump_user_location_t.
1558 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1559 dump_printf_loc to pass in a dump_location_t rather than a
1560 location_t, via the gimple stmt.
1561 * coverage.c (get_coverage_counts): Update calls to
1562 dump_printf_loc to pass in dump_location_t rather than a
1564 * doc/optinfo.texi (Dump types): Convert example of
1565 dump_printf_loc from taking "locus" to taking "insn". Update
1566 description of the "_loc" calls to cover dump_location_t.
1567 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1569 (dump_user_location_t::dump_user_location_t): New constructors,
1570 from gimple *stmt and rtx_insn *.
1571 (dump_user_location_t::from_function_decl): New function.
1572 (dump_loc): Make static.
1573 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1574 const dump_location_t &.
1575 (dump_generic_expr_loc): Delete.
1576 (dump_printf_loc): Convert param "loc" from location_t to
1577 const dump_location_t &.
1578 (selftest::test_impl_location): New function.
1579 (selftest::dumpfile_c_tests): New function.
1580 * dumpfile.h: Include "profile-count.h".
1581 (class dump_user_location_t): New class.
1582 (struct dump_impl_location_t): New struct.
1583 (class dump_location_t): New class.
1584 (dump_printf_loc): Convert 2nd param from source_location to
1585 const dump_location_t &.
1586 (dump_generic_expr_loc): Delete.
1587 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1588 const dump_location_t &.
1589 * gimple-fold.c (fold_gimple_assign): Update call to
1590 dump_printf_loc to pass in a dump_location_t rather than a
1591 location_t, via the gimple stmt.
1592 (gimple_fold_call): Likewise.
1593 * gimple-loop-interchange.cc
1594 (loop_cand::analyze_iloop_reduction_var): Update for change to
1595 check_reduction_path.
1596 (tree_loop_interchange::interchange): Update for change to
1598 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1599 change in return-type of find_loop_location.
1600 (graphite_regenerate_ast_isl): Likewise.
1601 * graphite-optimize-isl.c (optimize_isl): Likewise.
1602 * graphite.c (graphite_transform_loops): Likewise.
1603 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1604 pass in a dump_location_t rather than a location_t, via the
1606 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1607 * ipa.c (walk_polymorphic_call_targets): Likewise.
1608 * loop-unroll.c (report_unroll): Convert "locus" param from
1609 location_t to dump_location_t.
1610 (decide_unrolling): Update for change to get_loop_location's
1612 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1613 location_t to dump_user_location_t.
1614 (grid_find_single_omp_among_assignments_1): Updates calls to
1615 dump_printf_loc to pass in a dump_location_t rather than a
1616 location_t, via the gimple stmt.
1617 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1618 location_t to dump_location_t. Updates calls to dump_printf_loc
1619 to pass in a dump_location_t rather than a location_t, via the
1621 (grid_inner_loop_gridifiable_p): Likewise.
1622 (grid_dist_follows_simple_pattern): Likewise.
1623 (grid_gfor_follows_tiling_pattern): Likewise.
1624 (grid_target_follows_gridifiable_pattern): Likewise.
1625 (grid_attempt_target_gridification): Convert initialization
1626 of local "grid" from memset to zero-initialization; FIXME: does
1627 this require C++11? Update call to dump_printf_loc to pass in a
1628 optinfo_location rather than a location_t, via the gimple stmt.
1629 * profile.c (read_profile_edge_counts): Updates call to
1630 dump_printf_loc to pass in a dump_location_t rather than a
1632 (compute_branch_probabilities): Likewise.
1633 * selftest-run-tests.c (selftest::run_tests): Call
1635 * selftest.h (dumpfile_c_tests): New decl.
1636 * tree-loop-distribution.c (pass_loop_distribution::execute):
1637 Update for change in return type of find_loop_location.
1638 * tree-parloops.c (parallelize_loops): Likewise.
1639 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1640 "locus" from location_t to dump_user_location_t.
1641 (canonicalize_loop_induction_variables): Likewise.
1642 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1643 for change in return type of find_loop_location.
1644 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1645 to dump_printf_loc to pass in a dump_location_t rather than a
1646 location_t, via the stmt.
1647 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1649 * tree-vect-loop-manip.c (find_loop_location): Convert return
1650 type from source_location to dump_user_location_t.
1651 (vect_do_peeling): Update for above change.
1652 (vect_loop_versioning): Update for change in type of
1654 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1655 from location_t to dump_user_location_t.
1656 (vect_estimate_min_profitable_iters): Update for change in type
1658 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1659 location_t to dump_location_t.
1660 (vect_slp_bb): Update for change in type of vect_location.
1661 * tree-vectorizer.c (vect_location): Convert from source_location
1662 to dump_user_location_t.
1663 (try_vectorize_loop_1): Update for change in vect_location's type.
1664 (vectorize_loops): Likewise.
1665 (increase_alignment): Likewise.
1666 * tree-vectorizer.h (vect_location): Convert from source_location
1667 to dump_user_location_t.
1668 (find_loop_location): Convert return type from source_location to
1669 dump_user_location_t.
1670 (check_reduction_path): Convert 1st param from location_t to
1671 dump_user_location_t.
1672 * value-prof.c (check_counter): Update call to dump_printf_loc to
1673 pass in a dump_user_location_t rather than a location_t; update
1674 call to error_at for change in type of "locus".
1675 (check_ic_target): Update call to dump_printf_loc to
1676 pass in a dump_user_location_t rather than a location_t, via the
1679 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1681 * config/s390/s390.h (enum processor_flags): Do not use
1682 default tune parameter when -march was specified.
1684 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1687 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1688 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1691 2018-06-26 Richard Biener <rguenther@suse.de>
1693 PR tree-optimization/86287
1695 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1696 (vect_analyze_loop): Initialize n_stmts.
1698 2018-06-26 Richard Biener <rguenther@suse.de>
1701 * fold-const.c (fold_convertible_p): Pointer extension
1704 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1707 * dwarf2out.c (loc_list_has_views): Adjust comments.
1708 (dw_loc_list): Split single cross-partition range with
1711 2018-06-25 Jeff Law <law@redhat.com>
1713 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1714 on -mbig-switch by default.
1716 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1718 (const_float_0_operand): Remove unused predicate.
1719 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1720 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1721 (recipsf2): New expander. Original pattern now called
1723 (recipdf2, recipdf2_insn): Similarly.
1724 (rsqrtsf2, rsqrtsf2_insn): Similarly
1725 (rsqrtdf2, rsqrtdf2_insn): Similarly
1727 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1729 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1730 Simplify logic for FreeBSD (twice).
1732 2018-06-25 Martin Sebor <msebor@redhat.com>
1734 PR tree-optimization/86204
1735 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1736 a strnlen result if it's less than the length of the string.
1738 2018-06-25 Martin Sebor <msebor@redhat.com>
1740 PR tree-optimization/85700
1741 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1742 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1743 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1745 2018-06-25 Martin Sebor <msebor@redhat.com>
1747 * doc/extend.texi (Zero-length arrays): Update and clarify.
1749 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1751 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1752 added IEEE/IBM long double multilib support on PowerPC little
1753 endian Linux systems.
1754 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1755 (MULTILIB_DEFAULTS): Likewise.
1756 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1758 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1759 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1760 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1762 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1765 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1766 (REORDER_45): Likewise.
1768 2018-06-25 Jeff Law <law@redhat.com>
1770 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1771 dividend to 32 bits. Adjust length.
1772 (udivmodhi4): Cleanup output template. Fix length.
1774 2018-06-25 Carl Love <cel@us.ibm.com>
1776 * config/rs6000/vsx.md: Change word selector to prefered location.
1778 2018-06-25 Richard Biener <rguenther@suse.de>
1780 PR tree-optimization/86304
1781 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1782 epilogue-if-converted loops as well.
1784 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1786 * lto-section-out.c (lto_begin_section): Do not print section
1787 name for noaddr and unnumbered dumps.
1789 2018-06-25 Richard Biener <rguenther@suse.de>
1791 * tree-vectorizer.h (struct vec_info_shared): New structure
1792 with parts split out from struct vec_info and loop_nest from
1793 struct _loop_vec_info.
1794 (struct vec_info): Adjust accordingly.
1795 (struct _loop_vec_info): Likewise.
1796 (LOOP_VINFO_LOOP_NEST): Adjust.
1797 (LOOP_VINFO_DATAREFS): Likewise.
1798 (LOOP_VINFO_DDRS): Likewise.
1799 (struct _bb_vec_info): Likewise.
1800 (BB_VINFO_DATAREFS): Likewise.
1801 (BB_VINFO_DDRS): Likewise.
1802 (struct _stmt_vec_info): Add dr_aux member.
1803 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1804 (DR_MISALIGNMENT_UNINITIALIZED): New.
1805 (set_dr_misalignment): Adjust.
1806 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1807 (vect_analyze_loop): Adjust prototype.
1808 (vect_analyze_loop_form): Likewise.
1809 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1810 Compute dependences lazily.
1811 (vect_record_base_alignments): Use shared datarefs/ddrs.
1812 (vect_verify_datarefs_alignment): Likewise.
1813 (vect_analyze_data_refs_alignment): Likewise.
1814 (vect_analyze_data_ref_accesses): Likewise.
1815 (vect_analyze_data_refs): Likewise.
1816 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1817 constructor parameter for shared part.
1818 (vect_analyze_loop_form): Pass in shared part and adjust.
1819 (vect_analyze_loop_2): Pass in storage for the number of
1820 stmts. Move loop nest finding to the caller. Compute
1822 (vect_analyze_loop): Pass in shared part.
1823 (vect_transform_loop): Verify shared datarefs are unchanged.
1824 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1825 constructor parameter for shared part.
1826 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1827 (vect_slp_bb): Verify shared datarefs are unchanged before
1829 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1831 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1832 DR_MISALIGNMENT_UNINITIALIZED.
1833 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1834 parameter for shared part.
1835 (vec_info::~vec_info): Adjust.
1836 (vec_info_shared::vec_info_shared): New.
1837 (vec_info_shared::~vec_info_shared): Likewise.
1838 (vec_info_shared::save_datarefs): Likewise.
1839 (vec_info_shared::check_datarefs): Likewise.
1840 (try_vectorize_loop_1): Construct shared part live for analyses
1841 of a single loop for multiple vector sizes.
1842 * tree-parloops.c (gather_scalar_reductions): Adjust.
1844 2018-06-25 Richard Biener <rguenther@suse.de>
1846 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1847 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1848 (vect_analyze_data_refs): Remove similar code from here and
1849 simplify accordingly.
1851 2018-06-25 Richard Biener <rguenther@suse.de>
1853 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1854 for reverse storage order accesses rather than asserting
1855 they cannot happen here.
1857 2018-06-25 Tom de Vries <tdevries@suse.de>
1860 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1861 Use data16 instead of .byte for insn prefix.
1863 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1866 * parser.c (make_char_string_pack): Pass this literal chars
1867 through cpp_interpret_string.
1868 (cp_parser_userdef_numeric_literal): Check the result of
1869 make_char_string_pack.
1871 2018-06-24 Maya Rashish <coypu@sdf.org>
1873 * ginclude/stddef.h: Simplify conditions around avoiding
1874 re-definition of __size_t.
1876 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1878 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1879 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1881 2018-06-22 Maya Rashish <coypu@sdf.org>
1883 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1884 made to OpenBSD and NetBSD through the years.
1886 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1888 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1889 behavior of vec_pack (vector double, vector double) to match
1890 behavior of vec_float2 (vector double, vector double).
1892 2018-06-22 Olivier Hainque <hainque@adacore.com>
1894 * gimplify.c (gimplify_function_tree): Prevent creation
1895 of a trampoline for the address of the current function
1896 passed to entry/exit instrumentation hooks.
1898 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1901 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1904 2018-06-22 Martin Liska <mliska@suse.cz>
1906 PR tree-optimization/86263
1907 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1908 Bail out if is_enabled is false.
1909 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1911 (jump_table_cluster::is_enabled): New function.
1913 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1915 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1916 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1917 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1918 (lto_input_ts_binfo_tree_pointers): Likewise.
1919 * tree-streamer-out.c (streamer_write_tree_bitfields,
1920 write_ts_binfo_tree_pointers): Likewise.
1921 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1923 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1925 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1927 2018-06-22 Martin Liska <mliska@suse.cz>
1929 * symbol-summary.h (get): Make it pure and inline move
1930 functionality from ::get function.
1931 (get): Remove and inline into ::get and ::get_create.
1932 (get_create): Move code from ::get function.
1934 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1937 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1938 -x assembler-with-cpp.
1940 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1942 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1943 _FILE_OFFSET_BITS=64 for C++.
1945 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1947 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1948 conversion insn that shows up when pr85657-3.c is compiled using
1949 IEEE 128-bit long double.
1950 (neg<mode>2_internal): Use the correct mode to check whether the
1951 mode is IBM extended.
1952 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1953 multiply and divide external functions from being created more
1956 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1958 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1960 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1961 the edge can be forwarded.
1962 (cfg_layout_merge_blocks): Likewise.
1964 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1966 * except.c (finish_eh_generation): Commit edge insertions only after
1967 the EH edges have been redirected from post-landing to landing pads.
1969 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1971 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1972 create_tmp_var_for to create the FRAME decl.
1973 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1975 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1977 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1978 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1979 debug statement when resetting its value.
1980 (expand_call_inline): Copy the locus of the call onto the assignment
1981 of the return value, if any. Use local variable in more cases.
1983 2018-06-21 Martin Liska <mliska@suse.cz>
1985 * ipa-pure-const.c (propagate_nothrow): Use
1986 funct_state_summaries->get.
1987 (dump_malloc_lattice): Likewise.
1988 (propagate_malloc): Likewise.
1990 2018-06-21 Richard Biener <rguenther@suse.de>
1992 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1993 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1994 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1996 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1997 BLOCK_ABSTRACT_ORIGIN unconditionally.
1999 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2001 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2003 * ipa-reference.c (ipa_reference_c_finalize): Delete
2004 ipa_ref_opt_sum_summaries and set it to NULL.
2006 2018-06-21 Tom de Vries <tdevries@suse.de>
2008 PR tree-optimization/85859
2009 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2010 test with comment from bb_no_side_effects_p.
2012 2018-06-21 Richard Biener <rguenther@suse.de>
2014 PR tree-optimization/86232
2015 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2016 max for constant niter.
2018 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2020 * config/aarch64/aarch64-simd.md
2021 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2023 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2025 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2026 Make opernads of the unspec commutative.
2028 2018-06-21 Richard Biener <rguenther@suse.de>
2030 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2031 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2032 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2033 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2035 (vect_record_base_alignments): Do not record base alignment
2036 for gathers or scatters.
2037 (vect_compute_data_ref_alignment): Drop return value that is always
2038 true. Bail out early for gathers or scatters.
2039 (vect_enhance_data_refs_alignment): Bail out early for gathers
2041 (vect_find_same_alignment_drs): Likewise.
2042 (vect_analyze_data_refs_alignment): Remove dead code.
2043 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2044 (vect_analyze_data_refs): For possible gathers or scatters do
2045 not create an alternate DR, just check their possible validity
2046 and mark them. Adjust DECL_NONALIASED handling to not rely
2048 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2049 update inits of gathers or scatters.
2050 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2051 Also copy gather/scatter flag to pattern vinfo.
2053 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2055 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2056 behavior of vec_packsu (vector unsigned long long, vector unsigned
2057 long long) to match behavior of vec_packs with same signature.
2059 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2060 Thomas Schwinge <thomas@codesourcery.com>
2061 Cesar Philippidis <cesar@codesourcery.com>
2063 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2064 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2065 (gimplify_adjust_omp_clauses): Likewise.
2066 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2067 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2068 (gimplify_omp_target_update): Update handling of acc update and
2070 * omp-low.c (install_var_field): Remove unused parameter
2071 base_pointers_restrict.
2072 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2073 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2075 (omp_target_base_pointers_restrict_p): Delete.
2076 (scan_omp_target): Update call to scan_sharing_clauses.
2077 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2079 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2080 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2081 (convert_local_omp_clauses): Likewise.
2082 * tree-pretty-print.c (dump_omp_clause): Likewise.
2083 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2085 (omp_clause_code_name): Likewise.
2087 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2090 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2093 PR tree-optimization/86231
2094 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2095 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2097 2018-06-20 Tom de Vries <tdevries@suse.de>
2099 PR tree-optimization/86097
2100 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2101 iv type if signedness of iv type is not the same as that of *nit.
2103 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2105 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2106 EDGE_EH edges, verify they are all EDGE_EH.
2108 2018-06-20 Maya Rashish <coypu@sdf.org>
2110 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2112 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2114 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2115 * config/aarch64/aarch64.c (xgene1_tunings): Add
2116 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2117 (aarch64_mode_valid_for_sched_fusion_p):
2118 Allow 16-byte modes.
2119 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2120 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2122 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2124 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2125 * config/aarch64/iterators.md (VQ2): New mode iterator.
2127 2018-06-20 Martin Liska <mliska@suse.cz>
2129 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2130 Change default ratio from 10 to 8.
2132 2018-06-20 Martin Liska <mliska@suse.cz>
2134 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2136 (bit_test_cluster::find_bit_tests): Likewise.
2137 (switch_decision_tree::analyze_switch_statement): Find clusters.
2138 * tree-switch-conversion.h (struct jump_table_cluster): Document
2141 2018-06-20 Martin Liska <mliska@suse.cz>
2143 * tree-switch-conversion.c (switch_conversion::collect):
2144 Record m_uniq property.
2145 (switch_conversion::expand): Bail out for special conditions.
2146 (group_cluster::~group_cluster): New.
2147 (group_cluster::group_cluster): Likewise.
2148 (group_cluster::dump): Likewise.
2149 (jump_table_cluster::emit): New.
2150 (switch_decision_tree::fix_phi_operands_for_edges): New.
2151 (struct case_node): Remove struct.
2152 (jump_table_cluster::can_be_handled): New.
2153 (case_values_threshold): Moved to header.
2154 (reset_out_edges_aux): Likewise.
2155 (jump_table_cluster::is_beneficial): New.
2156 (bit_test_cluster::can_be_handled): Likewise.
2157 (add_case_node): Remove.
2158 (bit_test_cluster::is_beneficial): New.
2159 (case_bit_test::cmp): New.
2160 (bit_test_cluster::emit): New.
2161 (expand_switch_as_decision_tree_p): Remove.
2162 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2163 (fix_phi_operands_for_edge): Likewise.
2164 (switch_decision_tree::analyze_switch_statement): New.
2165 (compute_cases_per_edge): Move ...
2166 (switch_decision_tree::compute_cases_per_edge): ... here.
2167 (try_switch_expansion): Likewise.
2168 (switch_decision_tree::try_switch_expansion): Likewise.
2169 (record_phi_operand_mapping): Likewise.
2170 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2171 (emit_case_decision_tree): Likewise.
2172 (switch_decision_tree::emit): Likewise.
2173 (balance_case_nodes): Likewise.
2174 (switch_decision_tree::balance_case_nodes): Likewise.
2175 (dump_case_nodes): Likewise.
2176 (switch_decision_tree::dump_case_nodes): Likewise.
2177 (emit_jump): Likewise.
2178 (switch_decision_tree::emit_jump): Likewise.
2179 (emit_cmp_and_jump_insns): Likewise.
2180 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2181 (emit_case_nodes): Likewise.
2182 (switch_decision_tree::emit_case_nodes): Likewise.
2183 (conditional_probability): Remove.
2184 * tree-switch-conversion.h (enum cluster_type): New.
2186 (struct cluster): Likewise.
2187 (cluster::cluster): Likewise.
2188 (struct simple_cluster): Likewise.
2189 (simple_cluster::simple_cluster): Likewise.
2190 (struct group_cluster): Likewise.
2191 (struct jump_table_cluster): Likewise.
2192 (struct bit_test_cluster): Likewise.
2193 (struct min_cluster_item): Likewise.
2194 (struct case_tree_node): Likewise.
2195 (case_tree_node::case_tree_node): Likewise.
2196 (jump_table_cluster::case_values_threshold): Likewise.
2197 (struct case_bit_test): Likewise.
2198 (struct switch_decision_tree): Likewise.
2199 (struct switch_conversion): Likewise.
2200 (switch_decision_tree::reset_out_edges_aux): Likewise.
2202 2018-06-20 Martin Liska <mliska@suse.cz>
2204 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2205 (hoist_edge_and_branch_if_true): Likewise.
2206 (expand_switch_using_bit_tests_p): Likewise.
2207 (struct case_bit_test): Likewise.
2208 (case_bit_test_cmp): Likewise.
2209 (emit_case_bit_tests): Likewise.
2210 (switch_conversion::switch_conversion): New class.
2211 (struct switch_conv_info): Remove old struct.
2212 (collect_switch_conv_info): More to ...
2213 (switch_conversion::collect): ... this.
2214 (check_range): Likewise.
2215 (switch_conversion::check_range): Likewise.
2216 (check_all_empty_except_final): Likewise.
2217 (switch_conversion::check_all_empty_except_final): Likewise.
2218 (check_final_bb): Likewise.
2219 (switch_conversion::check_final_bb): Likewise.
2220 (create_temp_arrays): Likewise.
2221 (switch_conversion::create_temp_arrays): Likewise.
2222 (free_temp_arrays): Likewise.
2223 (gather_default_values): Likewise.
2224 (switch_conversion::gather_default_values): Likewise.
2225 (build_constructors): Likewise.
2226 (switch_conversion::build_constructors): Likewise.
2227 (constructor_contains_same_values_p): Likewise.
2228 (switch_conversion::contains_same_values_p): Likewise.
2229 (array_value_type): Likewise.
2230 (switch_conversion::array_value_type): Likewise.
2231 (build_one_array): Likewise.
2232 (switch_conversion::build_one_array): Likewise.
2233 (build_arrays): Likewise.
2234 (switch_conversion::build_arrays): Likewise.
2235 (gen_def_assigns): Likewise.
2236 (switch_conversion::gen_def_assigns): Likewise.
2237 (prune_bbs): Likewise.
2238 (switch_conversion::prune_bbs): Likewise.
2239 (fix_phi_nodes): Likewise.
2240 (switch_conversion::fix_phi_nodes): Likewise.
2241 (gen_inbound_check): Likewise.
2242 (switch_conversion::gen_inbound_check): Likewise.
2243 (process_switch): Use the newly created class.
2244 (switch_conversion::expand): New.
2245 (switch_conversion::~switch_conversion): New.
2246 * tree-switch-conversion.h: New file.
2248 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2250 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2251 tree-vect-patterns.c.
2252 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2253 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2254 (vect_recog_sad_pattern): Likewise.
2255 (vect_recog_widen_sum_pattern): Likewise.
2256 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2257 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2258 (vect_recog_rotate_pattern): Likewise.
2259 (vect_recog_mult_pattern): Likewise.
2260 (vect_recog_vector_vector_shift_pattern): Likewise.
2261 (vect_recog_divmod_pattern): Likewise.
2262 (vect_recog_mixed_size_cond_pattern): Likewise.
2263 (vect_recog_bool_pattern): Likewise.
2264 (vect_recog_mask_conversion_pattern): Likewise.
2265 (vect_try_gather_scatter_pattern): Likewise.
2266 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2267 (vect_recog_over_widening_pattern): Likewise.
2268 (vect_recog_gather_scatter_pattern): Likewise.
2269 (vect_recog_func_ptr): Move from tree-vectorizer.h
2270 (vect_vect_recog_func_ptrs): Move further down the file.
2271 (vect_recog_func): Likewise. Remove the third argument.
2272 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2273 (vect_pattern_recog_1): Expect the pattern function to do any
2274 necessary target tests. Also expect it to provide a vector type.
2275 Remove the type_in handling.
2277 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2279 * tree-vect-patterns.c (vect_pattern_detected): New function.
2280 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2281 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2282 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2283 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2284 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2285 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2286 (vect_recog_mask_conversion_pattern)
2287 (vect_try_gather_scatter_pattern): Likewise.
2289 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2291 * tree-vect-patterns.c (vect_get_internal_def): New function.
2292 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2293 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2294 (search_type_for_mask_1): Use it.
2296 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2298 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2299 redundant WIDEN_SUM_EXPR handling.
2300 (vect_recog_sad_pattern): Likewise.
2302 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2304 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2305 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2306 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2307 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2308 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2310 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2312 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2313 use the stmt_vec_info of the original bb statement for the
2314 new zero assignment, even if the call is part of a pattern.
2316 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2318 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2319 that the sequence is attached to the original statement rather
2320 than the pattern statement.
2321 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2322 PATTERN_DEF_SEQ from the original statement rather than
2323 the main pattern statement.
2324 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2325 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2326 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2328 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2330 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2331 definition statements before the early exit for statements that aren't
2333 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2335 (vect_transform_loop): ...here. Process pattern definition
2336 statements without first checking whether the main pattern
2337 statement is live or relevant.
2339 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2341 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2342 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2344 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2346 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2347 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2348 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2349 (expand_block_compare): Change select_block_compare_mode call.
2350 (expand_strncmp_align_check): Use new functions, fix comment.
2351 (emit_final_str_compare_gpr): New function.
2352 (expand_strn_compare): Refactor and clean up code.
2353 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2355 2018-06-19 Tony Reix <tony.reix@atos.com>
2356 Damien Bergamini <damien.bergamini@atos.com>
2357 David Edelsohn <dje.gcc@gmail.com>
2359 * collect2.c (static_obj): New variable.
2360 (static_libs): New variable.
2361 (is_in_list): Uncomment declaration.
2362 (main): Track AIX libraries linked statically.
2363 (is_in_list): Uncomment definition.
2364 (scan_prog_file): Don't add AIX shared libraries initializer
2365 to constructor list if linking statically.
2367 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2369 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2371 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2373 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2375 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2378 2018-06-19 Martin Liska <mliska@suse.cz>
2380 * config/i386/i386.c (ix86_can_inline_p): Do not use
2381 ipa_fn_summaries::get_create.
2382 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2384 (devirtualization_time_bonus): Likewise.
2385 (ipcp_propagate_stage): Likewise.
2386 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2387 (edge_set_predicate): Likewise.
2388 (evaluate_conditions_for_known_args): Likewise.
2389 (evaluate_properties_for_edge): Likewise.
2390 (ipa_call_summary::reset): Tranform to ...
2391 (ipa_call_summary::~ipa_call_summary): ... this.
2392 (ipa_fn_summary::reset): Transform to ...
2393 (ipa_fn_summary::~ipa_fn_summary): ... this.
2394 (ipa_fn_summary_t::remove): Rename to ...
2395 (ipa_fn_summary_t::remove_callees): ... this.
2396 (ipa_fn_summary_t::duplicate): Use placement new
2397 instead of memory copy.
2398 (ipa_call_summary_t::duplicate): Likewise.
2399 (ipa_call_summary_t::remove): Remove.
2400 (dump_ipa_call_summary): Change get_create to get.
2401 (ipa_dump_fn_summary): Dump only when summary exists.
2402 (analyze_function_body): Use symbol_summary::get instead
2404 (compute_fn_summary): Likewise.
2405 (estimate_edge_devirt_benefit): Likewise.
2406 (estimate_edge_size_and_time): Likewise.
2407 (inline_update_callee_summaries): Likewise.
2408 (remap_edge_change_prob): Likewise.
2409 (remap_edge_summaries): Likewise.
2410 (ipa_merge_fn_summary_after_inlining): Likewise.
2411 (write_ipa_call_summary): Likewise.
2412 (ipa_fn_summary_write): Likewise.
2413 (ipa_free_fn_summary): Likewise.
2414 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2415 (struct ipa_call_summary): Likewise.
2416 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2418 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2419 (estimate_size_after_inlining): Likewise.
2420 (estimate_growth): Likewise.
2421 (growth_likely_positive): Likewise.
2422 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2423 (inline_call): Likewise.
2424 * ipa-inline.c (caller_growth_limits): Likewise.
2425 (can_inline_edge_p): Likewise.
2426 (can_inline_edge_by_limits_p): Likewise.
2427 (compute_uninlined_call_time): Likewise.
2428 (compute_inlined_call_time): Likewise.
2429 (want_inline_small_function_p): Likewise.
2430 (edge_badness): Likewise.
2431 (update_caller_keys): Likewise.
2432 (update_callee_keys): Likewise.
2433 (inline_small_functions): Likewise.
2434 (inline_to_all_callers_1): Likewise.
2435 (dump_overall_stats): Likewise.
2436 (early_inline_small_functions): Likewise.
2437 (early_inliner): Likewise.
2438 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2439 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2440 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2441 * ipa-split.c (execute_split_functions): Likewise.
2442 * symbol-summary.h: Likewise.
2443 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2445 2018-06-19 Richard Biener <rguenther@suse.de>
2447 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2448 (vectorize_loops): ... here. Fix dbgcnt handling.
2449 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2451 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2454 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2455 ieee128 argument takes up only one (vector) register, not two (floating
2458 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2460 * gimplify.c (gimplify_init_constructor): Really never clear for an
2461 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2463 2018-06-19 Richard Biener <rguenther@suse.de>
2465 PR tree-optimization/86179
2466 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2467 after failed recognition.
2469 2018-06-18 Martin Sebor <msebor@redhat.com>
2472 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2473 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2474 Handle integer subtraction.
2475 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2476 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2478 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2480 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2481 param from rtx to rtx_insn *.
2482 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2484 (frv_ifcvt_modify_insn): Likwise.
2485 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2486 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2487 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2488 as_a <rtx_insn *> cast to local "unprotected_region" once
2489 it's been established that it's not NULL or pc_rtx.
2490 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2491 param "sethi" from rtx to rtx_insn *.
2492 (nds32_group_float_insns): Likewise for param "insn".
2493 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2495 (vax_output_int_subtract): Likewise.
2496 * config/vax/vax.c (vax_output_int_add): Likewise for param
2498 (vax_output_int_subtract): Likewise.
2499 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2500 (emit_pattern_after): Likewise for param "after".
2501 (emit_insn_after): Likewise.
2502 (emit_jump_insn_after): Likewise.
2503 (emit_call_insn_after): Likewise.
2504 (emit_debug_insn_after): Likewise.
2505 (emit_pattern_before): Likewise for param "before".
2506 (emit_insn_before): Likewise.
2507 (emit_jump_insn_before): Likewise.
2508 * final.c (get_insn_template): Likewise for param "insn", removing
2510 * output.h (get_insn_template): Likewise for 2nd param.
2511 * rtl.h (emit_insn_before): Likewise.
2512 (emit_jump_insn_before): Likewise.
2513 (emit_debug_insn_before_noloc): Likewise.
2514 (emit_insn_after): Likewise.
2515 (emit_jump_insn_after): Likewise.
2516 (emit_call_insn_after): Likewise.
2517 (emit_debug_insn_after): Likewise.
2518 (set_insn_deleted): Likewise for param.
2520 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2523 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2524 floating point modes, so that IFmode is numerically greater than
2525 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2526 to declare the ordering. This prevents IFmode from being
2527 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2528 machine. Include rs6000-modes.h to share the fractional values
2529 between genmodes* and the rest of the compiler.
2533 * config/rs6000/rs6000-modes.h: New file.
2534 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2535 meaning of rs6000_long_double_size so that 126..128 selects an
2536 appropriate 128-bit floating point type.
2537 (rs6000_option_override_internal): Likewise.
2538 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2539 (TARGET_LONG_DOUBLE_128): Change the meaning of
2540 rs6000_long_double_size so that 126..128 selects an appropriate
2541 128-bit floating point type.
2542 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2543 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2544 source and destination to match the standard usage.
2545 (truncifkf2): Likewise.
2546 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2547 ISA 2.07 to use an explicit clobber, instead of passing in a
2549 (copysign<mode>3_soft): Likewise.
2551 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2553 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2554 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2555 (vect_slp_analyze_instance_dependence): Likewise.
2556 (vect_enhance_data_refs_alignment): Likewise.
2557 (vect_analyze_data_refs_alignment): Likewise.
2558 (vect_slp_analyze_and_verify_instance_alignment
2559 (vect_analyze_data_ref_accesses): Likewise.
2560 (vect_prune_runtime_alias_test_list): Likewise.
2561 (vect_analyze_data_refs): Likewise.
2562 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2563 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2564 (vect_analyze_scalar_cycles_1): Likewise.
2565 (vect_get_loop_niters): Likewise.
2566 (vect_analyze_loop_form_1): Likewise.
2567 (vect_update_vf_for_slp): Likewise.
2568 (vect_analyze_loop_operations): Likewise.
2569 (vect_analyze_loop): Likewise.
2570 (vectorizable_induction): Likewise.
2571 (vect_transform_loop): Likewise.
2572 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2573 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2574 (vect_make_slp_decision): Likewise.
2575 (vect_detect_hybrid_slp): Likewise.
2576 (vect_slp_analyze_operations): Likewise.
2577 (vect_slp_bb): Likewise.
2578 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2579 (vectorizable_bswap): Likewise.
2580 (vectorizable_call): Likewise.
2581 (vectorizable_simd_clone_call): Likewise.
2582 (vectorizable_conversion): Likewise.
2583 (vectorizable_assignment): Likewise.
2584 (vectorizable_shift): Likewise.
2585 (vectorizable_operation): Likewise.
2586 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2588 2018-06-18 Martin Sebor <msebor@redhat.com>
2590 PR tree-optimization/81384
2591 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2592 * builtins.c (expand_builtin_strnlen): New function.
2593 (expand_builtin): Call it.
2594 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2595 * builtins.def (BUILT_IN_STRNLEN): New.
2596 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2597 Warn for bounds in excess of maximum object size.
2598 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2599 single-value ranges. Handle strnlen.
2600 (handle_builtin_strlen): Handle strnlen.
2601 (strlen_check_and_optimize_stmt): Same.
2602 * doc/extend.texi (Other Builtins): Document strnlen.
2604 2018-06-18 Maya Rashish <coypu@sdf.org>
2606 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2607 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2608 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2610 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2612 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2614 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2616 * tree.c (escaped_string::escape): Replace cast to char * by
2617 const_cast<char *> (unescaped).
2619 2018-06-18 Nick Clifton <nickc@redhat.com>
2622 * tree.c (escaped_string): New class. Converts an unescaped
2623 string into its escaped equivalent.
2624 (warn_deprecated_use): Use the new class to convert the
2625 deprecation message, if present.
2626 (test_escaped_strings): New self test.
2627 (test_c_tests): Add test_escaped_strings.
2628 * doc/extend.texi (deprecated): Add a note that the
2629 deprecation message is affected by the -fmessage-length
2630 option, and that control characters will be escaped.
2631 (#pragma GCC error): Document this pragma.
2632 (#pragma GCC warning): Likewise.
2633 * doc/invoke.texi (-fmessage-length): Document this option's
2634 effect on the #warning and #error preprocessor directives and
2635 the deprecated attribute.
2637 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2639 * tree.c (decl_value_expr_lookup): Revert latest change.
2640 (decl_value_expr_insert): Likewise.
2642 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2644 * gimplify.c (nonlocal_vlas): Delete.
2645 (nonlocal_vla_vars): Likewise.
2646 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2648 (gimplify_body): Do not create and destroy nonlocal_vlas.
2649 * tree-nested.c: Include diagnostic.h.
2650 (use_pointer_in_frame): Tweak.
2651 (lookup_field_for_decl): Add assertion and declare the transformation.
2652 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2653 internal error when the reference is in a wrong context. Do not
2654 create a debug decl by default.
2655 (note_nonlocal_block_vlas): Delete.
2656 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2657 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2658 create a debug decl by default.
2659 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2660 call is in a wrong context.
2661 (fixup_vla_decls): New function.
2662 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2663 debug variables were created.
2664 * tree.c (decl_value_expr_lookup): Add checking assertion.
2665 (decl_value_expr_insert): Likewise.
2667 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2670 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2671 * tree-scalar-evolution.c (interpret_expr): Likewise.
2672 (expression_expensive_p): Likewise.
2673 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2674 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2675 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2676 (ssa_defined_by_minus_one_stmt_p): New.
2678 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2681 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2682 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2683 * dojump.c (do_jump): Likewise.
2684 * expr.c (expand_expr_real_2): Check operand type's sign.
2685 * fold-const.c (const_unop): Handle ABSU_EXPR.
2686 (fold_abs_const): Likewise.
2687 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2688 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2689 (strip_sign_op_1): Likesise.
2690 * match.pd: Add new pattern to generate ABSU_EXPR.
2691 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2692 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2693 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2694 * tree-inline.c (estimate_operator_cost): Likewise.
2695 * tree-pretty-print.c (dump_generic_node): Likewise.
2696 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2697 * tree.def (ABSU_EXPR): New.
2699 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2702 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2703 documented as preserved for backward compatibility only.
2704 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2706 PR rtl-optimization/86108
2707 * bb-reorder.c (create_forwarder_block): Renamed to ...
2708 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2709 jump from new landing pad to the second part.
2710 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2713 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2716 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2717 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2718 Only call store_expr for halves if the mode is the same.
2721 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2722 Fix up comment formatting.
2724 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2726 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2727 (typed_splay_tree::closure,
2728 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2729 (typed_splay_tree::typed_splay_tree,
2730 typed_splay_tree::operator =): Declared private.
2731 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2732 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2733 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2734 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2735 typed_splay_tree::splay_tree_splay,
2736 typed_splay_tree::splay_tree_foreach_helper,
2737 typed_splay_tree::splay_tree_insert,
2738 typed_splay_tree::splay_tree_remove,
2739 typed_splay_tree::splay_tree_lookup,
2740 typed_splay_tree::splay_tree_predecessor,
2741 typed_splay_tree::splay_tree_successor,
2742 typed_splay_tree::splay_tree_min,
2743 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2744 (typed_splay_tree::root, typed_splay_tree::comp,
2745 typed_splay_tree::delete_key,
2746 typed_splay_tree::delete_value): New data members.
2747 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2748 typed_splay_tree::remove.
2750 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2752 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2753 -mginv and -mno-ginv to the assembler.
2754 * config/mips/mips.opt (-mcrc): New option.
2756 * doc/invoke.text (-mcrc): Document.
2759 2018-06-15 Nick Clifton <nickc@redhat.com>
2762 * tree.c (escaped_string): New class. Converts an unescaped
2763 string into its escaped equivalent.
2764 (warn_deprecated_use): Use the new class to convert the
2765 deprecation message, if present.
2766 (test_escaped_strings): New self test.
2767 (test_c_tests): Add test_escaped_strings.
2768 * doc/extend.texi (deprecated): Add a note that the
2769 deprecation message is affected by the -fmessage-length
2770 option, and that control characters will be escaped.
2771 (#pragma GCC error): Document this pragma.
2772 (#pragma GCC warning): Likewise.
2773 * doc/invoke.texi (-fmessage-length): Document this option's
2774 effect on the #warning and #error preprocessor directives and
2775 the deprecated attribute.
2777 2018-06-15 Richard Biener <rguenther@suse.de>
2779 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2780 here, also noting vector size used.
2781 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2782 size used in MSG_OPTIMIZED_LOCATIONS dump.
2783 (pass_slp_vectorize::execute): Adjust.
2785 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2788 * config/arc/arc.c (arc_return_address_register): Fix
2791 2018-06-15 Richard Biener <rguenther@suse.de>
2794 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2795 leave useless conversion stripping to force_gimple_operand_gsi.
2796 (gimplify_build2): Likewise.
2797 (gimplify_build1): Likewise.
2799 2018-06-15 Richard Biener <rguenther@suse.de>
2802 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2803 before adjusting their block.
2805 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2807 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2808 multilibs for *-*-rtems*.
2809 * config/riscv/t-rtems: New file.
2811 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2814 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2815 unsigned_type_for returns NULL.
2818 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2819 subregs of multi-word pseudos unless the float mode has word size.
2821 2018-06-14 Richard Biener <rguenther@suse.de>
2824 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2825 duplicate and harmful type_hash_canon.
2826 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2828 2018-06-14 Richard Biener <rguenther@suse.de>
2831 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2834 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2836 * config/rtems.h (STDINT_LONG32): Define.
2838 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2839 Prachi Godbole <prachi.godbole@imgtec.com>
2841 * config/mips/mips-cpus.def: Define P6600.
2842 * config/mips/mips-tables.opt: Regenerate.
2843 * config/mips/mips.c (mips_ucbranch_type): New enum.
2844 (mips_rtx_cost_data): Add support for P6600.
2845 (mips_issue_rate): Likewise.
2846 (mips_multipass_dfa_lookahead): Likewise.
2847 (mips_avoid_hazard): Likewise.
2848 (mips_reorg_process_insns): Likewise.
2849 (mips_classify_branch_p6600): New function.
2850 * config/mips/mips.h (TUNE_P6600): New define.
2851 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2852 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2853 * config/mips/mips.md: Include p6600.md.
2854 (processor): Add p6600.
2855 * config/mips/p6600.md: New file.
2856 * doc/invoke.texi: Add p6600 to supported architectures.
2858 2018-06-13 Martin Sebor <msebor@redhat.com>
2860 PR tree-optimization/86114
2861 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2863 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2865 2018-06-13 Richard Biener <rguenther@suse.de>
2867 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2868 Properly set vector type of the intermediate stmt.
2869 * tree-vect-stmts.c (vectorizable_operation): The destination
2870 var always has vectype_out type.
2872 2018-06-13 Jeff Law <law@redhat.com>
2874 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2875 integer 0 for argument to print_rtl_with_bb.
2876 (rl78_reorg): Likewise.
2878 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2880 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2881 from rtx to rtx_insn *.
2882 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2884 (add_sched_insns_for_speculation): Likewise for local "target",
2885 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2886 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2887 from rtx_insn ** to rtx_code_label **.
2888 (reorg_emit_nops): Likewise.
2889 (c6x_reorg): Likewise for local "call_labels".
2890 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2892 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2893 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2894 the loops over LABEL_REFS.
2895 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2897 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2898 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2899 (split_branches): Strengthen local "olabel" from rtx to
2900 rtx_insn *, adding a safe_as_a cast.
2901 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2903 (add_insn_after): Likewise for first two params.
2904 (add_insn_before): Likewise.
2905 (remove_insn): Likewise for param.
2906 (emit_pattern_before_noloc): Likewise for second and third params.
2907 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2908 (emit_call_insn_before_noloc): Likewise.
2909 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2911 (emit_barrier_before): Likewise.
2912 (emit_label_before): Strengthen "label" param from "rtx" to
2913 "rtx_code_label *". Strengthen "before" param from "rtx" to
2915 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2917 (emit_pattern_after_noloc): Likewise.
2918 (emit_insn_after_noloc): Likewise.
2919 (emit_jump_insn_after_noloc): Likewise.
2920 (emit_call_insn_after_noloc): Likewise.
2921 (emit_debug_insn_after_noloc): Likewise.
2922 (emit_barrier_after): Likewise.
2923 (emit_label_after): Likewise for both params.
2924 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2925 "loc" param from "int" to "location_t".
2926 (emit_insn_after_setloc): Likewise.
2927 (emit_jump_insn_after_setloc): Likewise.
2928 (emit_call_insn_after_setloc): Likewise.
2929 (emit_debug_insn_after_setloc): Likewise.
2930 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2931 "loc" param from "int" to "location_t".
2932 (emit_pattern_before): Convert NULL_RTX to NULL.
2933 (emit_insn_before_setloc): Convert "loc" param from "int" to
2935 (emit_jump_insn_before_setloc): Likewise.
2936 (emit_call_insn_before_setloc): Likewise.
2937 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2938 rtx_insn *. Convert "loc" param from "int" to "location_t".
2939 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2940 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2941 Convert 3rd param from "int" to "location_t".
2942 (emit_barrier_before, emit_barrier_after, next_real_insn):
2943 Strengthen param from rtx to rtx_insn *.
2944 (emit_label_before): Strengthen 1st param from "rtx" to
2945 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2947 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2948 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2949 Strengthen 2nd param from "rtx" to "rtx_insn *".
2950 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2951 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2952 Likewise. Convert 3rd param from "int" to "location_t".
2953 (emit_label_after): Strengthen 1st param from "rtx" to
2955 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2957 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2958 from "rtx" to "rtx_insn *".
2960 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2962 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2963 bodies streamed in with -Q.
2964 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2965 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2966 * gimple-streamer-out.c: Include gimple-pretty-print.h
2967 (output_bb): Dump stmts streamed.
2968 * lto-section-out.c: Include print-tree.h
2969 (lto_begin_section): Dump sections created.
2970 (lto_output_decl_index): Dump decl encoded.
2971 * lto-streamer-out.c: Include print-tree.h
2972 (create_output_block): Dump output block created.
2973 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2974 (output_function): Dump function output.
2975 (output_constructor): Dump constructor streamed.
2976 (write_global_stream): Output indexes encoded.
2977 (produce_asm_for_decls): Dump streams encoded.
2978 * lto-streamer.c (streamer_dump_file): New global var.
2979 * lto-streamer.h (streamer_dump_file): Declare.
2980 * passes.c (ipa_write_summaries): Initialize streamer dump.
2981 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2984 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2987 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2988 offsets for register save directives. Emit a second batch of save
2989 directives, if need be, when the function accesses prior frames.
2991 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2993 * config/arc/fpu.md (fmasf4): Force operand to register.
2994 (fnmasf4): Likewise.
2996 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2998 * config/arc/arc-protos.h (arc_pad_return): Remove.
2999 * config/arc/arc.c (machine_function): Remove force_short_suffix
3001 (arc_print_operand): Adjust printing of '&'.
3002 (arc_verify_short): Remove conditional printing of short suffix.
3003 (arc_final_prescan_insn): Remove reference to size_reason.
3004 (pad_return): New function.
3005 (arc_reorg): Call pad_return.
3006 (arc_pad_return): Remove.
3007 (arc_init_machine_status): Remove reference to force_short_suffix.
3008 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3009 (attr length): When attribute iscompact is true force to 2
3010 regardless; in the case of maybe check if we want to force the
3011 instruction to have 4 bytes length.
3012 (nopv): Change it to generate 4 byte long nop as well.
3013 (blockage): New pattern.
3014 (simple_return): Remove call to arc_pad_return.
3015 (p_return_i): Likewise.
3017 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3019 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3021 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3023 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3026 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3028 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3029 for ARC700 and ARCv2.
3031 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3034 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3035 operands[2] instead of operands[1].
3038 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3040 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3041 case, check whether the outer register overlaps an unallocatable
3042 register, not just whether it fits the required class.
3044 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3046 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3048 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3049 (memrefs_conflict_p): Likewise.
3050 (init_alias_analysis): Likewise.
3051 * cfgexpand.c (expand_debug_expr): Likewise.
3052 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3053 * cse.c (fold_rtx): Likewise.
3054 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3055 * expr.c (emit_block_move_hints): Likewise.
3056 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3057 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3058 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3059 (emit_group_store): Likewise.
3060 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3061 to read the PRE/POST_MODIFY increment.
3062 * calls.c (store_one_arg): Use strip_offset.
3063 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3065 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3067 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3068 (simplify_binary_operation_1): Extend CONST_INT handling to
3070 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3071 than a HOST_WIDE_INT.
3072 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3074 (adjust_mems, add_stores): Update accodingly.
3075 (vt_canonicalize_addr): Track polynomial offsets.
3076 (emit_note_insn_var_location): Likewise.
3077 (vt_add_function_parameter): Likewise.
3078 (vt_initialize): Likewise.
3080 2018-06-12 Jeff Law <law@redhat.com>
3082 * config.gcc (alpha*-*-freebsd*): Remove.
3083 * config/alpha/freebsd.h: Remove.
3085 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3088 * spellcheck-tree.c (levenshtein_distance): Rename to...
3089 (get_edit_distance): ...this, and update for underlying renaming.
3090 * spellcheck-tree.h (levenshtein_distance): Rename to...
3091 (get_edit_distance): ...this.
3092 * spellcheck.c (levenshtein_distance): Rename to...
3093 (get_edit_distance): ...this. Convert from Levenshtein distance
3094 to Damerau-Levenshtein distance by supporting transpositions of
3095 adjacent characters. Rename "v1" to "v_next" and "v0" to
3097 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3098 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3099 update for underlying renaming.
3100 (selftest::levenshtein_distance_unit_test): Rename to...
3101 (selftest::test_get_edit_distance_unit): ...this, and update for
3102 underlying renaming.
3103 (selftest::test_find_closest_string): Add example from PR 69968
3104 where transposition helps
3105 (selftest::test_metric_conditions): Update for renaming.
3106 (selftest::test_metric_conditions): Likewise.
3107 (selftest::spellcheck_c_tests): Likewise.
3108 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3109 (get_edit_distance): ...this.
3110 (best_match::consider): Update for renaming.
3112 2018-06-12 Martin Sebor <msebor@redhat.com>
3114 PR tree-optimization/85259
3115 * builtins.c (compute_objsize): Handle constant offsets.
3116 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3117 true iff a warning has been issued.
3118 * gimple.h (gimple_nonartificial_location): New function.
3119 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3120 gimple_nonartificial_location and handle -Wno-system-headers.
3121 (handle_builtin_stxncpy): Same.
3123 2018-06-12 Martin Sebor <msebor@redhat.com>
3126 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3128 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3130 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3131 BUILTIN_VEC_XST entries for pointer to double and long long.
3133 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3136 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3138 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3141 2018-06-12 Martin Liska <mliska@suse.cz>
3143 * doc/options.texi: Document IntegerRange.
3145 2018-06-12 Martin Liska <mliska@suse.cz>
3147 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3148 * opt-functions.awk: Handle Deprecated flag.
3149 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3151 (read_cmdline_option): Report warning for a deprecated option.
3152 * opts.h (struct cl_option): Add new field cl_deprecated.
3153 (CL_ERR_DEPRECATED): New.
3155 2018-06-12 Martin Liska <mliska@suse.cz>
3157 * doc/options.texi: Document Deprecated option flag.
3159 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3161 * config/arc/arc-arch.h (arc_extras): New enum.
3162 (arc_cpu_t):Add field extra.
3163 (arc_cpu_types): Consider the extras.
3164 * config/arc/arc-cpus.def: Add extras info.
3165 * config/arc/arc-opts.h (processor_type): Consider extra field.
3166 * config/arc/arc.c (arc_override_options): Handle extra field.
3168 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3170 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3171 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3173 * config/arc/arc.c (arc_init): Update pic warning.
3174 (irq_range): Update irq range parsing warnings.
3175 (arc_override_options): Update various warning messages.
3176 (arc_handle_aux_attribute): Likewise.
3178 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3180 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3182 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3184 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3185 compiler messages directives.
3187 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3189 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3190 * config/mips/mips-tables.opt: Regenerate.
3191 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3193 * doc/invoke.texi: Document -march=i6500.
3195 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3197 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3198 (i6400_gpmul): Add cpu_unit.
3199 (i6400_gpdiv): Likewise.
3200 (i6400_msa_add_d): Update reservations.
3201 (i6400_msa_int_add) Likewise.
3202 (i6400_msa_short_logic3) Likewise.
3203 (i6400_msa_short_logic2) Likewise.
3204 (i6400_msa_short_logic) Likewise.
3205 (i6400_msa_move) Likewise.
3206 (i6400_msa_cmp) Likewise.
3207 (i6400_msa_short_float2) Likewise.
3208 (i6400_msa_div_d) Likewise.
3209 (i6400_msa_long_logic1) Likewise.
3210 (i6400_msa_long_logic2) Likewise.
3211 (i6400_msa_mult) Likewise.
3212 (i6400_msa_long_float2) Likewise.
3213 (i6400_msa_long_float4) Likewise.
3214 (i6400_msa_long_float5) Likewise.
3215 (i6400_msa_long_float8) Likewise.
3216 (i6400_fpu_fadd): Include frint type.
3217 (i6400_fpu_store): New define_insn_reservation.
3218 (i6400_fpu_load): Likewise.
3219 (i6400_fpu_move): Likewise.
3220 (i6400_fpu_fcmp): Likewise.
3221 (i6400_fpu_fmadd): Likewise.
3222 (i6400_int_mult): Include imul3nc type and update reservation.
3223 (i6400_int_div): Include idiv3 type and update reservation.
3224 (i6400_int_load): Update to check type not move_type.
3225 (i6400_int_store): Likewise.
3226 (i6400_int_prefetch): Set zero latency.
3228 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3230 * gcc.c: Document new %@{...} sequence.
3231 (LINK_COMMAND_SPEC): Use it for the -L switches.
3232 (cpp_unique_options): Use it for the -I switches.
3233 (at_file_argbuf): New global variable.
3234 (in_at_file): Likewise.
3235 (alloc_args): Create at_file_argbuf.
3236 (clear_args): Truncate at_file_argbuf.
3237 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3238 (open_at_file): New function.
3239 (close_at_file): Likewise.
3240 (create_at_file): Delete.
3241 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3244 (validate_switches_from_spec): Deal with %@{...} sequence.
3245 (validate_switches): Likewise.
3246 (driver::finalize): Call clear_args.
3248 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3250 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3252 2018-06-11 Martin Sebor <msebor@redhat.com>
3254 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3255 (Wno-class-memaccess): Add @opindex.
3256 (Wno-templates, Wno-multiple-inheritance): Same.
3257 (Wno-virtual-inheritance, Wno-namespaces): Same.
3258 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3259 (Wno-format-overflow, Wno-format-truncation): Same.
3260 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3261 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3262 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3263 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3264 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3265 (Wno-misspelled-isr): Same.
3267 2018-06-11 Martin Sebor <msebor@redhat.com>
3269 * PR tree-optimization/86083
3270 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3272 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3274 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3276 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3279 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3280 on the correct operand.
3281 (*movdi_internal64): Ditto.
3283 2018-06-11 Martin Liska <mliska@suse.cz>
3285 PR tree-optimization/86089
3286 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3288 2018-06-11 Julia Koval <julia.koval@intel.com>
3290 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3291 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3292 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3294 2018-06-11 Olivier Hainque <hainque@adacore.com>
3296 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3297 for Ada with strict dwarf2.
3299 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3302 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3305 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3307 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3309 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3311 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3314 2018-06-08 Martin Liska <mliska@suse.cz>
3316 * tree-cfg.h (debug_function): Fix argument type to match
3319 2018-06-08 Martin Liska <mliska@suse.cz>
3321 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3322 Remove usage of MPX-related (and removed) fields.
3323 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3325 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3327 * cfg.c (debug): Use TDF_NONE rather than 0.
3328 * cfghooks.c (debug): Likewise.
3329 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3330 (struct dump_option_value_info): Convert to...
3331 (struct kv_pair): ...this template type.
3332 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3334 (optinfo_verbosity_options): Likewise.
3335 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3337 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3338 than int for "optgroup_flags" param.
3339 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3341 (dump_dec): Likewise.
3342 (dump_finish): Use TDF_NONE rather than 0.
3343 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3344 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3345 than 0. Update for change to option_ptr.
3346 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3347 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3348 0. Update for changes to optinfo_verbosity_options and
3350 (opt_info_switch_p): Convert optgroup_flags from int to
3352 (dump_basic_block): Use dump_flags_t rather than int
3353 for "dump_kind" param.
3354 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3355 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3356 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3357 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3358 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3359 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3360 TDF_NONE): Convert from macros to...
3361 (enum dump_flag): ...this new enum.
3362 (dump_flags_t): Update to use enum.
3363 (operator|, operator&, operator~, operator|=, operator&=):
3364 Implement for dump_flags_t.
3365 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3366 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3367 Convert from macros to...
3368 (enum optgroup_flag): ...this new enum.
3369 (optgroup_flags_t): New typedef.
3370 (operator|, operator|=): Implement for optgroup_flags_t.
3371 (struct dump_file_info): Convert field "alt_flags" to
3372 dump_flags_t. Convert field "optgroup_flags" to
3374 (dump_basic_block): Use dump_flags_t rather than int for param.
3375 (dump_generic_expr_loc): Likewise.
3376 (dump_dec): Likewise.
3377 (dump_register): Convert param "optgroup_flags" to
3379 (opt_info_enable_passes): Likewise.
3380 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3382 * gimple-pretty-print.c (debug): Likewise.
3383 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3384 (merged_store_group::apply_stores): Likewise.
3385 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3386 * gimple.c (verify_gimple_pp): Likewise.
3387 * graphite-poly.c (print_pbb_body): Likewise.
3388 * passes.c (pass_manager::register_one_dump_file): Convert
3389 local "optgroup_flags" to optgroup_flags_t.
3390 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3392 (debug_body): Likewise.
3393 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3394 to optgroup_flags_t.
3395 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3397 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3398 (convert_mult_to_fma): Likewise.
3399 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3400 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3401 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3402 "dump_kind" to dump_flags_t.
3404 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3406 * config/rs6000/rs6000.c (min, max): Delete.
3408 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3410 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3411 -mabi=spe and -mabi=no-spe.
3413 2018-06-08 Martin Liska <mliska@suse.cz>
3415 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3416 where we expect an existing summary.
3418 2018-06-08 Martin Liska <mliska@suse.cz>
3420 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3421 * ipa-inline.h (estimate_edge_growth): Likewise.
3423 2018-06-08 Martin Liska <mliska@suse.cz>
3425 * cgraph.c (function_version_hasher::hash): Use
3426 cgraph_node::get_uid ().
3427 (function_version_hasher::equal):
3428 * cgraph.h (cgraph_node::get_uid): New method.
3429 * ipa-inline.c (update_caller_keys): Use
3430 cgraph_node::get_uid ().
3431 (update_callee_keys): Likewise.
3432 * ipa-utils.c (searchc): Likewise.
3433 (ipa_reduced_postorder): Likewise.
3434 * lto-cgraph.c (input_node): Likewise.
3435 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3436 * symbol-summary.h (symtab_insertion): Likewise.
3437 (symtab_removal): Likewise.
3438 (symtab_duplication): Likewise.
3439 * tree-pretty-print.c (dump_function_header): Likewise.
3440 * tree-sra.c (convert_callers_for_node): Likewise.
3442 2018-06-08 Martin Liska <mliska@suse.cz>
3444 * cgraph.c (symbol_table::create_edge): Always assign a new
3446 (symbol_table::free_edge): Do not recycle numbers.
3447 * cgraph.h (cgraph_edge::get): New method.
3448 * symbol-summary.h (symtab_removal): Use it.
3449 (symtab_duplication): Likewise.
3450 (call_summary::hashable_uid): Remove.
3452 2018-06-08 Martin Liska <mliska@suse.cz>
3454 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3455 (initialize_growth_caches): Remove.
3456 (free_growth_caches): Likewise.
3457 (do_estimate_edge_time): Use edge_growth_cache.
3458 (do_estimate_edge_size): Likewise.
3459 (do_estimate_edge_hints): Likewise.
3460 * ipa-inline.c (reset_edge_caches): Likewise.
3461 (recursive_inlining): Likewise.
3462 (inline_small_functions): Likewise.
3463 * ipa-inline.h (initialize_growth_caches): Remove.
3464 (estimate_edge_size): Likewise.
3465 (estimate_edge_time): Likewise.
3466 (estimate_edge_hints): Likewise.
3467 (reset_edge_growth_cache): Likewise.
3468 * symbol-summary.h (call_summary::remove): New method.
3470 2018-06-08 Martin Liska <mliska@suse.cz>
3472 * ipa-cp.c (class edge_clone_summary): New summary.
3473 (grow_edge_clone_vectors): Remove.
3474 (ipcp_edge_duplication_hook): Remove.
3475 (class edge_clone_summary_t): New call_summary class.
3476 (ipcp_edge_removal_hook): Remove.
3477 (edge_clone_summary_t::duplicate): New function.
3478 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3479 (create_specialized_node): Likewise.
3480 (ipcp_driver): Initialize edge_clone_summaries and do not
3483 2018-06-08 Martin Liska <mliska@suse.cz>
3485 * symbol-summary.h (get): New function.
3486 (call_summary::m_initialize_when_cloning): New class member.
3488 2018-06-08 Martin Liska <mliska@suse.cz>
3490 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3491 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3492 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3493 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3494 (remove_cgraph_node_from_order): Use the removed_nodes set.
3495 (do_per_function_toporder): Likwise.
3496 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3497 instead of summary_uid.
3498 (symtab_removal): Likewise.
3499 (symtab_duplication): Likewise.
3501 2018-06-08 Martin Liska <mliska@suse.cz>
3503 * ipa-cp.c (ipcp_store_bits_results): Use
3504 ipcp_transformation_sum.
3505 (ipcp_store_vr_results): Likewise.
3506 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3508 (ipcp_transformation_initialize): ... this.
3509 (ipa_set_node_agg_value_chain):
3510 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3511 (write_ipcp_transformation_info): Likewise.
3512 (read_ipcp_transformation_info): Likewise.
3513 (ipcp_update_bits): Likewise.
3514 (ipcp_update_vr): Likewise.
3515 (ipcp_transform_function): Likewise.
3516 * ipa-prop.h: Rename ipcp_transformation_summary to
3517 ipcp_transformation.
3518 (class ipcp_transformation_t): New function summary.
3519 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3520 (ipa_get_agg_replacements_for_node): Likewise.
3522 2018-06-08 Martin Liska <mliska@suse.cz>
3524 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3526 (class funct_state_summary_t): New function_summary class.
3527 (has_function_state): Remove.
3528 (get_function_state): Likewise.
3529 (set_function_state): Likewise.
3530 (add_new_function): Likewise.
3531 (funct_state_summary_t::insert): New function.
3532 (duplicate_node_data): Remove.
3533 (remove_node_data): Remove.
3534 (funct_state_summary_t::duplicate): New function.
3535 (register_hooks): Create new funct_state_summaries.
3536 (pure_const_generate_summary): Use it.
3537 (pure_const_write_summary): Likewise.
3538 (pure_const_read_summary): Likewise.
3539 (propagate_pure_const): Likewise.
3540 (propagate_nothrow): Likewise.
3541 (dump_malloc_lattice): Likewise.
3542 (propagate_malloc): Likewise.
3543 (execute): Do not register hooks, just remove summary
3545 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3548 2018-06-08 Martin Liska <mliska@suse.cz>
3550 * ipa-reference.c (remove_node_data): Remove.
3551 (duplicate_node_data): Likewise.
3552 (class ipa_ref_var_info_summary_t): New class.
3553 (class ipa_ref_opt_summary_t): Likewise.
3554 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3555 (get_reference_optimization_summary): Use
3556 ipa_ref_opt_sum_summaries.
3557 (set_reference_vars_info): Remove.
3558 (set_reference_optimization_summary): Likewise.
3559 (ipa_init): Create summaries.
3560 (init_function_info): Use function summary.
3561 (ipa_ref_opt_summary_t::duplicate): New function.
3562 (ipa_ref_opt_summary_t::remove): New function.
3563 (get_read_write_all_from_node): Fix GNU coding style.
3564 (propagate): Use function summary.
3565 (write_node_summary_p): Fix GNU coding style.
3566 (stream_out_bitmap): Likewise.
3567 (ipa_reference_read_optimization_summary): Use function summary.
3568 (ipa_reference_c_finalize): Do not release hooks.
3570 2018-06-08 Martin Liska <mliska@suse.cz>
3572 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3573 (analyze_function_body): Extract multiple calls of get_create.
3574 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3575 * ipa-inline.c (recursive_inlining): Use ::get method.
3576 * ipa-inline.h (estimate_edge_growth): Likewise.
3578 2018-06-08 Martin Liska <mliska@suse.cz>
3580 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3582 (hsa_function_summary::hsa_function_summary): Use the new enum
3584 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3585 * hsa-gen.c (hsa_get_host_function): Likewise.
3586 (get_brig_function_name): Likewise.
3587 * ipa-hsa.c (process_hsa_functions): Likewise.
3588 (ipa_hsa_write_summary): Likewise.
3589 * symbol-summary.h (symtab_duplication): Use ::get function/
3590 (get): New function.
3592 2018-06-08 Martin Liska <mliska@suse.cz>
3594 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3596 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3597 (hsa_register_kernel): Likewise.
3598 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3599 * hsa-gen.c (hsa_get_host_function): Likewise.
3600 (get_brig_function_name): Likewise.
3601 (generate_hsa): Likewise.
3602 (pass_gen_hsail::execute): Likewise.
3603 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3604 (devirtualization_time_bonus): Likewise.
3605 (ipcp_propagate_stage): Likewise.
3606 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3607 (edge_set_predicate): Likewise.
3608 (evaluate_conditions_for_known_args): Likewise.
3609 (evaluate_properties_for_edge): Likewise.
3610 (ipa_fn_summary::reset): Likewise.
3611 (ipa_fn_summary_t::duplicate): Likewise.
3612 (dump_ipa_call_summary): Likewise.
3613 (ipa_dump_fn_summary): Likewise.
3614 (analyze_function_body): Likewise.
3615 (compute_fn_summary): Likewise.
3616 (estimate_edge_devirt_benefit): Likewise.
3617 (estimate_edge_size_and_time): Likewise.
3618 (estimate_calls_size_and_time): Likewise.
3619 (estimate_node_size_and_time): Likewise.
3620 (inline_update_callee_summaries): Likewise.
3621 (remap_edge_change_prob): Likewise.
3622 (remap_edge_summaries): Likewise.
3623 (ipa_merge_fn_summary_after_inlining): Likewise.
3624 (ipa_update_overall_fn_summary): Likewise.
3625 (read_ipa_call_summary): Likewise.
3626 (inline_read_section): Likewise.
3627 (write_ipa_call_summary): Likewise.
3628 (ipa_fn_summary_write): Likewise.
3629 (ipa_free_fn_summary): Likewise.
3630 * ipa-hsa.c (process_hsa_functions): Likewise.
3631 (ipa_hsa_write_summary): Likewise.
3632 (ipa_hsa_read_section): Likewise.
3633 * ipa-icf.c (sem_function::merge): Likewise.
3634 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3635 (do_estimate_edge_time): Likewise.
3636 (estimate_size_after_inlining): Likewise.
3637 (estimate_growth): Likewise.
3638 (growth_likely_positive): Likewise.
3639 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3640 (inline_call): Likewise.
3641 * ipa-inline.c (caller_growth_limits): Likewise.
3642 (can_inline_edge_p): Likewise.
3643 (can_inline_edge_by_limits_p): Likewise.
3644 (compute_uninlined_call_time): Likewise.
3645 (compute_inlined_call_time): Likewise.
3646 (want_inline_small_function_p): Likewise.
3647 (edge_badness): Likewise.
3648 (update_caller_keys): Likewise.
3649 (update_callee_keys): Likewise.
3650 (recursive_inlining): Likewise.
3651 (inline_small_functions): Likewise.
3652 (inline_to_all_callers_1): Likewise.
3653 (dump_overall_stats): Likewise.
3654 (early_inline_small_functions): Likewise.
3655 (early_inliner): Likewise.
3656 * ipa-inline.h (estimate_edge_growth): Likewise.
3657 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3658 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3659 * ipa-prop.h (IPA_NODE_REF): Likewise.
3660 (IPA_EDGE_REF): Likewise.
3661 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3662 (propagate_malloc): Likewise.
3663 * ipa-split.c (execute_split_functions): Likewise.
3664 * symbol-summary.h: Rename get to get_create.
3666 (get_create): Likewise.
3667 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3669 2018-06-08 Martin Liska <mliska@suse.cz>
3671 * symbol-summary.h (release): Move definition out of class
3673 (symtab_removal): Likewise.
3674 (symtab_duplication): Likewise.
3676 2018-06-08 Martin Liska <mliska@suse.cz>
3678 * symbol-summary.h (function_summary): Move constructor
3679 implementation out of class declaration.
3680 (release): Likewise.
3681 (symtab_insertion): Likewise.
3682 (symtab_removal): Likewise.
3683 (symtab_duplication): Likewise.
3686 2018-06-08 Martin Liska <mliska@suse.cz>
3688 * Makefile.in: Remove support for MPX (macros, related functions,
3689 fields in cgraph_node, ...).
3690 * builtin-types.def (BT_BND): Likewise.
3691 (BT_FN_BND_CONST_PTR): Likewise.
3692 (BT_FN_CONST_PTR_BND): Likewise.
3693 (BT_FN_VOID_PTR_BND): Likewise.
3694 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3695 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3696 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3697 (expand_builtin_mempcpy_with_bounds): Likewise.
3698 (expand_builtin_memset_with_bounds): Likewise.
3699 (expand_builtin_memset_args): Likewise.
3700 (std_expand_builtin_va_start): Likewise.
3701 (expand_builtin): Likewise.
3702 (expand_builtin_with_bounds): Likewise.
3703 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3704 (DEF_LIB_BUILTIN_CHKP): Likewise.
3705 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3706 (DEF_CHKP_BUILTIN): Likewise.
3707 (BUILT_IN_MEMCPY): Likewise.
3708 (BUILT_IN_MEMMOVE): Likewise.
3709 (BUILT_IN_MEMPCPY): Likewise.
3710 (BUILT_IN_MEMSET): Likewise.
3711 (BUILT_IN_STPCPY): Likewise.
3712 (BUILT_IN_STRCAT): Likewise.
3713 (BUILT_IN_STRCHR): Likewise.
3714 (BUILT_IN_STRCPY): Likewise.
3715 (BUILT_IN_STRLEN): Likewise.
3716 (BUILT_IN_MEMCPY_CHK): Likewise.
3717 (BUILT_IN_MEMMOVE_CHK): Likewise.
3718 (BUILT_IN_MEMPCPY_CHK): Likewise.
3719 (BUILT_IN_MEMSET_CHK): Likewise.
3720 (BUILT_IN_STPCPY_CHK): Likewise.
3721 (BUILT_IN_STRCAT_CHK): Likewise.
3722 (BUILT_IN_STRCPY_CHK): Likewise.
3723 * calls.c (store_bounds): Likewise.
3724 (emit_call_1): Likewise.
3725 (special_function_p): Likewise.
3726 (maybe_warn_nonstring_arg): Likewise.
3727 (initialize_argument_information): Likewise.
3728 (finalize_must_preallocate): Likewise.
3729 (compute_argument_addresses): Likewise.
3730 (expand_call): Likewise.
3731 * cfgexpand.c (expand_call_stmt): Likewise.
3732 (expand_return): Likewise.
3733 (expand_gimple_stmt_1): Likewise.
3734 (pass_expand::execute): Likewise.
3735 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3736 (cgraph_node::remove): Likewise.
3737 (cgraph_node::dump): Likewise.
3738 (cgraph_node::verify_node): Likewise.
3739 * cgraph.h (chkp_function_instrumented_p): Likewise.
3740 (symtab_node::get_alias_target): Likewise.
3741 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3742 (cgraph_local_p): Likewise.
3743 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3744 (cgraph_edge::rebuild_references): Likewise.
3745 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3746 (walk_polymorphic_call_targets): Likewise.
3747 (cgraph_node::expand_thunk): Likewise.
3748 (symbol_table::output_weakrefs): Likewise.
3749 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3750 (ix86_handle_option): Likewise.
3751 * config/i386/constraints.md: Likewise.
3752 * config/i386/i386-builtin-types.def (BND): Likewise.
3756 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3757 (BDESC_FIRST): Likewise.
3759 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3760 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3761 * config/i386/i386.c (enum reg_class): Likewise.
3762 (ix86_target_string): Likewise.
3763 (ix86_option_override_internal): Likewise.
3764 (ix86_conditional_register_usage): Likewise.
3765 (ix86_valid_target_attribute_inner_p): Likewise.
3766 (ix86_set_indirect_branch_type): Likewise.
3767 (ix86_set_current_function): Likewise.
3768 (ix86_function_arg_regno_p): Likewise.
3769 (init_cumulative_args): Likewise.
3770 (ix86_function_arg_advance): Likewise.
3771 (ix86_function_arg): Likewise.
3772 (ix86_pass_by_reference): Likewise.
3773 (ix86_function_value_regno_p): Likewise.
3774 (ix86_function_value_1): Likewise.
3775 (ix86_function_value_bounds): Likewise.
3776 (ix86_return_in_memory): Likewise.
3777 (ix86_setup_incoming_vararg_bounds): Likewise.
3778 (ix86_va_start): Likewise.
3779 (indirect_thunk_need_prefix): Likewise.
3780 (print_reg): Likewise.
3781 (ix86_print_operand): Likewise.
3782 (ix86_expand_call): Likewise.
3783 (ix86_output_function_return): Likewise.
3784 (reg_encoded_number): Likewise.
3785 (BDESC_VERIFYS): Likewise.
3786 (ix86_init_mpx_builtins): Likewise.
3787 (ix86_init_builtins): Likewise.
3788 (ix86_emit_cmove): Likewise.
3789 (ix86_emit_move_max): Likewise.
3790 (ix86_expand_builtin): Likewise.
3791 (ix86_builtin_mpx_function): Likewise.
3792 (ix86_get_arg_address_for_bt): Likewise.
3793 (ix86_load_bounds): Likewise.
3794 (ix86_store_bounds): Likewise.
3795 (ix86_load_returned_bounds): Likewise.
3796 (ix86_store_returned_bounds): Likewise.
3797 (ix86_class_likely_spilled_p): Likewise.
3798 (ix86_hard_regno_mode_ok): Likewise.
3799 (x86_order_regs_for_local_alloc): Likewise.
3800 (ix86_mitigate_rop): Likewise.
3801 (ix86_bnd_prefixed_insn_p): Likewise.
3802 (ix86_mpx_bound_mode): Likewise.
3803 (ix86_make_bounds_constant): Likewise.
3804 (ix86_initialize_bounds): Likewise.
3805 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3806 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3807 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3808 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3809 (TARGET_CHKP_BOUND_MODE): Likewise.
3810 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3811 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3812 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3813 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3814 * config/i386/i386.h (TARGET_MPX): Likewise.
3815 (TARGET_MPX_P): Likewise.
3816 (VALID_BND_REG_MODE): Likewise.
3817 (FIRST_BND_REG): Likewise.
3818 (LAST_BND_REG): Likewise.
3819 (enum reg_class): Likewise.
3820 (BND_REG_P): Likewise.
3821 (BND_REGNO_P): Likewise.
3822 (BNDmode): Likewise.
3823 (ADJUST_INSN_LENGTH): Likewise.
3824 * config/i386/i386.md: Likewise.
3825 * config/i386/i386.opt: Likewise.
3826 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3827 (defined): Likewise.
3828 (LINK_MPX): Likewise.
3829 (MPX_SPEC): Likewise.
3830 (LIBMPX_SPEC): Likewise.
3831 (LIBMPXWRAPPERS_SPEC): Likewise.
3832 (CHKP_SPEC): Likewise.
3833 * config/i386/predicates.md: Likewise.
3834 * dbxout.c (dbxout_type): Likewise.
3835 * doc/extend.texi: Likewise.
3836 * doc/invoke.texi: Likewise.
3837 * doc/md.texi: Likewise.
3838 * doc/tm.texi: Likewise.
3839 * doc/tm.texi.in: Likewise.
3840 * dwarf2out.c (is_base_type): Likewise.
3841 (gen_formal_types_die): Likewise.
3842 (gen_subprogram_die): Likewise.
3843 (gen_type_die_with_usage): Likewise.
3844 (gen_decl_die): Likewise.
3845 (dwarf2out_late_global_decl): Likewise.
3846 * expr.c (expand_assignment): Likewise.
3847 (emit_storent_insn): Likewise.
3848 (store_expr_with_bounds): Likewise.
3849 (store_expr): Likewise.
3850 (expand_expr_real_1): Likewise.
3851 * expr.h (store_expr_with_bounds): Likewise.
3852 * function.c (use_register_for_decl): Likewise.
3853 (struct bounds_parm_data): Likewise.
3854 (assign_parms_augmented_arg_list): Likewise.
3855 (assign_parm_find_entry_rtl): Likewise.
3856 (assign_parm_is_stack_parm): Likewise.
3857 (assign_parm_load_bounds): Likewise.
3858 (assign_bounds): Likewise.
3859 (assign_parms): Likewise.
3860 (expand_function_start): Likewise.
3861 * gcc.c (CHKP_SPEC): Likewise.
3862 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3863 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3864 (wrestrict_dom_walker::check_call): Likewise.
3865 * gimple.c (gimple_build_call_from_tree): Likewise.
3866 * gimple.h (enum gf_mask): Likewise.
3867 (gimple_call_with_bounds_p): Likewise.
3868 (gimple_call_set_with_bounds): Likewise.
3869 * gimplify.c (gimplify_init_constructor): Likewise.
3870 * ipa-cp.c (initialize_node_lattices): Likewise.
3871 (propagate_constants_across_call): Likewise.
3872 (find_more_scalar_values_for_callers_subset): Likewise.
3873 * ipa-hsa.c (process_hsa_functions): Likewise.
3874 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3875 * ipa-icf.c (sem_function::merge): Likewise.
3876 * ipa-inline.c (early_inliner): Likewise.
3877 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3878 (warn_function_cold): Likewise.
3879 (propagate_pure_const): Likewise.
3880 * ipa-ref.h (enum GTY): Likewise.
3881 * ipa-split.c (find_retbnd): Likewise.
3882 (consider_split): Likewise.
3883 (split_function): Likewise.
3884 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3885 * ipa.c (walk_polymorphic_call_targets): Likewise.
3886 (symbol_table::remove_unreachable_nodes): Likewise.
3887 (process_references): Likewise.
3888 (cgraph_build_static_cdtor_1): Likewise.
3889 * lto-cgraph.c (lto_output_node): Likewise.
3890 (output_refs): Likewise.
3891 (compute_ltrans_boundary): Likewise.
3892 (input_overwrite_node): Likewise.
3893 (input_node): Likewise.
3894 (input_cgraph_1): Likewise.
3895 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3896 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3897 (class pass_chkp_instrumentation_passes): Likewise.
3898 (make_pass_chkp_instrumentation_passes): Likewise.
3899 * passes.def: Likewise.
3900 * rtl.h (struct GTY): Likewise.
3901 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3902 * stor-layout.c (layout_type): Likewise.
3903 * symtab.c: Likewise.
3904 * target.def: Likewise.
3905 * targhooks.c (default_chkp_bound_type): Likewise.
3906 (default_chkp_bound_mode): Likewise.
3907 (default_builtin_chkp_function): Likewise.
3908 (default_chkp_function_value_bounds): Likewise.
3909 (default_chkp_make_bounds_constant): Likewise.
3910 (default_chkp_initialize_bounds): Likewise.
3911 * targhooks.h (default_chkp_bound_type): Likewise.
3912 (default_chkp_bound_mode): Likewise.
3913 (default_builtin_chkp_function): Likewise.
3914 (default_chkp_function_value_bounds): Likewise.
3915 (default_chkp_make_bounds_constant): Likewise.
3916 (default_chkp_initialize_bounds): Likewise.
3917 * toplev.c (compile_file): Likewise.
3918 (process_options): Likewise.
3919 * tree-core.h (DEF_BUILTIN): Likewise.
3920 (DEF_BUILTIN_CHKP): Likewise.
3921 * tree-inline.c (declare_return_variable): Likewise.
3922 (remap_gimple_stmt): Likewise.
3923 (copy_bb): Likewise.
3924 (initialize_inlined_parameters): Likewise.
3925 (expand_call_inline): Likewise.
3926 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3927 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3928 (make_pass_ipa_chkp_produce_thunks): Likewise.
3929 (make_pass_chkp): Likewise.
3930 (make_pass_chkp_opt): Likewise.
3931 (make_pass_chkp_instrumentation_passes): Likewise.
3932 * tree-pretty-print.c (dump_generic_node): Likewise.
3933 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3934 * tree-ssa-dce.c (propagate_necessity): Likewise.
3935 (eliminate_unnecessary_stmts): Likewise.
3936 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3937 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3938 * tree-ssa-sccvn.h: Likewise.
3939 * tree-ssa-strlen.c (get_string_length): Likewise.
3940 (valid_builtin_call): Likewise.
3941 (adjust_last_stmt): Likewise.
3942 (handle_builtin_strchr): Likewise.
3943 (handle_builtin_strcpy): Likewise.
3944 (handle_builtin_stxncpy): Likewise.
3945 (handle_builtin_memcpy): Likewise.
3946 (handle_builtin_strcat): Likewise.
3947 (strlen_check_and_optimize_stmt): Likewise.
3948 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3949 * tree-streamer-in.c: Likewise.
3950 * tree-streamer.c (record_common_node): Likewise.
3951 * tree.c (tree_code_size): Likewise.
3952 (wide_int_to_tree_1): Likewise.
3953 (type_contains_placeholder_1): Likewise.
3954 (build_common_tree_nodes): Likewise.
3955 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3956 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3957 (POINTER_BOUNDS_P): Likewise.
3958 (BOUNDED_TYPE_P): Likewise.
3959 (BOUNDED_P): Likewise.
3960 (CALL_WITH_BOUNDS_P): Likewise.
3961 (pointer_bounds_type_node): Likewise.
3962 * value-prof.c (gimple_ic): Likewise.
3963 * var-tracking.c (vt_add_function_parameters): Likewise.
3964 * varasm.c (make_decl_rtl): Likewise.
3965 (assemble_start_function): Likewise.
3966 (output_constant): Likewise.
3967 (maybe_assemble_visibility): Likewise.
3968 * varpool.c (ctor_for_folding): Likewise.
3969 * chkp-builtins.def: Remove.
3970 * ipa-chkp.c: Remove.
3971 * ipa-chkp.h: Remove.
3972 * rtl-chkp.c: Remove.
3973 * rtl-chkp.h: Remove.
3974 * tree-chkp-opt.c: Remove.
3975 * tree-chkp.c: Remove.
3976 * tree-chkp.h: Remove.
3978 2018-06-07 Carl Love <cel@us.ibm.com>
3980 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3981 vextract_fp_from_shortl): Add BE support.
3983 2018-06-07 Paul Koning <ni1d@arrl.net>
3985 * compare-elim.c (try_merge_compare): Don't merge compare if
3986 address contains a side effect.
3987 (try_eliminate_compare): Likewise.
3989 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3991 * config.gcc: Support "tremont".
3992 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3993 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3995 * config/i386/i386.c (m_TREMONT): Define.
3996 (processor_target_table): Add "tremont".
3997 (PTA_TREMONT): Define.
3998 (ix86_lea_outperforms): Add TARGET_TREMONT.
3999 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4000 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4001 and M_INTEL_GOLDMONT_PLUS.
4002 (fold_builtin_cpu): Add "tremont".
4003 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4004 (ix86_option_override_internal): Add "tremont".
4005 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4006 (processor_type): Add PROCESSOR_TREMONT.
4007 * config/i386/x86-tune.def: Add m_TREMONT.
4008 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4010 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4012 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4013 symbol defined for msp430i* devices to be lower case.
4015 gcc/testsuite/gcc.target/msp430/
4016 * msp430i-device-symbol.c: New test.
4017 * msp430f-device-symbol.c: New test.
4018 * msp430.h: New test header file.
4020 2018-06-07 Richard Biener <rguenther@suse.de>
4022 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4023 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4024 Properly wrap signed arithmetic if overflow wraps.
4026 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4028 PR tree-optimization/69615
4029 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4030 of a cast from a same precision integral SSA_NAME in a bb dominated
4031 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4032 cast to utype if rhs2 has already a compatible type.
4034 2018-06-07 Richard Biener <rguenther@suse.de>
4036 PR tree-optimization/85935
4037 * graphite-scop-detection.c (find_params_in_bb): Analyze
4038 condition operands with respect to the correct loop. Assert
4039 the analysis doesn't fail.
4041 2018-06-04 Carl Love <cel@us.ibm.com>
4043 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4044 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4045 as it is slightly cheaper.
4046 (first_match_or_eos_index_<mode>):
4047 Calculate index using natural element order.
4048 (first_match_index_<mode>):
4049 Calculate index using natural element order.
4050 (first_match_or_eos_index_<mode>):
4051 Calculate index using natural order.
4052 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4053 for BE and LE modes.
4054 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4055 P9V_BUILTIN_VCLZLSBB_V16QI.
4056 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4059 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4061 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4062 indentation and line wrap for many prototypes. Add missing
4063 @smallexample directives around block of prototypes for vec_xl and
4066 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4068 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4069 track if we pass or return IEEE 128-bit floating point.
4070 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4071 C++ mangling that is compatible with GCC 8.1.
4072 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4073 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4074 floating point types.
4075 (rs6000_function_arg_advance_1): Likewise.
4076 (rs6000_mangle_type): Optionally generate mangled names that match
4077 what GCC 8.1 generated for IEEE 128-bit floating point types.
4078 (rs6000_globalize_decl_name): If we have an external function that
4079 passes or returns IEEE 128-bit types, generate a weak reference
4080 from the mangled name used in GCC 8.1 to the current mangled
4082 (rs6000_init_builtins): Make __ibm128 use the long double type if
4083 long double is IBM extended double. Make __float128 use the long
4084 double type if long double is IEEE 128-bit.
4087 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4088 macro for __ibm128 built-in functions.
4089 (PACK_IF): Add __ibm128 pack/unpack functions.
4090 (UNPACK_IF): Likewise.
4091 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4092 enable long double built-in functions if long double is IEEE
4093 128-bit floating point.
4094 (rs6000_invalid_builtin): Update long double built-in function
4096 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4097 functions, adjust the built-in function to use the long double
4098 built-in function if __ibm128 and long double are the same type.
4099 * doc/extend.texi (PowerPC builtins): Update documention for
4100 __builtin_{,un}pack_longdouble. Add documentation for
4101 __builtin_{,un}pack_ibm128.
4103 2018-06-06 Jim Wilson <jimw@sifive.com>
4105 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4106 (struct machine_function): New field interrupt_mode.
4107 (riscv_handle_type_attribute): New function. Add forward declaration.
4108 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4109 (riscv_expand_epilogue): Check interrupt_mode field.
4110 (riscv_set_current_function): Check interrupt attribute args and
4111 set interrupt_mode field.
4112 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4113 (riscv_sret, riscv_uret): New.
4114 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4115 new arguments to interrupt attribute.
4117 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4120 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4121 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4123 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4125 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4126 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4127 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4128 several redundant entries.
4130 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4132 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4133 type from "rtx" to "rtx_insn *".
4134 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4135 for local "call_insn", removing cast.
4136 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4138 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4140 PR tree-optimization/86066
4141 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4142 for BIT_INSERT_EXPR stores.
4144 2018-06-06 Richard Biener <rguenther@suse.de>
4146 PR tree-optimization/86062
4147 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4148 component refs ontop
4149 of to be offsetted base.
4151 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4153 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4154 to be static and remove check on interrupt attribute name.
4156 gcc/testsuite/gcc.target/msp430/
4157 * function-attributes-4.c: New test.
4158 * static-interrupts.c: New test.
4160 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4162 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4163 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4165 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4168 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4170 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4171 Remove second argument, change how error is called.
4172 (aarch64_layout_arg): Remove second argument from
4173 aarch64_err_no_fpadvsimd call.
4174 (aarch64_init_cumulative_args): Ditto.
4175 (aarch64_gimplify_va_arg_expr): Ditto.
4176 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4178 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4180 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4181 (*simple_return_indirect_internal<mode>): Rename from
4182 simple_return_indirect_internal. Use W mode iterator.
4183 (rstorssp): New expander.
4184 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4185 (clrssbsy): New expander.
4186 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4188 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4190 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4192 (cmse_check_pointed_object): Likewise.
4194 2018-06-05 Martin Liska <mliska@suse.cz>
4196 PR gcov-profile/47618
4197 * doc/invoke.texi: Document how -fprofile-dir format
4200 2018-06-05 Richard Biener <rguenther@suse.de>
4202 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4203 removal pretend DOM info isn't available so we do not update
4204 it and only remove edges, not dominated blocks. Actually free
4205 DOM info in case we removed something. Remove unreachable blocks.
4206 (mfb_keep_latches): Work with either DOM info or marked backedges.
4207 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4208 first. Mark backedges if DOM info isn't available.
4209 (Re-)compute DOM info after cleanup_control_flow_pre.
4211 2018-06-05 Richard Biener <rguenther@suse.de>
4213 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4214 (locus_discrim_hasher::hash): Adjust.
4215 (locus_discrim_hasher::equal): Likewise.
4216 (next_discriminator_for_locus): Work on line directly.
4217 (same_line_p): Pass in expanded locus1 as well.
4218 (assign_discriminators): Avoid redundant location expansions.
4220 2018-06-05 Richard Biener <rguenther@suse.de>
4222 PR tree-optimization/86046
4223 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4224 if required after clearing TREE_ADDRESSABLE.
4226 2018-06-05 Richard Biener <rguenther@suse.de>
4228 PR tree-optimization/86047
4229 * tree-ssa-loop.c (for_each_index): Glob handling of all
4230 decls and constants and really handle all of them.
4232 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4235 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4236 qualifier_void_pointer and qualifier_const_void_pointer.
4237 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4238 (arm_init_builtins): Handle the above.
4239 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4240 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4241 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4244 2018-06-05 Martin Liska <mliska@suse.cz>
4246 * auto-profile.c (read_autofdo_file): Do not use
4247 gcov_ctr_summary struct.
4248 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4249 * coverage.c (struct counts_entry): Likewise.
4250 (read_counts_file): Read just single summary entry.
4251 (get_coverage_counts): Use gcov_summary struct.
4252 * coverage.h (get_coverage_counts): Likewise.
4253 * gcov-dump.c (dump_working_sets): Likewise.
4254 (tag_summary): Dump just single summary.
4255 * gcov-io.c (gcov_write_summary): Write just histogram
4257 (gcov_read_summary): Read just single summary.
4258 (compute_working_sets): Use gcov_summary struct.
4259 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4260 of GCOV_COUNTERS_SUMMABLE.
4261 (GCOV_COUNTERS_SUMMABLE): Remove.
4262 (GCOV_FIRST_VALUE_COUNTER): Replace with
4263 GCOV_COUNTER_V_INTERVAL.
4264 (struct gcov_ctr_summary): Remove.
4265 (struct gcov_summary): Directly use fields of former
4267 (compute_working_sets): Use gcov_summary struct.
4268 * gcov.c (read_count_file): Do not use ctrs fields.
4269 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4271 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4273 * profile.c: Likewise.
4274 * profile.h: Likewise.
4276 2018-06-05 Martin Liska <mliska@suse.cz>
4278 PR gcov-profile/84846
4279 * gcov.c (output_lines): Print working directory only
4280 in intermediate format.
4282 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4284 * config/s390/s390-builtin-types.def: Add void function type.
4285 * config/s390/s390-builtins.def: Use the function type for the
4288 2018-06-04 Jim Wilson <jimw@sifive.com>
4290 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4292 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4293 and maybe_eh_return. Change regno to unsigned int. Use new args to
4294 handle EH_RETURN_DATA_REGNO registers properly.
4295 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4296 (riscv_expand_epilogue): Update comment. Change argument name and
4297 type. Update code to use new name and type. Pass new args to
4298 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4300 * config/riscv/riscv.md (NORMAL_RETURN): New.
4301 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4302 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4303 (eh_return): Call gen_eh_return_internal and emit barrier.
4304 (eh_return_internal): Call riscv_expand_epilogue.
4306 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4308 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4309 bit_insertion field and declare can_be_merged_into method.
4310 (merged_store_group::can_be_merged_into): New method.
4311 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4312 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4313 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4315 2018-06-04 Richard Biener <rguenther@suse.de>
4317 PR tree-optimization/85955
4318 * builtins.c (fold_builtin_sincos): Convert pointers to
4319 destination to appropriate type before dereferencing.
4321 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4323 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4325 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4327 * expr.c (expand_expr_real_1): Force the operand into memory if
4328 its TYPE_MODE is BLKmode and if there is no integer mode for
4329 the number of bits being extracted.
4331 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4335 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4336 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4338 2018-06-04 Richard Biener <rguenther@suse.de>
4340 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4341 (cleanup_tree_cfg_noloop): ... single caller. Do
4342 start_recording_case_labels later.
4344 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4346 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4347 to _IMMINTRIN_H_INCLUDED.
4348 * config/i386/pconfigintrin.h: Ditto.
4349 * config/i386/waitpkgintrin.h: Ditto.
4350 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4351 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4352 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4353 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4354 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4355 waitpkgintrin.h and cldemoteintrin.h.
4357 2018-06-04 Richard Biener <rguenther@suse.de>
4359 PR tree-optimization/86038
4360 * tracer.c (find_best_successor): Check probability for
4361 being initialized, bail out if not.
4363 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4366 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4367 of bits to ignore when comparing architectures.
4369 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4371 PR tree-optimization/69615
4372 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4373 maximum or minimum of the type, try to merge it also as if
4374 range1 is + [-, x - 1] or + [x + 1, -].
4377 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4379 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4381 PR tree-optimization/86034
4382 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4383 the unsigned bitfield type in a bit insertion sequence if it does not
4384 have a larger precision than the bitfield size.
4385 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4387 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4389 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4391 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4393 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4394 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4395 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4396 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4398 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4400 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4401 Disable -fdelete-null-pointer-checks for ELF toolchain.
4403 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4404 Kito Cheng <kito.cheng@gmail.com>
4406 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4407 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4408 (nds32*-*-*): Add glibc and uclibc conditions.
4409 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4410 (TARGET_EXCEPT_UNWIND_INFO): Define.
4411 * config/nds32/elf.h: New file.
4412 * config/nds32/linux.h: New file.
4413 * config/nds32/nds32-elf.opt: New file.
4414 * config/nds32/nds32-linux.opt: New file.
4415 * config/nds32/nds32-fp-as-gp.c
4416 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4417 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4419 (nds32_asm_file_end): Ditto.
4420 (nds32_print_operand): Ditto.
4421 (nds32_insert_attributes): Ditto.
4422 (nds32_init_libfuncs): New function.
4423 (TARGET_HAVE_TLS): Define.
4424 (TARGET_INIT_LIBFUNCS): Define.
4425 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4427 (TARGET_ELF): Apply different mcmodel setting.
4428 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4429 been migrated into elf.h and linux.h files.
4430 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4431 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4432 (mcmodel): The content has been migrated into nds32-elf.opt and
4433 nds32-linux.opt files.
4434 * config/nds32/t-elf: New file.
4435 * config/nds32/t-linux: New file.
4437 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4438 Shiva Chen <shiva0217@gmail.com>
4440 * config/nds32/constants.md (unspec_volatile_element): Add
4441 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4442 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4444 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4445 (make_pass_nds32_fp_as_gp): Declare.
4446 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4448 (nds32_asm_function_end_prologue): Remove unused asm output.
4449 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4450 (nds32_asm_file_start): Output necessary fp_as_gp information.
4451 (nds32_option_override): Adjust register usage.
4452 (nds32_expand_prologue): Consider fp_as_gp situation.
4453 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4454 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4457 (simple_return): Ditto.
4458 (omit_fp_begin): Output special directive for fp_as_gp.
4459 (omit_fp_end): Output special directive for fp_as_gp.
4460 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4461 mforbid-fp-as-gp): New options.
4463 2018-06-01 Mark Wielaard <mark@klomp.org>
4465 * dwarf2out.c (dwarf2out_finish): Remove generation of
4466 DW_AT_loclists_base.
4468 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4470 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4471 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4472 (struct merged_store_group): Add bit_insertion field.
4473 (dump_char_array): Use standard hexadecimal format.
4474 (merged_store_group::merged_store_group): Set bit_insertion to false.
4475 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4476 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4477 also print the mask in the dump file.
4478 (pass_store_merging::gate): Minor tweak.
4479 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4480 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4481 stores with INTEGER_CST stores.
4482 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4483 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4484 and use it throughout. Generate bit insertion sequences if need be.
4485 (pass_store_merging::process_store): Remove redundant condition.
4486 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4488 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4490 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4491 the 128-bit floating point types. Fix function comment.
4493 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4495 * config/aarch64/aarch64-simd.md
4496 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4498 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4501 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4503 PR tree-optimization/85989
4504 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4506 (backprop::intersect_uses): Check it when deciding whether this
4507 is a backedge reference.
4508 (backprop::process_block): Add each phi to m_visited_phis
4509 after visiting it, then clear it at the end.
4511 2018-06-01 Richard Biener <rguenther@suse.de>
4513 * tree-vectorizer.h (vect_dr_stmt): New function.
4514 (vect_get_load_cost): Adjust.
4515 (vect_get_store_cost): Likewise.
4516 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4517 Use vect_dr_stmt instead of DR_SMTT.
4518 (vect_record_base_alignments): Likewise.
4519 (vect_calculate_target_alignment): Likewise.
4520 (vect_compute_data_ref_alignment): Likewise and make static.
4521 (vect_update_misalignment_for_peel): Likewise.
4522 (vect_verify_datarefs_alignment): Likewise.
4523 (vector_alignment_reachable_p): Likewise.
4524 (vect_get_data_access_cost): Likewise. Pass down
4525 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4526 (vect_get_peeling_costs_all_drs): Likewise.
4527 (vect_peeling_hash_get_lowest_cost): Likewise.
4528 (vect_enhance_data_refs_alignment): Likewise.
4529 (vect_find_same_alignment_drs): Likewise.
4530 (vect_analyze_data_refs_alignment): Likewise.
4531 (vect_analyze_group_access_1): Likewise.
4532 (vect_analyze_group_access): Likewise.
4533 (vect_analyze_data_ref_access): Likewise.
4534 (vect_analyze_data_ref_accesses): Likewise.
4535 (vect_vfa_segment_size): Likewise.
4536 (vect_small_gap_p): Likewise.
4537 (vectorizable_with_step_bound_p): Likewise.
4538 (vect_prune_runtime_alias_test_list): Likewise.
4539 (vect_analyze_data_refs): Likewise.
4540 (vect_supportable_dr_alignment): Likewise.
4541 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4542 (vect_gen_prolog_loop_niters): Likewise.
4543 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4544 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4546 (vect_recog_mask_conversion_pattern): Likewise.
4547 (vect_try_gather_scatter_pattern): Likewise.
4548 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4549 to vect_get_store_cost.
4550 (vect_get_store_cost): Get stmt_info instead of DR.
4551 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4552 (vect_get_load_cost): Get stmt_info instead of DR.
4554 2018-06-01 Richard Biener <rguenther@suse.de>
4557 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4558 vars wrapped in MEM_REFs.
4560 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4562 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4563 Fix subreg tests so that we only return a choice between
4564 GENERAL_REGS and FP_REGS if the original classes included both.
4566 2018-06-01 Richard Biener <rguenther@suse.de>
4569 * tree-ssa-structalias.c (get_function_part_constraint):
4570 Handle NULL fi->decl.
4571 (find_func_aliases_for_call): Properly handle indirect
4572 fi from direct call.
4573 (find_func_clobbers): Likewise.
4574 (ipa_pta_execute): Likewise.
4575 (create_variable_info_for): For functions that are ifunc_resolver
4576 resolve to a varinfo that contains the result of the resolver call.
4577 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4580 2018-05-31 Michael Collison <michael.collison@arm.com>
4582 * config/aarch64/aarch64.md:
4583 (*fix_to_zero_extenddfdi2): New pattern.
4584 * gcc.target/aarch64/fix_extend1.c: New testcase.
4586 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4590 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4591 and BUILT_IN_STRNCMP_EQ.
4592 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4593 BUILT_IN_STRNCMP_EQ.
4594 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4595 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4596 (gimple_fold_builtin): Likewise.
4597 * tree-ssa-strlen.c (compute_string_length): New function.
4598 (determine_min_obsize): New function.
4599 (handle_builtin_string_cmp): New function to handle calls to
4600 string compare functions.
4601 (strlen_optimize_stmt): Add handling to builtin string compare
4603 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4604 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4605 * tree.c (build_common_builtin_nodes): Add new defines of
4606 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4608 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4611 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4612 functions with naked attribute.
4614 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4616 * config/i386/sse.md (avx_vec_concat<mode>):
4617 Substitute concat_tg_mode mode attribute with xtg_mode.
4618 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4619 (concat_tg_mode): Remove mode attribute.
4621 2018-05-31 Martin Sebor <msebor@redhat.com>
4624 * calls.c (alloc_max_size): Correct a logic error/typo.
4625 Treat excessive arguments as infinite. Warn for invalid arguments.
4626 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4628 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4631 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4632 and movx for Haswell.
4634 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4635 Cesar Philippidis <cesar@codesourcery.com>
4638 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4639 when emitting error on private/firstprivate reductions.
4640 * omp-low.c (lower_omp_target): Avoid reference-type processing
4641 on pointers for firstprivate clause.
4643 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4645 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4648 * config/aarch64/aarch64-simd.md
4649 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4650 (aarch64_ld1_x3_<mode>): Likewise
4651 (aarch64_st1x2<VALLDIF:mode>): Likewise
4652 (aarch64_st1_x2_<mode>): Likewise
4653 (aarch64_st1x3<VALLDIF:mode>): Likewise
4654 (aarch64_st1_x3_<mode>): Likewise
4655 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4656 (vld1_s8_x3): Likewise.
4657 (vld1_u16_x3): Likewise.
4658 (vld1_s16_x3): Likewise.
4659 (vld1_u32_x3): Likewise.
4660 (vld1_s32_x3): Likewise.
4661 (vld1_u64_x3): Likewise.
4662 (vld1_s64_x3): Likewise.
4663 (vld1_f16_x3): Likewise.
4664 (vld1_f32_x3): Likewise.
4665 (vld1_f64_x3): Likewise.
4666 (vld1_p8_x3): Likewise.
4667 (vld1_p16_x3): Likewise.
4668 (vld1_p64_x3): Likewise.
4669 (vld1q_u8_x3): Likewise.
4670 (vld1q_s8_x3): Likewise.
4671 (vld1q_u16_x3): Likewise.
4672 (vld1q_s16_x3): Likewise.
4673 (vld1q_u32_x3): Likewise.
4674 (vld1q_s32_x3): Likewise.
4675 (vld1q_u64_x3): Likewise.
4676 (vld1q_s64_x3): Likewise.
4677 (vld1q_f16_x3): Likewise.
4678 (vld1q_f32_x3): Likewise.
4679 (vld1q_f64_x3): Likewise.
4680 (vld1q_p8_x3): Likewise.
4681 (vld1q_p16_x3): Likewise.
4682 (vld1q_p64_x3): Likewise.
4683 (vst1_s64_x2): Likewise.
4684 (vst1_u64_x2): Likewise.
4685 (vst1_f64_x2): Likewise.
4686 (vst1_s8_x2): Likewise.
4687 (vst1_p8_x2): Likewise.
4688 (vst1_s16_x2): Likewise.
4689 (vst1_p16_x2): Likewise.
4690 (vst1_s32_x2): Likewise.
4691 (vst1_u8_x2): Likewise.
4692 (vst1_u16_x2): Likewise.
4693 (vst1_u32_x2): Likewise.
4694 (vst1_f16_x2): Likewise.
4695 (vst1_f32_x2): Likewise.
4696 (vst1_p64_x2): Likewise.
4697 (vst1q_s8_x2): Likewise.
4698 (vst1q_p8_x2): Likewise.
4699 (vst1q_s16_x2): Likewise.
4700 (vst1q_p16_x2): Likewise.
4701 (vst1q_s32_x2): Likewise.
4702 (vst1q_s64_x2): Likewise.
4703 (vst1q_u8_x2): Likewise.
4704 (vst1q_u16_x2): Likewise.
4705 (vst1q_u32_x2): Likewise.
4706 (vst1q_u64_x2): Likewise.
4707 (vst1q_f16_x2): Likewise.
4708 (vst1q_f32_x2): Likewise.
4709 (vst1q_f64_x2): Likewise.
4710 (vst1q_p64_x2): Likewise.
4711 (vst1_s64_x3): Likewise.
4712 (vst1_u64_x3): Likewise.
4713 (vst1_f64_x3): Likewise.
4714 (vst1_s8_x3): Likewise.
4715 (vst1_p8_x3): Likewise.
4716 (vst1_s16_x3): Likewise.
4717 (vst1_p16_x3): Likewise.
4718 (vst1_s32_x3): Likewise.
4719 (vst1_u8_x3): Likewise.
4720 (vst1_u16_x3): Likewise.
4721 (vst1_u32_x3): Likewise.
4722 (vst1_f16_x3): Likewise.
4723 (vst1_f32_x3): Likewise.
4724 (vst1_p64_x3): Likewise.
4725 (vst1q_s8_x3): Likewise.
4726 (vst1q_p8_x3): Likewise.
4727 (vst1q_s16_x3): Likewise.
4728 (vst1q_p16_x3): Likewise.
4729 (vst1q_s32_x3): Likewise.
4730 (vst1q_s64_x3): Likewise.
4731 (vst1q_u8_x3): Likewise.
4732 (vst1q_u16_x3): Likewise.
4733 (vst1q_u32_x3): Likewise.
4734 (vst1q_u64_x3): Likewise.
4735 (vst1q_f16_x3): Likewise.
4736 (vst1q_f32_x3): Likewise.
4737 (vst1q_f64_x3): Likewise.
4738 (vst1q_p64_x3): Likewise.
4740 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4742 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4743 user_label_prefix to name.
4745 * tree-core.h: Update comment about the format of NAME string
4746 passed to handler in attribute_spec.
4748 * config/msp430/msp430.md: Remove erroneous subreg expression from
4749 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4750 zero_extend{q,h}isi2.
4752 2018-05-30 Borislav Petkov <bp@suse.de>
4754 * doc/extend.texi: Document some architecture specific
4755 constraints and sort entries.
4757 2018-05-30 Martin Sebor <msebor@redhat.com>
4760 * builtins.c (expand_builtin_stpcpy_1): New function.
4761 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4762 only if the former succeeds.
4764 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4766 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4769 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4771 * doc/invoke.texi (-flinker-output): Document
4773 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4775 * passes.c (ipa_write_summaries): Only modify statements if body
4777 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4778 incrementally linking.
4779 (ipa_passes): Likewise.
4780 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4781 pass down resolution info.
4782 * common.opt (flag_incremental_link): Update info.
4783 * gcc.c (plugin specs): Turn flinker-output=* to
4784 -plugin-opt=-linker-output-known
4785 * toplev.c (compile_file): Also cut compilation when doing incremental
4787 * flag-types. (enum lto_partition_model): Add
4788 LTO_LINKER_OUTPUT_NOLTOREL.
4789 (invoke.texi): Add -flinker-output docs.
4790 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4791 link same way as WPA; do not stream in dead initializers.
4793 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4794 darf2out_register_external_decl): Support incremental link.
4796 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4798 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4800 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4802 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4803 it down to simple_object_copy_lto_debug_sections.
4804 (run_gcc): Determine incremental LTO link time and configure
4805 lto1 into non-wpa mode, disable renaming of debug sections.
4807 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4809 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4810 descriptions of various incorrectly documented functions.
4812 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4815 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4816 address check not strict.
4818 2018-05-30 Richard Biener <rguenther@suse.de>
4820 PR tree-optimization/85964
4821 * tracer.c (better_p): Drop initialized count check, we only
4822 call the function with initialized counts now.
4823 (find_best_successor): Do find a best edge if one
4824 has uninitialized count.
4825 (find_best_predecessor): Likewise. Do BB frequency check only
4826 if count is initialized.
4828 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4830 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4831 (aarch64_ldrstr_offset_compare): New.
4832 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4833 load/store orderings.
4834 (aarch64_gen_adjusted_ldpstp): Likewise.
4836 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4838 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4839 Check for subset of GENERAL_REGS and FP_REGS.
4840 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4843 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4845 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4846 and wi::to_poly_offset. Add the current offset and then check
4847 whether the sum fits, rather than using an unchecked addition of
4848 a checked term. Check for a shwi rather than a uhwi.
4849 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4850 (store_constructor): Use poly_int_tree_p.
4851 (expand_expr_real_1): Likewise.
4852 * function.c (assign_temp): Likewise.
4853 * fold-const.c (const_binop): Use poly_int_tree_p and
4855 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4857 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4858 to_poly_offset for MEM offsets.
4859 * ipa-icf.c (sem_variable::equals): Likewise.
4860 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4861 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4862 wi::to_poly_offset for BIT_FIELD_REF offsets.
4863 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4865 * var-tracking.c (emit_note_insn_var_location): Use
4866 tree_to_poly_uint64.
4868 2018-05-29 Jim Wilson <jimw@sifive.com>
4870 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4872 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4875 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4876 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4878 (sse4_1_round<mode>2): Use nonimmediate_operand
4879 for operand 1 predicate.
4881 2018-05-29 Martin Sebor <msebor@redhat.com>
4882 Richard Biener <rguenther@suse.de>
4885 * calls.c (get_size_range): Call determine_value_range instead
4886 of get_value_range..
4887 * tree-vrp.h (determine_value_range): Declared new function.
4888 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4890 2018-05-29 Richard Biener <rguenther@suse.de>
4892 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4893 sure to use non-pattern stmts for get_earlier_stmt arguments.
4894 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4895 called on pattern stmts.
4896 (get_later_stmt): Likewise.
4898 2018-05-29 Martin Liska <mliska@suse.cz>
4900 PR gcov-profile/85759
4901 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4904 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4906 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4908 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4912 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4913 VEC_PACK_FLOAT_EXPR): New tree codes.
4914 * tree-pretty-print.c (op_code_prio): Handle
4915 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4916 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4917 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4918 * tree-inline.c (estimate_operator_cost): Likewise.
4919 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4920 * fold-const.c (const_binop): Likewise.
4921 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4922 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4923 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4924 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4925 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4926 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4927 * expr.c (expand_expr_real_2): Likewise.
4928 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4929 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4930 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4932 * optabs.c (expand_widen_pattern_expr): For
4933 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4934 sign from result type rather than operand's type.
4935 (expand_binop_directly): For vec_packu_float_optab and
4936 vec_packs_float_optab allow result type to be different from operand's
4938 * optabs-tree.c (optab_for_tree_code): Handle
4939 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4940 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4941 * tree-vect-generic.c (expand_vector_operations_1): Handle
4942 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4943 VEC_PACK_FLOAT_EXPR.
4944 * tree-vect-stmts.c (supportable_widening_operation): Handle
4946 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4947 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4948 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4949 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4950 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4952 (vec_pack<floatprefix>_float_<mode>): New expander.
4953 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4955 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4956 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4957 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4958 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4959 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4961 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4962 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4963 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4964 VEC_PACK_FLOAT_EXPR): Document.
4966 2018-05-29 Richard Biener <rguenther@suse.de>
4968 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4970 (stmt_vec_info_vec): Make pointer.
4971 (init_stmt_vec_info_vec): Remove.
4972 (free_stmt_vec_info_vec): Likewise.
4973 (set_stmt_vec_info_vec): New function.
4974 (free_stmt_vec_infos): Likewise.
4975 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4976 (set_vinfo_for_stmt): Likewise.
4977 (get_earlier_stmt): Likewise.
4978 (get_later_stmt): Likewise.
4979 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4980 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4981 (vec_info::~vec_info): Free stmt_vec_infos.
4982 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4983 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4984 (pass_slp_vectorize::execute): Likewise.
4985 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4986 (free_stmt_vec_info_vec): Likewise.
4987 (set_stmt_vec_info_vec): New function.
4988 (free_stmt_vec_infos): Likewise.
4989 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4990 the global stmt_vec_info_vec.
4991 * tree-parloops.c (gather_scalar_reductions): Use
4992 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4995 2018-05-29 Richard Biener <rguenther@suse.de>
4997 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4999 2018-05-29 Martin Liska <mliska@suse.cz>
5000 David Malcolm <dmalcolm@redhat.com>
5002 * vec.c (test_reverse): New.
5003 (vec_c_tests): Add new test.
5004 * vec.h (vl_ptr>::reverse): New function.
5006 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5008 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5010 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5013 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5015 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5017 2018-05-28 Richard Biener <rguenther@suse.de>
5019 PR tree-optimization/85933
5020 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5021 look at stmts marked as vectorizable.
5023 2018-05-28 Richard Biener <rguenther@suse.de>
5025 PR tree-optimization/85934
5026 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5027 vector boolean check before scalar optimization.
5029 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5031 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5034 2018-05-28 Mark Wielaard <mark@klomp.org>
5036 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5037 if it is an expression containing a minus sign.
5039 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5041 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5043 2018-05-27 Paul Koning <ni1d@arrl.net>
5045 * config/pdp11/pdp11.md (truncsihi2): Remove.
5047 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5048 Chung-Ju Wu <jasonwucj@gmail.com>
5050 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5052 (unaligned_store_dw): Ditto.
5053 * config/nds32/nds32-memory-manipulation.c
5054 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5055 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5056 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5057 (emit_setmem_word_loop): Rename to ...
5058 (emit_setmem_doubleword_loop): ... this.
5059 (nds32_gen_dup_4_byte_to_word_value): New function.
5060 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5061 (nds32_expand_setmem_loop): Refine implementation.
5062 (nds32_expand_setmem_loop_v3m): Ditto.
5063 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5066 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5068 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5070 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5072 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5073 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5074 machine->attr_no_prologue_p.
5075 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5076 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5077 (nds32_expand_epilogue): Consider attr_naked_p.
5078 (nds32_expand_epilogue_v3pop): Likewise.
5079 (nds32_can_use_return_insn): Likewise.
5080 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5081 attr_no_prologue_p fields.
5082 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5084 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5087 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5089 * config/i386/sse.md
5090 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5092 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5094 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5096 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5098 (*<floatsuffix>floatv2div2sf2): Rename to ...
5099 (*float<floatunssuffix>v2div2sf2): ... this.
5100 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5101 (float<floatunssuffix>v2div2sf2_mask): ... this.
5102 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5103 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5104 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5106 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5108 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5110 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5112 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5114 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5116 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5117 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5118 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5119 gen_ufix_truncv8dfv8si2.
5120 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5121 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5122 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5123 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5124 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5125 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5126 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5127 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5129 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5133 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5135 2018-05-25 Jim Wilson <jimw@sifive.com>
5137 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5138 * config/riscv/riscv.c (struct machine_function): Add
5139 interrupt_handler_p and attribute_checked_p fields.
5140 (riscv_attribute_table): Add interrupt.
5141 (riscv_interrupt_type_p): New.
5142 (riscv_save_reg_p): Save extra regs for interrupt handler.
5143 (riscv_use_save_libcall): Return false for interrupt handler.
5144 (riscv_first_stack_step): Add forward declaration.
5145 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5146 for interrupt handler with large frame. Use it for saved reg list.
5147 (riscv_expand_prologue): Move flag_stack_usage_info support to
5148 eliminate duplication.
5149 (riscv_expand_epilogue): Generate mret for interrupt handler.
5150 (riscv_epilogue_uses): New.
5151 (riscv_can_use_return_insn): Return false for interrupt handler.
5152 (riscv_function_ok_for_sibcall): Likewise.
5153 (riscv_set_current_function): Add interrupt handler support.
5154 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5155 * config/riscv/riscv.md (UNSPECV_MRET): New.
5157 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5159 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5161 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5163 PR tree-optimization/85712
5164 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5165 this candidate has already been replaced in-situ by a copy.
5167 2018-05-25 Jason Merrill <jason@redhat.com>
5169 PR c++/80485 - inline function non-zero address.
5170 * symtab.c (nonzero_address): Check DECL_COMDAT.
5172 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5175 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5176 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5177 extension of SImode operation. Use const123_operand predicate.
5178 (*saddsi_1): Remove.
5179 (*saddl_se_1): Ditto.
5181 (*ssubl_se_1): Ditto.
5182 * config/alpha/predicates.md (const123_operand): New predicate.
5183 * config/alpha/constraints.md (P): Use IN_RANGE.
5185 2018-05-25 Richard Biener <rguenther@suse.de>
5187 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5189 (ref_maybe_used_by_stmt_p): Likewise.
5190 (stmt_may_clobber_ref_p): Likewise.
5191 (stmt_may_clobber_ref_p_1): Likewise.
5192 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5194 (ref_maybe_used_by_stmt_p): Likewise.
5195 (stmt_may_clobber_ref_p): Likewise.
5196 (stmt_may_clobber_ref_p_1): Likewise.
5197 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5198 the alias oracle to disambiguate DRs with stmts DR analysis
5200 (vect_analyze_data_refs): Do not give up on not analyzable
5201 DRs for BB vectorization. Remove code truncating the dataref
5204 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5207 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5208 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5209 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5211 2018-05-25 Richard Biener <rguenther@suse.de>
5213 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5214 function, combining stmt data ref gathering and fatal analysis
5216 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5217 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5218 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5219 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5220 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5222 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5224 PR tree-optimization/85720
5225 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5226 SCC if all partitions are builtins.
5227 (version_loop_by_alias_check): New parameter. Generate cancelable
5228 runtime alias check if all partitions are builtins.
5229 (distribute_loop): Update call to above function.
5231 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5233 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5234 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5235 (parm_default_def_partition_arg): Ditto.
5236 (set_parm_default_def_partition): Ditto.
5237 (get_parm_default_def_partitions): Ditto and make it static.
5238 (get_undefined_value_partitions): Ditto and make it static.
5239 (remove_ssa_form): Refactor call to init_var_map here.
5240 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5241 computation for loop region.
5242 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5243 (register_default_def): Delete.
5244 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5245 (parm_default_def_partition_arg): Ditto.
5246 (set_parm_default_def_partition): Ditto.
5247 (get_parm_default_def_partitions): Ditto and make it static.
5248 (get_undefined_value_partitions): Ditto and make it static.
5249 (coalesce_with_default, coalesce_with_default): Update comment.
5250 (create_coalesce_list_for_region): New func factored out from
5251 create_outofssa_var_map.
5252 (populate_coalesce_list_for_outofssa): New func factored out from
5253 create_outofssa_var_map and coalesce_ssa_name.
5254 (create_outofssa_var_map): Delete.
5255 (coalesce_ssa_name): Refactor to support live range computation.
5256 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5257 (get_parm_default_def_partitions): Delete.
5258 (get_undefined_value_partitions): Ditto.
5259 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5260 computation for loop region.
5261 (new_tree_live_info, loe_visit_block): Ditto.
5262 (live_worklist, set_var_live_on_entry): Ditto.
5263 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5264 * tree-ssa-live.h (struct _var_map): New fields.
5265 (init_var_map): Change decl.
5266 (region_contains_p): New.
5268 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5270 * tree-ssa-live.h (live_merge_and_clear): Delete.
5272 2018-05-25 Richard Biener <rguenther@suse.de>
5275 * tree-dump.c (dequeue_and_dump): Remove access to removed
5276 operand 2 of a SWITCH_EXPR.
5278 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5280 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5281 multiplication and division.
5282 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5283 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5284 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5285 (cond_udiv_optab, cond_umod_optab): New optabs.
5286 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5287 (IFN_COND_RDIV): New internal functions.
5288 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5289 TRUNC_MOD_EXPR and RDIV_EXPR.
5290 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5291 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5293 (SVE_INT_BINARY): Include mult.
5294 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5295 (optab, sve_int_op): Handle mult.
5296 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5298 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5299 for SVE_INT_BINARY_SD.
5301 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5303 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5304 (optab, sve_int_op): Handle div and udiv.
5305 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5306 for SVE_INT_BINARY_SD.
5307 (*<optab><mode>3): New insn for the same.
5309 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5311 * tree-vect-patterns.c: Include predict.h.
5312 (vect_recog_divmod_pattern): Restrict check for division support
5313 to when optimizing for size.
5315 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5317 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5318 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5319 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5320 (gimple_match_op::set_op): Likewise.
5321 (gimple_resimplify4): Declare.
5322 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5323 (expr::gen_transform): Likewise.
5324 (decision_tree::gen): Generate a simplification routine for 4 operands.
5325 * gimple-match-head.c (gimple_simplify): Add an overload for
5326 4 operands. In the top-level function, handle up to 4 call
5327 arguments and call gimple_resimplify4.
5328 (gimple_resimplify4): New function.
5329 (build_call_internal): Pass a fourth operand.
5330 (maybe_push_to_seq): Likewise.
5331 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5332 Fold VEC_COND_EXPRs of an operation and a default value into
5333 an IFN_COND_* function if possible.
5334 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5336 (SVE_COND_FP_BINARY): Include them.
5337 (optab, sve_fp_op): Handle them.
5338 (SVE_INT_BINARY_REV): New code iterator.
5339 (SVE_COND_FP_BINARY_REV): New int iterator.
5340 (commutative): New int attribute.
5341 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5343 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5345 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5346 (*cond_<optab><mode>): New patterns for reversed operands.
5348 2018-05-25 Richard Biener <rguenther@suse.de>
5350 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5351 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5352 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5353 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5354 * tree-vect-data-refs.c (everywhere): Adjust users.
5355 * tree-vect-loop.c (everywhere): Likewise.
5356 * tree-vect-slp.c (everywhere): Likewise.
5357 * tree-vect-stmts.c (everywhere): Likewise.
5358 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5360 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5362 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5364 (gcc_cv_as_section_exclude): ... this.
5365 Try Solaris as #exclude syntax.
5366 * configure: Regenerate.
5367 * config.in: Regenerate.
5368 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5370 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5371 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5373 * varasm.c (default_elf_asm_named_section): Don't check if
5374 HAVE_GAS_SECTION_EXCLUDE is defined.
5376 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5378 * doc/md.texi: Update the documentation of the cond_* optabs
5379 to mention the new final operand. Fix GET_MODE_NUNITS call.
5380 Describe the scalar case too.
5381 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5382 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5384 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5385 (get_conditional_internal_fn): Update comment.
5386 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5387 accumulator value as a final argument to conditional functions.
5388 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5389 a define_expand and add an "else" operand. Assert for now that
5390 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5391 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5392 (*cond_<optab><mode>): New patterns.
5393 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5394 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5395 (UNSPEC_COND_EOR): Delete.
5396 (optab): Remove associated mappings.
5397 (SVE_INT_BINARY): New code iterator.
5398 (sve_int_op): Remove int attribute and add "minus" to the code
5400 (SVE_COND_INT_OP): Delete.
5401 (SVE_COND_FP_OP): Rename to...
5402 (SVE_COND_FP_BINARY): ...this.
5404 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5406 * optabs.c (can_reuse_operands_p): New function.
5407 (maybe_legitimize_operands): Try to reuse the results for
5410 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5412 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5413 Add {q} suffix to insn mnemonic.
5415 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5417 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5418 (msp430_warn_func_return): New.
5420 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5422 * fold-const.c (tree_nonzero_bits): New function.
5423 * fold-const.h (tree_nonzero_bits): Likewise.
5424 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5425 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5427 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5431 * varasm.c (assemble_alias): Check ifunc_resolver only on
5434 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5437 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5438 when memory input operand is handled.
5440 2018-05-24 Luis Machado <luis.machado@linaro.org>
5442 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5444 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5446 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5448 * match.pd: Delay FMA folds until after vectorization.
5450 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5453 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5454 address check not strict.
5456 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5458 * gimple-match.h (gimple_match_op): New class.
5459 (mprts_hook): Replace parameters with a gimple_match_op *.
5460 (maybe_build_generic_op): Likewise.
5461 (gimple_simplified_result_is_gimple_val): Replace parameters with
5462 a const gimple_match_op *.
5463 (gimple_simplify): Replace code_helper * and tree * parameters with
5464 a gimple_match_op * parameter.
5465 (gimple_resimplify1): Replace code_helper *, tree and tree *
5466 parameters with a gimple_match_op * parameter.
5467 (gimple_resimplify2): Likewise.
5468 (gimple_resimplify3): Likewise.
5469 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5470 parameters with a gimple_match_op * parameter.
5471 * gimple-match-head.c (gimple_simplify): Change prototypes of
5472 auto-generated functions to take a gimple_match_op * instead of
5473 separate code_helper * and tree * parameters. Make the same
5474 change in the top-level overload and update calls to the
5475 gimple_resimplify routines. Update calls to the auto-generated
5476 functions and to maybe_push_res_to_seq in the publicly-facing
5477 operation-specific gimple_simplify overloads.
5478 (gimple_match_op::MAX_NUM_OPS): Define.
5479 (gimple_resimplify1): Replace rcode and ops with a single res_op
5480 parameter. Update call to gimple_simplify.
5481 (gimple_resimplify2): Likewise.
5482 (gimple_resimplify3): Likewise.
5483 (mprts_hook): Replace parameters with a gimple_match_op *.
5484 (maybe_build_generic_op): Likewise.
5485 (build_call_internal): Replace type, nargs and ops with
5486 a gimple_match_op *.
5487 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5488 with a single gimple_match_op *. Update calls to mprts_hook,
5489 build_call_internal and gimple_simplified_result_is_gimple_val.
5490 Factor out code that is common to the tree_code and combined_fn cases.
5491 * genmatch.c (expr::gen_transform): Replace tem_code and
5492 tem_ops with a gimple_match_op called tem_op. Update calls
5493 to the gimple_resimplify functions and maybe_push_res_to_seq.
5494 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5495 res_ops. Update call to the gimple_resimplify functions.
5496 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5497 (decision_tree::gen): Make the functions take a gimple_match_op *
5498 called res_op instead of separate res_code and res_ops parameters.
5499 Update call accordingly.
5500 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5501 and ops with a single res_op parameter. Update calls to
5502 maybe_build_generic_op and maybe_push_res_to_seq.
5503 (fold_stmt_1): Update calls to gimple_simplify and
5504 replace_stmt_with_simplification.
5505 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5506 and gimple_simplified_result_is_gimple_val.
5507 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5509 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5510 with a gimple_match_op *.
5511 (vn_nary_build_or_lookup): Likewise. Update call to
5512 vn_nary_build_or_lookup_1.
5513 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5514 gimple_match_op *. Update calls to the gimple_resimplify routines
5515 and to gimple_simplified_result_is_gimple_val.
5516 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5517 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5518 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5519 (visit_nary_op): Likewise.
5520 (visit_reference_op_load): Likewise.
5522 2018-05-23 Luis Machado <luis.machado@linaro.org>
5524 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5525 modifier for printing the step amount.
5527 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5530 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5533 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5535 * doc/sourcebuild.texi (Endianness): New subsubsection.
5537 2018-05-23 Luis Machado <luis.machado@linaro.org>
5539 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5540 <prefetch_dynamic_strides>: New const bool field.
5541 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5542 prefetch_dynamic_strides.
5543 (exynosm1_prefetch_tune): Likewise.
5544 (thunderxt88_prefetch_tune): Likewise.
5545 (thunderx_prefetch_tune): Likewise.
5546 (thunderx2t99_prefetch_tune): Likewise.
5547 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5549 (aarch64_override_options_internal): Update to set
5550 PARAM_PREFETCH_DYNAMIC_STRIDES.
5551 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5552 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5553 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5554 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5555 prefetch-dynamic-strides setting.
5557 2018-05-23 Luis Machado <luis.machado@linaro.org>
5559 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5560 <minimum_stride>: New const int field.
5561 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5562 minimum_stride field defaulting to -1.
5563 (exynosm1_prefetch_tune): Likewise.
5564 (thunderxt88_prefetch_tune): Likewise.
5565 (thunderx_prefetch_tune): Likewise.
5566 (thunderx2t99_prefetch_tune): Likewise.
5567 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5568 <default_opt_level>: Set to 3.
5569 (aarch64_override_options_internal): Update to set
5570 PARAM_PREFETCH_MINIMUM_STRIDE.
5571 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5572 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5573 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5574 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5575 stride is constant and is below the minimum stride threshold.
5577 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5579 * config/arm/arm-cpus.in (mode26): Delete.
5580 (armv4): Delete mode26 reference.
5581 * config/arm/arm.c (arm_configure_build_target): Delete use of
5584 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5586 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5588 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5589 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5591 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5592 (floatunsdidf2): Ditto.
5594 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5596 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5597 (fixuns_trunc<mode>si2_avx512f): Ditto.
5598 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5599 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5600 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5602 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5604 PR rtl-optimization/79985
5605 * df-scan.c (df_insn_refs_collect): Remove special case for
5606 global registers and asm statements.
5608 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5610 * extend.texi (Global Register Variables): Rewrite the bullet list.
5611 Note that the register is available for allocation. Note that access
5612 via inline asm must use constraints. Add note about async-signal
5613 handlers. Remove paragraph about automagic register selection.
5615 2018-05-23 Richard Biener <rguenther@suse.de>
5617 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5618 of fixed offset from memset VN.
5620 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5622 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5624 (alloc_cand_and_find_basis): Initialize first_interp field.
5625 (slsr_process_mul): Modify first_interp field.
5626 (slsr_process_add): Likewise.
5627 (slsr_process_cast): Modify first_interp field for each new
5629 (slsr_process_copy): Likewise.
5630 (dump_candidate): Dump first_interp field.
5631 (replace_mult_candidate): Process all interpretations, not just
5633 (replace_rhs_if_not_dup): Likewise.
5634 (replace_one_candidate): Likewise.
5636 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5638 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5640 (aarch64_needs_frame_chain): New function.
5641 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5643 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5646 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5647 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5648 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5649 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5650 as true for strict-align.
5651 (aarch64_can_inline_p): Perform checks even when callee has no
5652 attributes to check for strict alignment.
5653 * doc/extend.texi (AArch64 Function Attributes): Document
5655 * doc/invoke.texi: (AArch64 Options): Likewise.
5657 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5659 PR tree-optimization/85853
5660 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5661 the handling of the root of the node to...
5662 (vect_slp_analyze_node_operations_1): ...this new function,
5663 and run the whole thing with the child nodes' def types
5664 set according to their SLP node's def type.
5666 2018-05-23 Richard Biener <rguenther@suse.de>
5669 * tree-data-ref.c (create_runtime_alias_checks): Defer
5670 and ignore overflow warnings.
5672 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5674 PR tree-optimization/85822
5675 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5678 2018-05-23 Richard Biener <rguenther@suse.de>
5680 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5681 memset constants via native_interpret_expr.
5683 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5686 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5688 (cgraph_node::create_alias): Likewise.
5689 (cgraph_node::get_availability): Check ifunc_resolver instead
5690 of looking up ifunc attribute.
5691 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5692 * varasm.c (do_assemble_alias): Likewise.
5693 (assemble_alias): Likewise.
5694 (default_binds_local_p_3): Likewise.
5695 * cgraph.h (cgraph_node): Add ifunc_resolver.
5696 (cgraph_node::only_called_directly_or_aliased_p): Return false
5698 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5700 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5701 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5702 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5703 instead of looking up ifunc attribute.
5705 2018-05-22 Luis Machado <luis.machado@linaro.org>
5707 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5709 2018-05-22 Martin Sebor <msebor@redhat.com>
5712 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5713 only when expasion succeeds.
5714 (expand_builtin_strcmp): Same.
5715 (expand_builtin_strncmp): Same.
5717 2018-05-22 Martin Sebor <msebor@redhat.com>
5719 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5721 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5722 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5724 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5725 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5727 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5728 Add check for legitimate_address.
5729 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5730 (aarch64_swap_ldrstr_operands): New.
5731 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5734 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5735 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5737 * config/aarch64/aarch64.md: New patterns to generate stp
5739 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5740 single words and double words.
5741 (load_pair_sw, load_pair_dw): Likewise.
5742 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5744 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5746 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5747 for different mode ldpstp and add peephole for merged zero stores.
5749 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5751 (aarch64_gen_store_pair): Rename calls to match new patterns.
5752 (aarch64_gen_load_pair): Rename calls to match new patterns.
5753 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5754 (load_pair<DREG:mode><DREG2:mode>): ... This.
5755 (store_pair<mode>): Rename to...
5756 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5757 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5759 (V_INT_EQUIV): Handle SImode.
5760 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5763 2018-05-22 Martin Sebor <msebor@redhat.com>
5766 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5767 or ajust the presumed bound on an operation to avoid unnecessary
5770 2018-05-22 Martin Sebor <msebor@redhat.com>
5772 PR tree-optimization/85826
5773 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5774 assuming that a DECL necesarily has a constant size.
5776 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5779 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5781 2018-05-22 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/85834
5784 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5785 non-constant and non-zero memset arguments.
5787 2018-05-22 Martin Liska <mliska@suse.cz>
5790 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5792 2018-05-22 Richard Biener <rguenther@suse.de>
5794 PR tree-optimization/85863
5795 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5796 comparisons when vectype is specified.
5797 (vectorizable_condition): Do not specify vectype for
5798 vect_is_simple_cond when SLP vectorizing.
5800 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5803 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5804 define __ibm128 as long double.
5805 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5806 as a distinct type when IEEE 128-bit support is enabled.
5807 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5808 128-bit types to use the correct functions.
5809 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5810 convert between 128-bit floating point types that have different
5811 modes but the same representation, instead of using gen_lowpart to
5813 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5815 (IFKF_reg): New attributes to give the register constraints for
5817 (extend<mode>tf2_internal): New insns to mark an explicit
5818 conversion between 128-bit floating point types that have a
5819 different mode but share the same representation.
5821 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5823 PR tree-optimization/85814
5824 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5825 a null return from get_strinfo when unsharing the next
5826 strinfo in the chain.
5828 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5831 * varasm.c (weak_finish): Clean up weak_decls.
5833 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5835 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5836 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5837 UNSPEC_UADALP values.
5838 * config/aarch64/iterators.md (ABAL): New int iterator.
5841 (sur): Add mappings for the above.
5842 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5844 (aarch64_<sur>abal<mode>_4): Likewise.
5845 (aarch64_<sur>adalp<mode>_3): Likewise.
5846 (<sur>sadv16qi): New define_expand.
5848 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5850 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5851 (*movdf_internal): Ditto.
5852 (*rcpsf2_sse): Ditto.
5853 (*rsqrtsf2_sse): Ditto.
5854 (*sqrt<mode>2_sse): Ditto.
5856 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5858 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5860 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5861 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5862 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5863 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5865 * config/aarch64/arm_neon.h: Likewise.
5866 * config/aarch64/iterators.md (VQ_I): New.
5868 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5870 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5872 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5875 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5877 * config/nds32/constraints.md (S): New constraint.
5878 * config/nds32/nds32.md (call_internal): Use constraint S.
5879 (call_value_internal): Likewise.
5880 (sibcall_internal): Likewise.
5881 (sibcall_value_internal): Likewise.
5883 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5884 Chung-Ju Wu <jasonwucj@gmail.com>
5886 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5889 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5890 Chung-Ju Wu <jasonwucj@gmail.com>
5892 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5893 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5894 (nds32_rtx_costs_impl): Simplify.
5895 (nds32_address_cost_impl): Simplify.
5896 (nds32_init_rtx_costs): New function.
5897 (nds32_rtx_costs_speed_prefer): Likewise.
5898 (nds32_rtx_costs_size_prefer): Likewise.
5899 (nds32_address_cost_speed_prefer): Likewise.
5900 (nds32_address_cost_speed_fwprop): Likewise.
5901 (nds32_address_cost_size_prefer): Likewise.
5902 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5903 * config/nds32/nds32.c (nds32_option_override): Use
5904 nds32_init_rtx_costs function.
5906 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5908 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5909 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5910 (TARGET_PIPELINE_N8): Likewise.
5911 (TARGET_PIPELINE_N10): Likewise.
5912 (TARGET_PIPELINE_N13): Likewise.
5913 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5915 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5917 * config/nds32/nds32-fpu.md: Update copyright year.
5919 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5921 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5923 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5925 * config/nds32/nds32.c
5926 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5927 * config/nds32/nds32.opt (minline-asm-r15): New option.
5929 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5931 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5933 * config/nds32/nds32.md (abssi2): New pattern.
5935 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5937 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5938 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5939 Merge insn pattern from sse_cvtsi2ss<round_name> and
5940 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5941 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5942 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5943 using SWI48 mode iterator.
5944 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5945 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5946 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5947 pattern from sse_cvttss2si<round_saeonly_name>
5948 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5949 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5950 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5951 using SWI48 mode iterator.
5952 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5953 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5954 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5955 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5956 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5957 using SWI48 mode iterator.
5958 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5959 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5960 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5961 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5962 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5963 SWI48 mode iterator.
5964 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5965 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5966 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5967 pattern from sse_cvttsd2si<round_saeonly_name>
5968 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5970 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5972 * config/nds32/nds32-md-auxiliary.c
5973 (nds32_valid_smw_lwm_base_p): Refine.
5974 (nds32_output_smw_single_word): Refine.
5975 (nds32_output_smw_double_word): New.
5976 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5978 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5980 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5981 (nds32_output_stack_pop): Refine.
5982 (nds32_expand_unaligned_load): Refine.
5983 (nds32_expand_unaligned_store): Refine.
5985 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5986 Chung-Ju Wu <jasonwucj@gmail.com>
5988 * config/nds32/constants.md: Add TP_REGNUM constant.
5989 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5990 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5992 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5993 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5994 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5995 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5996 and PIC code generation.
5997 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5999 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6001 * config/nds32/nds32.md: Support TLS and PIC.
6002 * config/nds32/nds32.c: Support TLS and PIC.
6003 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6004 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6007 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6009 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6010 mode with E_ prefix.
6012 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6013 Chung-Ju Wu <jasonwucj@gmail.com>
6015 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6016 * config/nds32/nds32-md-auxiliary.c
6017 (symbolic_reference_mentioned_p): New.
6018 (nds32_legitimize_ict_address): New.
6019 (nds32_expand_ict_move): New.
6020 (nds32_indirect_call_referenced_p): New.
6021 (nds32_symbol_binds_local_p): Delete.
6022 (nds32_long_call_p): Modify.
6023 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6024 * config/nds32/nds32-protos.h
6025 (symbolic_reference_mentioned_p): Declare.
6026 (nds32_legitimize_ict_address): Declare.
6027 (nds32_expand_ict_move): Declare.
6028 (nds32_indirect_call_referenced_p): Declare.
6029 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6030 (nds32_relax_group): Use nds32_ict_const_p as condition.
6031 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6032 (nds32_asm_file_start): Output ict_model directive in asm code.
6033 (nds32_legitimate_address_p): Consider indirect call.
6034 (nds32_print_operand): Consider indirect call.
6035 (nds32_print_operand_address): Consider indirect call.
6036 (nds32_insert_attributes): Handle "indirect_call" attribute.
6037 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6038 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6039 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6040 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6041 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6042 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6043 (TARGET_ICT_MODEL_SMALL): Define.
6044 (TARGET_ICT_MODEL_LARGE): Define.
6045 * config/nds32/nds32.md (movsi): Consider ict model.
6046 (call, call_value): Consider ict model.
6047 (sibcall, sibcall_value): Consider ict model.
6048 * config/nds32/nds32.opt (mict-model): New option.
6049 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6052 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6053 Monk Chiang <sh.chiang04@gmail.com>
6054 Jim Wilson <jimw@sifive.com>
6056 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6057 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6058 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6059 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6060 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6061 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6062 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6063 compute save_libcall_adjustment properly.
6064 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6065 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6066 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6067 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6068 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6069 (ABI_SPEC): Handle mabi=ilp32e.
6070 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6071 (RVE): Add RVE mask.
6072 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6073 <-march>: Add rv32e as an example.
6075 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6078 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6079 (intra_create_variable_infos): Handle C++ constructors.
6081 2018-05-18 Martin Liska <mliska@suse.cz>
6083 * passes.def: Remove a redundant pass.
6085 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6088 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6090 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6092 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6094 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6096 (armv2, armv2a, armv3, armv3m): Delete architectures.
6097 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6098 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6099 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6101 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6102 (*mulsidi3adddi): Likewise.
6103 (mulsidi3): Likewise.
6104 (*mulsidi3_nov6): Likewise.
6105 (umulsidi3): Likewise.
6106 (umulsidi3_nov6): Likewise.
6107 (umaddsidi4): Likewise.
6108 (*umulsidi3adddi): Likewise.
6109 (smulsi3_highpart): Likewise.
6110 (*smulsi3_highpart_nov6): Likewise.
6111 (umulsi3_highpart): Likewise.
6112 (*umulsi3_highpart_nov6): Likewise.
6113 * config/arm/arm.h (arm_arch3m): Delete.
6114 * config/arm/arm.c (arm_arch3m): Delete.
6115 (arm_option_override_internal): Update armv3-related comment.
6116 (arm_configure_build_target): Delete use of isa_bit_mode32.
6117 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6118 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6119 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6121 (mulusa3): Likewise.
6122 * config/arm/arm-protos.h (arm_arch3m): Delete.
6123 * config/arm/arm-tables.opt: Regenerate.
6124 * config/arm/arm-tune.md: Likewise.
6125 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6126 deleted architectures.
6128 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6130 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6131 (armv5t, armv5te): New features.
6132 (ARMv5, ARMv5e): Delete fgroups.
6133 (ARMv5t, ARMv5te): Adjust for above changes.
6135 (armv5, armv5e): Delete arches.
6136 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6138 (*call_reg_arm): Likewise.
6139 (*call_value_reg_armv5): Likewise.
6140 (*call_value_reg_arm): Likewise.
6141 (*call_symbol): Likewise.
6142 (*call_value_symbol): Likewise.
6143 (*sibcall_insn): Likewise.
6144 (*sibcall_value_insn): Likewise.
6146 (prefetch): Likewise.
6147 (define_split and define_peephole2 dependent on arm_arch5):
6149 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6151 (TARGET_ARM_QBIT): Likewise.
6152 (TARGET_DSP_MULTIPLY): Likewise.
6153 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6154 (arm_arch5, arm_arch5e): Delete.
6155 (arm_arch5t, arm_arch5te): Declare.
6156 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6157 (arm_arch5t): Declare.
6158 (arm_option_reconfigure_globals): Update for the above.
6159 (arm_options_perform_arch_sanity_checks): Update comment, replace
6160 use of arm_arch5 with arm_arch5t.
6161 (use_return_insn): Likewise.
6162 (arm_emit_call_insn): Likewise.
6163 (output_return_instruction): Likewise.
6164 (arm_final_prescan_insn): Likewise.
6165 (arm_coproc_builtin_available): Likewise.
6166 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6167 arm_arch5e with arm_arch5t and arm_arch5te.
6168 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6169 (arm_arch5t, arm_arch5te): Declare.
6170 * config/arm/arm-tables.opt: Regenerate.
6171 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6172 * config/arm/t-multilib: Likewise.
6173 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6174 instead of arm_arch5.
6175 (*call_reg_thumb1): Likewise.
6176 (*call_value_reg_thumb1_v5): Likewise.
6177 (*call_value_reg_thumb1): Likewise.
6178 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6180 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6182 2018-05-18 Martin Liska <mliska@suse.cz>
6184 PR gcov-profile/84846
6185 * doc/gcov.texi: Document -t option of gcov tool.
6187 2018-05-18 Martin Liska <mliska@suse.cz>
6189 PR gcov-profile/84846
6190 * gcov.c (print_usage): Add new -t option.
6191 (process_args): Handle the option.
6192 (generate_results): Use stdout as output when requested by
6195 2018-05-18 Martin Liska <mliska@suse.cz>
6197 PR gcov-profile/84846
6198 * coverage.c (coverage_init): Write PWD to .gcno file.
6199 * doc/gcov.texi: Document how working directory is printed.
6200 * gcov-dump.c (dump_gcov_file): Print PWD.
6201 * gcov.c (output_intermediate_file): Likewise.
6202 (read_graph_file): Read PWD string.
6203 (output_lines): Print PWD.
6205 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6208 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6209 for retval and return false if all args to phi are zero.
6211 2018-05-18 Richard Biener <rguenther@suse.de>
6213 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6215 (evrp_dom_walker::before_dom_children): Call it.
6217 2018-05-18 Richard Biener <rguenther@suse.de>
6219 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6220 results when processing array refs with variable index.
6222 2018-05-18 Toon Moene <toon@moene.org>
6224 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6225 directly after that of -floop-interchange. Indicate that both
6226 options are enabled by default when specifying -O3.
6228 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6230 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6231 iterator. Delete separate integer-mode vec_set<mode> expander.
6232 (aarch64_simd_vec_setv2di): Delete.
6233 (vec_setv2di): Delete.
6234 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6235 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6236 the "w, r" alternative.
6238 2018-05-18 Martin Liska <mliska@suse.cz>
6240 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6241 * tree-pass.h (make_pass_lower_switch_O0): New function.
6242 * tree-switch-conversion.c (node_has_low_bound): Remove.
6243 (node_has_high_bound): Likewise.
6244 (node_is_bounded): Likewise.
6245 (class pass_lower_switch): Make it a template type and create
6247 (pass_lower_switch::execute): Add template argument.
6248 (make_pass_lower_switch): New function.
6249 (make_pass_lower_switch_O0): New function.
6250 (do_jump_if_equal): Remove.
6251 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6252 all the hard work to tree optimization passes.
6254 2018-05-18 Martin Liska <mliska@suse.cz>
6256 * dbgcnt.c (limit_low): Renamed from limit.
6257 (limit_high): New variable.
6258 (dbg_cnt_is_enabled): Check for upper limit.
6259 (dbg_cnt): Adjust dumping.
6260 (dbg_cnt_set_limit_by_index): Add new argument for high
6262 (dbg_cnt_set_limit_by_name): Likewise.
6263 (dbg_cnt_process_single_pair): Parse new format.
6264 (dbg_cnt_process_opt): Use strtok.
6265 (dbg_cnt_list_all_counters): Remove 'value' and add
6267 * doc/invoke.texi: Document changes.
6269 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6271 * doc/sourcebuild.texi (scalar_all_fma): Document.
6272 * tree.def (FMA_EXPR): Delete.
6273 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6274 * internal-fn.c (ternary_direct): New macro.
6275 (expand_ternary_optab_fn): Likewise.
6276 (direct_ternary_optab_supported_p): Likewise.
6277 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6278 * builtins.c (fold_builtin_fma): Delete.
6279 (fold_builtin_3): Don't call it.
6280 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6281 * expr.c (expand_expr_real_2): Likewise.
6282 * fold-const.c (operand_equal_p): Likewise.
6283 (fold_ternary_loc): Likewise.
6284 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6285 * gimple.c (DEFTREECODE): Likewise.
6286 * gimplify.c (gimplify_expr): Likewise.
6287 * optabs-tree.c (optab_for_tree_code): Likewise.
6288 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6289 * tree-eh.c (operation_could_trap_p): Likewise.
6290 (stmt_could_throw_1_p): Likewise.
6291 * tree-inline.c (estimate_operator_cost): Likewise.
6292 * tree-pretty-print.c (dump_generic_node): Likewise.
6293 (op_code_prio): Likewise.
6294 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6295 * tree-ssa-operands.c (get_expr_operands): Likewise.
6296 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6297 * fold-const-call.h (fold_fma): Delete.
6298 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6299 CFN_FNMA and CFN_FNMS.
6301 * genmatch.c (combined_fn): New enum.
6302 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6303 (commutative_op): New function.
6304 (commutate): Use it. Handle more than 2 operands.
6305 (dt_operand::gen_gimple_expr): Use commutative_op.
6306 (parser::parse_expr): Allow :c to be used with non-binary
6307 operators if the commutative operand is known.
6308 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6309 CFN_FMS, CFN_FNMA and CFN_FNMS.
6310 (backprop::process_assign_use): Remove FMA_EXPR handling.
6311 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6312 (gen_hsa_fma): New function.
6313 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6314 IFN_FNMA and IFN_FNMS.
6315 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6316 * gimple-fold.h (follow_all_ssa_edges): Declare.
6317 * gimple-fold.c (follow_all_ssa_edges): New function.
6318 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6319 gimple_build interface and use follow_all_ssa_edges to fold the result.
6320 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6321 instead of checking for optabs directly.
6322 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6323 rather than FMA_EXPRs.
6324 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6325 call to IFN_FMA instead of an FMA_EXPR.
6327 2018-05-17 Jim Wilson <jimw@sifive.com>
6329 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6330 sign extended promoted subreg, and the range does not have the sign bit
6331 set, then do a sign extend.
6333 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6334 test, check for sign extended subreg and/or constant operands, and
6335 do a sign extend in that case.
6337 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6339 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6340 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6342 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6343 Change logics_shift_reg to logics_shift_imm.
6344 (thunderx2t99_fp_loadpair_basic): Delete.
6345 (thunderx2t99_fp_storepair_basic): Delete.
6346 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6347 (thunderx2t99_asimd_polynomial): Delete.
6348 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6349 and neon_fp_mul_d_scalar_q.
6350 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6351 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6352 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6353 (thunderx2t99_asimd_lut): Add missing tbl types.
6354 (thunderx2t99_asimd_ext): Delete.
6355 (thunderx2t99_asimd_load1_1_mult): Delete.
6356 (thunderx2t99_asimd_load1_2_mult): Delete.
6357 (thunderx2t99_asimd_load1_ldp): New.
6358 (thunderx2t99_asimd_load1): New.
6359 (thunderx2t99_asimd_load2): Add missing *load2* types.
6360 (thunderx2t99_asimd_load3): New.
6361 (thunderx2t99_asimd_load4): New.
6362 (thunderx2t99_asimd_store1_1_mult): Delete.
6363 (thunderx2t99_asimd_store1_2_mult): Delete.
6364 (thunderx2t99_asimd_store2_mult): Delete.
6365 (thunderx2t99_asimd_store2_onelane): Delete.
6366 (thunderx2t99_asimd_store_stp): New.
6367 (thunderx2t99_asimd_store1): New.
6368 (thunderx2t99_asimd_store2): New.
6369 (thunderx2t99_asimd_store3): New.
6370 (thunderx2t99_asimd_store4): New.
6372 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6374 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6375 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6377 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6378 Segher Boessenkool <segher@kernel.crashing.org>
6381 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6384 2018-05-17 Richard Biener <rguenther@suse.de>
6386 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6387 for pruning loop and prune defs feeding only already visited PHIs.
6389 2018-05-17 Richard Biener <rguenther@suse.de>
6391 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6393 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6394 Richard Biener <rguenther@suse.de>
6396 PR tree-optimization/85793
6397 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6398 for VMAT_ELEMENTWISE.
6400 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6402 * internal-fn.h (lookup_internal_fn): Declare
6403 * internal-fn.c (lookup_internal_fn): New function.
6404 * gimple.c (gimple_build_call_from_tree): Handle calls to
6406 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6407 internal function names.
6408 * tree-pretty-print.c (dump_generic_node): Likewise.
6409 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6411 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6413 * gimple-fold.h (gimple_build): Make the function forms take
6414 combined_fn rather than built_in_function.
6415 (gimple_simplify): Likewise.
6416 * gimple-match-head.c (gimple_simplify): Likewise.
6417 * gimple-fold.c (gimple_build): Likewise.
6418 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6419 rather than gimple_build_call_internal.
6420 (get_initial_defs_for_reduction): Likewise.
6421 (vect_create_epilog_for_reduction): Likewise.
6422 (vectorizable_live_operation): Likewise.
6424 2018-05-17 Martin Liska <mliska@suse.cz>
6426 * gimple-ssa-sprintf.c (format_directive): Do not use
6427 space in between 'G_' and '('.
6429 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6432 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6433 even if the mask is not all ones.
6436 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6438 (ix86_gimple_fold_builtin): Likewise.
6441 * config/i386/i386.c: Include tree-vector-builder.h.
6442 (ix86_vector_shift_count): New function.
6443 (ix86_fold_builtin): Fold shift builtins by scalar count.
6444 (ix86_gimple_fold_builtin): Likewise.
6446 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6447 _mm512_setzero): New intrinsics.
6449 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6450 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6452 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6453 code generation for cases where splatting a value is not useful.
6454 * simplify-rtx.c (simplify_ternary_operation): Simplify
6455 vec_merge across a vec_duplicate and a paradoxical subreg forming
6456 a vector mode to a vec_concat.
6458 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6460 * config.gcc: Support "goldmont-plus".
6461 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6463 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6464 PROCESSOR_GOLDMONT_PLUS.
6465 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6466 (processor_target_table): Add "goldmont-plus".
6467 (PTA_GOLDMONT_PLUS): Define.
6468 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6469 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6470 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6471 (fold_builtin_cpu): Add "goldmont-plus".
6472 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6473 (ix86_option_override_internal): Add "goldmont-plus".
6474 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6475 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6476 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6477 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6479 2018-05-17 Richard Biener <rguenther@suse.de>
6481 PR tree-optimization/85757
6482 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6483 remove defs that only feed that PHI from further processing.
6485 2018-05-16 Jim Wilson <jimw@sifive.com>
6487 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6489 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6491 2018-05-16 Mark Wielaard <mark@klomp.org>
6493 * dwarf2out.c (count_index_strings): New function.
6494 (output_indirect_strings): Call count_index_strings and generate
6495 header for dwarf_version >= 5.
6497 2018-05-16 Mark Wielaard <mark@klomp.org>
6499 * dwarf2out.c (dwarf_FORM): New function.
6500 (set_indirect_string): Use dwarf_FORM.
6501 (reset_indirect_string): Likewise.
6502 (size_of_die): Likewise.
6503 (value_format): Likewise.
6504 (output_die): Likewise.
6505 (add_skeleton_AT_string): Likewise.
6506 (output_macinfo_op): Likewise.
6507 (index_string): Likewise.
6508 (output_index_string_offset): Likewise.
6509 (output_index_string): Likewise.
6510 (count_index_strings): Likewise.
6512 2018-05-16 Carl Love <cel@us.ibm.com>
6514 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6515 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6517 2018-05-16 Martin Jambor <mjambor@suse.cz>
6519 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6520 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6522 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6524 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6525 (fnma<mode>4): Likewise.
6526 (fms<mode>4): Likewise.
6527 (fnms<mode>4): Likewise.
6528 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6529 (aarch64_fnma<mode>4): Likewise.
6530 (aarch64_fms<mode>4): Likewise.
6531 (aarch64_fnms<mode>4): Likewise.
6532 (aarch64_fnmadd<mode>4): Likewise.
6534 2018-05-16 Jason Merrill <jason@redhat.com>
6536 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6538 2018-05-16 Richard Biener <rguenther@suse.de>
6540 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6541 (dump_stmt_cost): Declare.
6542 (add_stmt_cost): Dump cost we add.
6543 (add_stmt_costs): New function.
6544 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6546 (vect_analyze_stmt): Adjust prototype.
6547 (vectorizable_condition): Likewise.
6548 (vectorizable_live_operation): Likewise.
6549 (vectorizable_reduction): Likewise.
6550 (vectorizable_induction): Likewise.
6551 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6552 cost vector to pass to vectorizable_ and record afterwards.
6553 (vect_model_reduction_cost): Take cost vector argument and adjust.
6554 (vect_model_induction_cost): Likewise.
6555 (vectorizable_reduction): Likewise.
6556 (vectorizable_induction): Likewise.
6557 (vectorizable_live_operation): Likewise.
6558 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6559 SLP_TREE_NUMBER_OF_VEC_STMTS.
6560 (vect_analyze_slp_cost_1): Remove.
6561 (vect_analyze_slp_cost): Likewise.
6562 (vect_slp_analyze_node_operations): Take visited args and
6563 a target cost vector. Avoid processing already visited stmt sets.
6564 (vect_slp_analyze_operations): Use a local cost vector to gather
6565 costs and register those of non-discarded instances.
6566 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6567 (vect_schedule_slp_instance): Remove copying of
6568 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6570 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6571 adding cost. Record cost entry location.
6572 (vect_prologue_cost_for_slp_op): Function to compute cost of
6573 a constant or invariant generated for SLP vect in the prologue,
6574 split out from vect_analyze_slp_cost_1.
6575 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6576 (vect_model_promotion_demotion_cost): Likewise.
6577 (vect_model_store_cost): Likewise, make static.
6578 (vect_model_load_cost): Likewise.
6579 (vectorizable_bswap): Add cost vector arg and adjust.
6580 (vectorizable_call): Likewise.
6581 (vectorizable_simd_clone_call): Likewise.
6582 (vectorizable_conversion): Likewise.
6583 (vectorizable_assignment): Likewise.
6584 (vectorizable_shift): Likewise.
6585 (vectorizable_operation): Likewise.
6586 (vectorizable_store): Likewise.
6587 (vectorizable_load): Likewise.
6588 (vectorizable_condition): Likewise.
6589 (vectorizable_comparison): Likewise.
6590 (can_vectorize_live_stmts): Likewise.
6591 (vect_analyze_stmt): Likewise.
6592 (vect_transform_stmt): Adjust calls to vectorizable_*.
6593 * tree-vectorizer.c: Include gimple-pretty-print.h.
6594 (dump_stmt_cost): New function.
6596 2018-05-16 Richard Biener <rguenther@suse.de>
6598 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6599 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6600 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6601 (check_name): New callback.
6602 (dse_classify_store): Track cycles via a visited bitmap of PHI
6603 defs and simplify handling of in-loop and across loop dead stores
6604 and properly fail for loop-variant refs. Handle byte-tracking with
6605 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6608 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6610 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6611 (vect_get_mask_type_for_stmt): Likewise.
6612 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6614 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6615 to determine the statement's vector type and the vector type that
6616 should be used for calculating nunits. Deal with cases in which
6617 the type has to be deferred.
6618 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6619 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6620 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6621 (vect_determine_vf_for_stmt): New functions, split out from...
6622 (vect_determine_vectorization_factor): ...here.
6623 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6624 (vect_get_mask_type_for_stmt): New functions, split out from
6625 vect_determine_vectorization_factor.
6627 2018-05-16 Richard Biener <rguenther@suse.de>
6629 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6630 verify the [VEC_]COND_EXPR embedded comparison.
6632 2018-05-15 Martin Sebor <msebor@redhat.com>
6634 PR tree-optimization/85753
6635 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6636 RECORD_TYPE in addition to ARRAY_TYPE.
6638 2018-05-15 Martin Sebor <msebor@redhat.com>
6641 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6643 2018-05-15 Richard Biener <rguenther@suse.de>
6645 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6646 add by_clobber_p one. Change algorithm to collect all defs
6647 representing uses we need to walk and try reducing them to
6648 a single one before failing.
6649 (dse_dom_walker::dse_optimize_stmt): Adjust.
6651 2018-05-13 Mark Wielaard <mark@klomp.org>
6653 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6654 (size_of_loc_descr): Likewise.
6655 (output_loc_operands): Likewise.
6656 (output_loc_operands_raw): Likewise.
6657 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6658 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6659 (hash_loc_operands): Likewise.
6660 (compare_loc_operands): Likewise.
6662 2018-05-14 Mark Wielaard <mark@klomp.org>
6664 * dwarf2out.c (count_index_addrs): New function.
6665 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6667 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6669 PR tree-optimization/83648
6670 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6671 return value as malloc candidate.
6673 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6676 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6677 param as true in call to suggest_attribute.
6679 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6681 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6682 -mreadonly-in-sdata.
6684 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6686 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6688 (aarch64_crypto_aesd_fused): Likewise.
6690 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6692 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6693 (movsi_aarch64): Likewise.
6694 (load_pairsi): Likewise.
6695 (load_pairdi): Likewise.
6696 (store_pairsi): Likewise.
6697 (store_pairdi): Likewise.
6698 (load_pairsf): Likewise.
6699 (load_pairdf): Likewise.
6700 (store_pairsf): Likewise.
6701 (store_pairdf): Likewise.
6702 (zero_extend): Likewise.
6703 (trunc): Swap alternatives.
6704 (fcvt_target): Add '?' to prefer w over r.
6706 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6709 * config/i386/i386.md: Disallow non-commutative arithmetics in
6710 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6711 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6712 in the peephole2 before it.
6714 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6716 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6717 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6718 (ix86_handle_option): Handle -mcldemote.
6719 * config.gcc: New header.
6720 * config/i386/cldemoteintrin.h: New file.
6721 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6722 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6724 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6725 OPTION_MASK_ISA_CLDEMOTE.
6726 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6727 (ix86_valid_target_attribute_inner_p): Ditto.
6728 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6729 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6730 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6731 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6732 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6734 * config/i386/i386.opt: Add -mcldemote.
6735 * config/i386/x86intrin.h: New header.
6736 * doc/invoke.texi: Add -mcldemote.
6738 2018-05-14 Richard Biener <rguenther@suse.de>
6740 * doc/match-and-simplify.texi: Adjust :s documentation.
6742 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6744 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6745 intended memcpy size.
6746 (REORDER_45): Likewise.
6748 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6750 * sort.cc: New file.
6751 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6752 * vec.c (qsort_chk): Use gcc_qsort.
6753 * Makefile.in (OBJS-libcommon): Add sort.o.
6754 (build/sort.o): New target. Use it...
6755 (BUILD_RTL): ... here, and...
6756 (build/gencfn-macros): ... here, and...
6757 (build/genmatch): ... here.
6759 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6760 Chung-Ju Wu <jasonwucj@gmail.com>
6762 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6763 * config/nds32/nds32-graywolf.md: New file.
6764 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6765 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6767 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6768 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6769 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6770 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6771 * config/nds32/pipelines.md: Include n15 settings.
6773 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6774 Chung-Ju Wu <jasonwucj@gmail.com>
6776 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6777 * config/nds32/nds32-n13.md: New file.
6778 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6779 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6781 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6782 * config/nds32/nds32.md (pipeline_model): Add n13.
6783 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6784 * config/nds32/pipelines.md: Include n13 settings.
6786 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6787 Chung-Ju Wu <jasonwucj@gmail.com>
6789 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6790 * config/nds32/nds32-n10.md: New file.
6791 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6792 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6794 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6795 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6796 * config/nds32/nds32.md (pipeline_model): Add n10.
6797 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6798 * config/nds32/pipelines.md: Include n10 settings.
6800 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6801 Kito Cheng <kito.cheng@gmail.com>
6802 Chung-Ju Wu <jasonwucj@gmail.com>
6804 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6805 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6806 Add enum values for DSP extension instructions.
6807 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6809 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6810 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6812 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6813 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6814 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6815 * config/nds32/nds32-intrinsic.md: Likewise.
6816 * config/nds32/nds32_intrinsic.h: Likewise.
6817 * config/nds32/nds32-md-auxiliary.c: Likewise.
6818 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6819 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6820 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6821 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6822 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6823 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6824 TYPE_DMAC in switch statement.
6825 * config/nds32/nds32.c: New checking and implementation for DSP
6826 extension instructions.
6827 * config/nds32/nds32.h: Likewise.
6828 * config/nds32/nds32.md: Likewise.
6829 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6830 * config/nds32/predicates.md: Implement new predicates for DSP
6833 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6835 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6836 Reformat alternatives and attributes so it is easier to identify
6837 which constraints/attributes go with which instruction.
6838 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6839 (mov<mode>_softfloat32, FMOVE64): Likewise.
6840 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6841 (mov<mode>_softfloat64, FMOVE64): Likewise.
6843 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6845 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6847 (Basic PowerPC Built-in Functions): The new name of the
6848 subsection previously known as "PowerPC Built-in Functions".
6849 (Basic PowerPC Built-in Functions Available on all Configurations):
6851 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6852 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6853 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6854 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6856 2018-05-11 Martin Jambor <mjambor@suse.cz>
6859 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6862 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6865 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6867 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6869 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6870 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6871 (ix86_handle_option): Handle -mwaitpkg.
6872 * config.gcc: New header.
6873 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6874 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6875 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6877 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6878 OPTION_MASK_ISA_WAITPKG.
6879 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6880 (ix86_option_override_internal): Add PTA_WAITPKG.
6881 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6882 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6883 IX86_BUILTIN_TPAUSE.
6884 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6885 __builtin_ia32_umwait and __builtin_ia32_tpause.
6886 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6887 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6888 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6889 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6890 UNSPECV_TPAUSE): New.
6891 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6892 * config/i386/i386.opt: Add -mwaitpkg.
6893 * config/i386/waitpkgintrin.h: New file.
6894 * config/i386/x86intrin.h: New header.
6895 * doc/invoke.texi: Add -mwaitpkg.
6897 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6900 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6902 (cortex-m0): Use armv6s-m isa.
6903 (cortex-m0plus): Likewise.
6904 (cortex-m1): Likewise.
6905 (cortex-m0.small-multiply): Likewise.
6906 (cortex-m0plus.small-multiply): Likewise.
6907 (cortex-m1.small-multiply): Likewise.
6909 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6910 Jakub Jelinek <jakub@redhat.com>
6912 PR tree-optimization/85692
6913 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6914 source permute as well.
6916 2018-05-11 Martin Liska <mliska@suse.cz>
6919 * doc/extend.texi: Document LLVM style format for no_sanitize
6922 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6924 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6925 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6926 (mode_supports_vsx_dform_quad): Likewise.
6927 (mode_supports_vmx_dform): Move these functions to be next to the
6928 other mode_supports functions.
6929 (mode_supports_dq_form): Likewise.
6930 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6931 mode_supports_dq_form.
6932 (reg_offset_addressing_ok_p): Likewise.
6933 (offsettable_ok_by_alignment): Likewise.
6934 (rs6000_legitimate_offset_address_p): Likewise.
6935 (legitimate_lo_sum_address_p): Likewise.
6936 (rs6000_legitimize_address): Likewise.
6937 (rs6000_legitimize_reload_address): Likewise.
6938 (rs6000_secondary_reload_inner): Likewise.
6939 (rs6000_preferred_reload_class): Likewise.
6940 (rs6000_output_move_128bit): Likewise.
6942 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6944 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6945 Generate SImode target register for null target.
6946 <case IX86_BUILTIN_XGETBV>: Ditto.
6947 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6948 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6950 2018-05-10 Carl Love <cel@us.ibm.com>
6952 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6953 dcbtt and dcbtstt if operands[2] is 0.
6955 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6958 * config/i386/sse.md (usadv64qi): New expander.
6960 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6962 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6963 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6964 -maltivec=be support.
6965 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6966 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6967 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6968 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6969 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6970 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6971 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6972 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6973 altivec_vsumsws): Adjust.
6974 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6975 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6976 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6978 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6979 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6980 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6981 (altivec_lve<VI_char>x): Delete expand.
6982 (*altivec_lve<VI_char>x_internal): Rename to...
6983 (altivec_lve<VI_char>x): ... this.
6984 (altivec_lvxl_<mode>): Delete expand.
6985 (*altivec_lvxl_<mode>_internal): Rename to ...
6986 (altivec_lvxl_<mode>): ... this.
6987 (altivec_stvxl_<mode>): Delete expand.
6988 (*altivec_stvxl_<mode>_internal): Rename to ...
6989 (altivec_stvxl_<mode>): ... this.
6990 (altivec_stve<VI_char>x): Delete expand.
6991 (*altivec_stve<VI_char>x_internal): Rename to ...
6992 (altivec_stve<VI_char>x): ... this.
6993 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6994 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6995 reduc_plus_scal_<mode>): Adjust.
6996 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6998 (rs6000_cpu_cpp_builtins): Adjust.
6999 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7000 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7001 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7002 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7003 -maltivec=be support.
7004 (rs6000_split_vec_extract_var): Adjust.
7005 (rs6000_split_v4si_init): Adjust.
7006 (swap_selector_for_mode): Delete.
7007 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7008 altivec_expand_stvex_be): Delete.
7009 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7010 -maltivec=be support.
7011 (rs6000_gimple_fold_builtin): Ditto.
7012 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7014 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7015 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7016 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7017 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7018 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7019 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7020 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7021 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7022 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7023 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7024 anonymous split): Adjust.
7025 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7026 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7028 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7030 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7031 when --with-gxx-include-dir is also specified.
7032 * configure: Regenerate.
7034 2018-05-09 Jim Wilson <jimw@sifive.com>
7037 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7038 * config/riscv/t-withmultilib: New.
7039 * config/riscv/withmultilib.h: New.
7040 * doc/install.texi: Document RISC-V --with-multilib-list support.
7042 2018-05-09 Richard Biener <rguenther@suse.de>
7044 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7046 (vect_bb_vectorization_profitable_p): Adjust. Compute
7047 actual scalar cost using the cost vector and the add_stmt_cost
7050 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7052 PR rtl-optimization/85645
7053 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7054 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7056 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7058 PR rtl-optimization/85645
7059 * shrink-wrap.c (spread_components): Return a boolean saying if
7060 anything was changed.
7061 (try_shrink_wrapping_separate): Iterate spread_components until
7062 nothing changes anymore.
7064 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7066 PR rtl-optimization/85645
7067 * regrename.c (build_def_use): Also kill the chains that include the
7068 destination of a REG_CFA_REGISTER note.
7070 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7072 PR rtl-optimization/85645
7073 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7074 insn that has a REG_CFA_REGISTER note.
7076 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7078 * cfgexpand.c (expand_clobber): New function.
7079 (expand_gimple_stmt_1): Use it.
7080 * tree-vect-stmts.c (vect_clobber_variable): New function,
7082 (vectorizable_simd_clone_call): ...here.
7083 (vectorizable_store): Emit a clobber either side of an
7084 IFN_STORE_LANES sequence.
7085 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7087 2018-05-09 Tom de Vries <tom@codesourcery.com>
7090 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7091 (define_insn "trap_if_false"): Add exit after trap.
7093 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7095 PR rtl-optimization/85638
7096 * bb-reorder.c: Include common/common-target.h.
7097 (create_forwarder_block): New function extracted from...
7098 (fix_up_crossing_landing_pad): ...here. Rename into...
7099 (dw2_fix_up_crossing_landing_pad): ...this.
7100 (sjlj_fix_up_crossing_landing_pad): New function.
7101 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7102 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7103 from both partitions and exit the loop after one iteration.
7105 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7108 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7110 (Basic PowerPC Built-in Functions): The new name of the
7111 subsection previously known as "PowerPC Built-in Functions".
7112 (Basic PowerPC Built-in Functions Available on all Configurations):
7114 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7116 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7118 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7120 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7123 2018-05-08 Jim Wilson <jimw@sifive.com>
7125 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7126 (LD_EMUL_SUFFIX): New.
7127 (LINK_SPEC): Use it.
7129 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7131 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7133 (Basic PowerPC Built-in Functions): The new name of the
7134 subsection previously known as "PowerPC Built-in Functions".
7135 (Basic PowerPC Built-in Functions Available on all Configurations):
7137 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7139 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7141 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7143 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7146 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7149 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7150 after cmpelim optimization.
7152 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7154 * config.gcc: Support "goldmont".
7155 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7156 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7158 * config/i386/i386.c (m_GOLDMONT): Define.
7159 (processor_target_table): Add "goldmont".
7160 (PTA_GOLDMONT): Define.
7161 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7162 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7163 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7164 (fold_builtin_cpu): Add "goldmont".
7165 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7166 (ix86_option_override_internal): Add "goldmont".
7167 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7168 (processor_type): Add PROCESSOR_GOLDMONT.
7169 * config/i386/i386.md: Add CPU "glm".
7170 * config/i386/glm.md: New file.
7171 * config/i386/x86-tune.def: Add m_GOLDMONT.
7172 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7174 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7177 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7179 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7180 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7181 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7184 * config/i386/i386.c (ix86_fold_builtin): Handle
7185 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7188 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7189 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7191 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7194 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7195 (check_arch): Likewise.
7196 (check_fpu): Return the result rather than printing it.
7197 (end arch): Fix operator precedence.
7198 (end cpu): Likewise.
7199 (END): Print the result from check_fpu.
7201 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7202 Alan Hayward <alan.hayward@arm.com>
7203 David Sherwood <david.sherwood@arm.com>
7205 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7206 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7207 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7208 (*fcmuo<mode>_and): New patterns.
7210 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7212 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7213 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7214 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7215 (cmp_op, sve_imm_con): New code attributes.
7216 (SVE_COND_INT_CMP, imm_con): Delete.
7217 (cmp_op): Remove above unspecs from int attribute.
7218 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7220 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7221 comparison-specific unspecs.
7222 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7223 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7224 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7225 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7226 (*vec_fcm<cmp_op><mode>): Rename to...
7227 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7228 (*vec_fcmuo<mode>): Rename to...
7229 (*fcmuo<mode>): ...this and adjust likewise.
7230 (*pred_fcm<cmp_op><mode>): New pattern.
7231 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7232 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7234 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7236 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7237 (aarch64_emit_sve_predicated_cond): New function.
7238 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7239 (aarch64_emit_unspec_cond_or): Replace with...
7240 (aarch64_emit_sve_or_conds): ...this new function. Use
7241 aarch64_emit_sve_ptrue_op for the individual comparisons and
7242 aarch64_emit_binop to OR them together.
7243 (aarch64_emit_inverted_unspec_cond): Replace with...
7244 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7245 aarch64_emit_sve_ptrue_op for the comparison and
7246 aarch64_emit_unop to invert the result.
7247 (aarch64_expand_sve_vec_cmp_float): Update after the above
7248 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7250 2018-05-07 Nathan Sidwell <nathan@acm.org>
7252 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7253 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7254 (Backwards Compatibility): Likewise.
7256 2018-05-07 Luis Machado <luis.machado@linaro.org>
7260 2018-05-07 Luis Machado <luis.machado@linaro.org>
7262 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7263 <prefetch_dynamic_strides>: New const bool field.
7264 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7265 prefetch_dynamic_strides.
7266 (exynosm1_prefetch_tune): Likewise.
7267 (thunderxt88_prefetch_tune): Likewise.
7268 (thunderx_prefetch_tune): Likewise.
7269 (thunderx2t99_prefetch_tune): Likewise.
7270 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7272 (aarch64_override_options_internal): Update to set
7273 PARAM_PREFETCH_DYNAMIC_STRIDES.
7274 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7275 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7276 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7277 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7278 prefetch-dynamic-strides setting.
7280 2018-05-07 Luis Machado <luis.machado@linaro.org>
7282 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7283 <minimum_stride>: New const int field.
7284 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7285 minimum_stride field.
7286 (exynosm1_prefetch_tune): Likewise.
7287 (thunderxt88_prefetch_tune): Likewise.
7288 (thunderx_prefetch_tune): Likewise.
7289 (thunderx2t99_prefetch_tune): Likewise.
7290 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7291 (aarch64_override_options_internal): Update to set
7292 PARAM_PREFETCH_MINIMUM_STRIDE.
7293 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7294 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7295 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7296 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7297 stride is constant and is below the minimum stride threshold.
7299 2018-05-07 Luis Machado <luis.machado@linaro.org>
7301 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7304 2018-05-07 Luis Machado <luis.machado@linaro.org>
7306 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7307 <prefetch_dynamic_strides>: New const bool field.
7308 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7309 prefetch_dynamic_strides.
7310 (exynosm1_prefetch_tune): Likewise.
7311 (thunderxt88_prefetch_tune): Likewise.
7312 (thunderx_prefetch_tune): Likewise.
7313 (thunderx2t99_prefetch_tune): Likewise.
7314 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7316 (aarch64_override_options_internal): Update to set
7317 PARAM_PREFETCH_DYNAMIC_STRIDES.
7318 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7319 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7320 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7321 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7322 prefetch-dynamic-strides setting.
7324 2018-05-07 Luis Machado <luis.machado@linaro.org>
7326 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7327 <minimum_stride>: New const int field.
7328 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7329 minimum_stride field.
7330 (exynosm1_prefetch_tune): Likewise.
7331 (thunderxt88_prefetch_tune): Likewise.
7332 (thunderx_prefetch_tune): Likewise.
7333 (thunderx2t99_prefetch_tune): Likewise.
7334 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7335 (aarch64_override_options_internal): Update to set
7336 PARAM_PREFETCH_MINIMUM_STRIDE.
7337 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7338 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7339 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7340 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7341 stride is constant and is below the minimum stride threshold.
7343 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7346 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7347 the type is addressable. Don't force op into register if it has
7350 2018-05-05 Roland McGrath <mcgrathr@google.com>
7353 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7354 any section for which we don't know a specific type it should have,
7355 regardless of name. Previously this was done only for the exact
7356 names ".init_array", ".fini_array", and ".preinit_array".
7357 (default_elf_asm_named_section): Add comment about
7358 relationship with default_section_type_flags and SECTION_NOTYPE.
7359 (get_section): Don't consider it a type conflict if one side has
7360 SECTION_NOTYPE and the other doesn't, as long as neither has the
7361 SECTION_BSS et al used in the default_section_type_flags logic.
7363 2018-05-05 Tom de Vries <tom@codesourcery.com>
7366 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7367 (workaround_barsyncs): New function.
7368 (nvptx_reorg): Use workaround_barsyncs.
7369 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7370 (define_expand "nvptx_membar_cta"): New define_expand.
7371 (define_insn "*nvptx_membar_cta"): New insn.
7373 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7375 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7376 To improve optimization opportunities.
7377 * builtin-types.def: The new needed builtin types for the above.
7379 2018-05-04 Richard Biener <rguenther@suse.de>
7381 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7382 * gimple-ssa-store-merging.c
7383 (imm_store_chain_info::output_merged_store): Remove redundant create,
7384 release split_store vector contents on failure.
7385 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7386 scalar stmt vector on cache hit.
7388 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7390 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7392 * config.gcc (powerpc-xilinx-eabi*): Remove.
7393 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7395 (fusion_addis_mem_combo_load): Ditto.
7396 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7398 (rs6000_cpu_cpp_builtins): Ditto.
7399 * config/rs6000/rs6000-linux.c
7400 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7401 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7402 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7404 (rs6000_setup_reg_addr_masks): Ditto.
7405 (rs6000_init_hard_regno_mode_ok): Ditto.
7406 (rs6000_option_override_internal): Ditto.
7407 (legitimate_lo_sum_address_p): Ditto.
7408 (rs6000_legitimize_address): Ditto.
7409 (rs6000_legitimize_reload_address): Ditto.
7410 (rs6000_legitimate_address_p): Ditto.
7411 (abi_v4_pass_in_fpr): Ditto.
7412 (setup_incoming_varargs): Ditto.
7413 (rs6000_gimplify_va_arg): Ditto.
7414 (rs6000_split_multireg_move): Ditto.
7415 (rs6000_savres_strategy): Ditto.
7416 (rs6000_emit_prologue_components): Ditto.
7417 (rs6000_emit_epilogue_components): Ditto.
7418 (rs6000_emit_prologue): Ditto.
7419 (rs6000_emit_epilogue): Ditto.
7420 (rs6000_elf_file_end): Ditto.
7421 (rs6000_function_value): Ditto.
7422 (rs6000_libcall_value): Ditto.
7423 * config/rs6000/rs6000.h: Ditto.
7424 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7425 (TARGET_MINMAX): ... this. New.
7426 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7427 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7428 (*movsi_internal1_single): Delete.
7429 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7430 mfpu=, mxilinx-fpu): Delete.
7431 * config/rs6000/singlefp.h: Delete.
7432 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7433 * config/rs6000/t-rs6000: Ditto.
7434 * config/rs6000/t-xilinx: Delete.
7435 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7436 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7437 (VStype_simple): Delete.
7438 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7439 * config/rs6000/xfpu.h: Delete.
7440 * config/rs6000/xfpu.md: Delete.
7441 * config/rs6000/xilinx.h: Delete.
7442 * config/rs6000/xilinx.opt: Delete.
7443 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7444 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7446 2018-05-04 Tom de Vries <tom@codesourcery.com>
7449 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7452 2018-05-04 Richard Biener <rguenther@suse.de>
7455 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7456 (expand_complex_div_wide): Likewise.
7457 (expand_complex_operations_1): Likewise.
7458 (expand_complex_libcall): Preserve EH info of the original stmt.
7459 (tree_lower_complex): Handle removed blocks.
7460 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7461 on complex multiplication and division libcall builtins.
7463 2018-05-04 Richard Biener <rguenther@suse.de>
7466 * fold-const.c (negate_expr_p): Restrict negation of operand
7467 zero of a division to when we know that can happen without
7469 (fold_negate_expr_1): Likewise.
7471 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7474 * real.h (real_nextafter): Declare.
7475 * real.c (real_nextafter): New function.
7476 * fold-const-call.c (fold_const_nextafter): New function.
7477 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7478 CASE_CFN_NEXTTOWARD.
7479 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7480 even when arg1_mode is different from arg0_mode.
7482 2018-05-03 Nathan Sidwell <nathan@acm.org>
7484 * doc/extend.texi (Deprecated Features): Remove
7486 (Backwards Compatibility): Likewise.
7487 * doc/invoke.texi (C++ Language Options): Likewise.
7488 (C++ Dialect Options): Likewise.
7490 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7493 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7494 _mm512_mask_mullox_epi64): New intrinsics.
7496 2018-05-03 Tom de Vries <tom@codesourcery.com>
7499 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7500 dump files): Add offload-tree.
7502 2018-05-03 Richard Biener <rguenther@suse.de>
7504 PR tree-optimization/85615
7505 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7506 to loops not nested in BBs loop father to avoid creating multi-entry
7509 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7511 PR tree-optimization/70291
7512 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7513 arguments. Change return type to tree. Emit libcall as a new
7514 statement rather than replacing existing one when inplace_p is true.
7515 (expand_complex_multiplication_components): New function.
7516 (expand_complex_multiplication): Expand floating-point complex
7517 multiplication using the above.
7518 (expand_complex_division): Rename inner_type parameter to type.
7519 Update expand_complex_libcall call-site.
7520 (expand_complex_operations_1): Update expand_complex_multiplication
7521 and expand_complex_division call-sites.
7523 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7526 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7527 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7528 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7529 the highest significant bit of the shift count mask is clear. In
7530 check whether and[sq]i3 is needed verify that all significant bits
7531 of the shift count other than the highest are set.
7533 2018-05-02 Tom de Vries <tom@codesourcery.com>
7536 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7537 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7538 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7539 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7540 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7541 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7542 __builtin_goacc_parlevel_size.
7544 2018-05-02 Richard Biener <rguenther@suse.de>
7546 PR tree-optimization/85597
7547 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7548 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7551 2018-05-02 Tom de Vries <tom@codesourcery.com>
7554 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7555 dump files): Add ltrans-tree.
7557 2018-05-02 Tom de Vries <tom@codesourcery.com>
7560 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7561 dump files): Add wpa-ipa.
7563 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7565 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7566 powerpc*-*-linux*paired* target.
7567 * config/rs6000/750cl.h: Delete.
7568 * config/rs6000/paired.h: Delete.
7569 * config/rs6000/paired.md: Delete.
7570 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7572 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7573 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7574 comment. Remove paired float support.
7575 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7576 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7578 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7579 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7581 * config/rs6000/rs6000.c: Remove paired float support.
7582 (paired_expand_vector_init, paired_expand_vector_move,
7583 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7584 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7585 paired_expand_builtin, paired_expand_predicate_builtin,
7586 paired_init_builtins): Delete.
7587 * config/rs6000/rs6000.h: Remove paired float support.
7588 * config/rs6000/rs6000.md: Remove paired float support.
7589 (move_from_CR_ov_bit): Delete.
7590 * config/rs6000/rs6000.opt (mpaired): Delete.
7591 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7592 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7594 2018-05-02 Richard Biener <rguenther@suse.de>
7597 * gimplify.c (gimplify_save_expr): When in SSA form allow
7598 SAVE_EXPRs to compute to SSA vars.
7600 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7603 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7604 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7605 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7606 clobber operands[2], instead use a new pseudo. Formatting fixes.
7608 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7610 PR tree-optimization/85586
7611 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7612 exit early for statements in the same group if the accesses are
7615 2018-05-02 Tom de Vries <tom@codesourcery.com>
7618 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7621 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7623 PR tree-optimization/85143
7624 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7626 2018-05-01 Tom de Vries <tom@codesourcery.com>
7629 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7630 not found" error message.
7632 2018-05-01 Tom de Vries <tom@codesourcery.com>
7635 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7636 * vec.c (test_ordered_remove_if): New function.
7637 (vec_c_tests): Call test_ordered_remove_if.
7638 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7639 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7640 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7641 VEC_ORDERED_REMOVE_IF.
7643 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7645 PR tree-optimization/82665
7646 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7647 pointer subtraction where arguments come from a memchr call.
7649 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7651 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7652 --push-state --as-needed and --pop-state instead of --as-needed and
7653 --no-as-needed if ld supports it.
7654 * configure: Regenerated.
7657 * doc/install.texi2html: Replace _002d with - and _002a with * in
7658 generated html files using sed.
7660 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7663 * gcc-rich-location.c (blank_line_before_p): New function.
7664 (use_new_line): New function.
7665 (gcc_rich_location::add_fixit_insert_formatted): New function.
7666 * gcc-rich-location.h
7667 (gcc_rich_location::add_fixit_insert_formatted): New function.
7669 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7671 * selftest.c (assert_streq): Rename "expected" and "actual" to
7672 "val1" and "val2". Extend NULL-handling to cover both inputs
7673 symmetrically, while still requiring both to be non-NULL for a pass.
7674 * selftest.h (assert_streq): Rename "expected" and "actual" to
7676 (ASSERT_EQ): Likewise.
7677 (ASSERT_EQ_AT): Likewise.
7678 (ASSERT_KNOWN_EQ): Likewise.
7679 (ASSERT_KNOWN_EQ_AT): Likewise.
7680 (ASSERT_NE): Likewise.
7681 (ASSERT_MAYBE_NE): Likewise.
7682 (ASSERT_MAYBE_NE_AT): Likewise.
7683 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7684 the assertion to pass.
7685 (ASSERT_STREQ_AT): Likewise.
7687 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7689 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7690 interaction with -pie.
7692 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7694 * selftest.h: Fix alphabetization of per-source-file selftest
7697 2018-04-30 Jason Merrill <jason@redhat.com>
7699 PR c++/61982 - dead stores to destroyed objects.
7700 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7703 2018-04-30 Jason Merrill <jason@redhat.com>
7705 * tree.c (build_clobber): New.
7706 * tree.h: Declare it.
7707 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7709 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7711 * diagnostic-show-locus.c (layout::layout): Update for
7712 location_get_source_line returning a char_span.
7713 (struct char_span): Move to input.h.
7714 (struct correction): Update for fields in char_span becoming
7716 (struct source_line): Update for location_get_source_line
7717 returning a char_span.
7718 (layout::print_line): Likewise.
7719 * edit-context.c (edited_file::print_content): Likewise.
7720 (edited_file::print_diff_hunk): Likewise.
7721 (edited_file::print_run_of_changed_lines): Likewise.
7722 (edited_file::get_num_lines): Likewise.
7723 (edited_line::edited_line): Likewise.
7724 * final.c (asm_show_source): Likewise.
7725 * input.c (location_get_source_line): Convert return type
7726 from const char * to char_span, losing the final "line_len"
7728 (dump_location_info): Update for the above.
7729 (get_substring_ranges_for_loc): Likewise. Use a char_span
7730 when handling the literal within the line.
7731 (test_reading_source_line): Update for location_get_source_line
7732 returning a char_span.
7733 * input.h (class char_span): Move here from
7734 diagnostic-show-locus.c, converting from a struct to a class.
7735 Make data members private.
7736 (char_span::operator bool): New.
7737 (char_span::length): New.
7738 (char_span::get_buffer): New.
7739 (char_span::operator[]): New.
7740 (char_span::subspan): Make const.
7741 (char_span::xstrdup): New.
7742 (location_get_source_line): Convert return type from const char *
7743 to char_span, losing the final "line_size" param.
7745 2018-04-30 Jan Hubicka <jh@suse.cz>
7747 * lto-wrapper.c (ltrans_priorities): New static var.
7748 (cmp_priority): New.
7749 (run_gcc): Read priorities and if doing parallel build order
7750 the Makefile by them.
7752 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7754 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7756 2018-04-30 Richard Biener <rguenther@suse.de>
7758 * tree-cfg.c (verify_address): Remove base argument, add
7759 flag whether to check TREE_ADDRESSABLE and do that.
7760 (verify_expr): Remove.
7761 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7762 (verify_gimple_assign_single): Likewise.
7763 (verify_gimple_switch): Likewise.
7764 (verify_expr_location_1): Dereference tp once. Add (disabled)
7765 piece from verify_expr.
7766 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7768 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7770 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7772 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7774 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7775 (small_data_pattern): Likewise.
7776 (arc_rewrite_small_data): Likewise.
7777 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7778 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7779 (get_symbol_alignment): New function.
7780 (legitimate_small_data_address_p): Likewise.
7781 (legitimate_scaled_address): Update, call
7782 legitimate_small_data_address_p.
7783 (output_sdata): New static variable.
7784 (arc_print_operand): Update how we handle small data operands.
7785 (arc_print_operand_address): Likewise.
7786 (arc_legitimate_address_p): Update, use
7787 legitimate_small_data_address_p.
7788 (arc_rewrite_small_data_p): Remove.
7789 (arc_rewrite_small_data_1): Likewise.
7790 (arc_rewrite_small_data): Likewise.
7791 (small_data_pattern): Likewise.
7792 (compact_sda_memory_operand): Update to use
7793 legitimate_small_data_address_p and get_symbol_alignment.
7794 (prepare_move_operands): Don't rewite sdata pattern.
7795 (prepare_extend_operands): Remove.
7796 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7798 (zero_extendqisi2): Likewise.
7799 (zero_extendhisi2): Likewise.
7800 (extendqihi2): Likewise.
7801 (extendqisi2): Likewise.
7802 (extendhisi2): Likewise.
7806 * config/arc/constraints.md (Usd): Change it to memory constraint.
7808 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7810 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7811 as source of std instructions.
7812 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7813 allow 6-bit constants as source for store instructions.
7814 (movdi_insn): Update instruction pattern to allow 6-bit constants
7815 as source for store instructions.
7817 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7819 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7821 2018-04-30 Nathan Sidwell <nathan@acm.org>
7822 Sandra Loosemore <sandra@codesourcery.com>
7824 * dumpfile.c (dump_open): Allow '-' for stdout.
7825 * doc/invoke.texi (Developer Options): Document dump filename
7826 determination early. Document stdin/stdout selection.
7828 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7830 Microblaze Target: PIC data text relative
7832 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7833 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7835 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7836 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7837 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7838 New addressing mode for data-text relative position indepenedent code.
7839 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7840 'ADDRESS_SYMBOLIC_TXT_REL'.
7841 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7842 (microblaze_legitimate_pic_operand): Exclude function calls from
7843 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7844 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7846 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7847 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7848 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7849 for 'address + offset'.
7850 (microblaze_expand_prologue): Add new function prologue call for
7852 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7853 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7854 table in case of TARGET_PIC_DATA_TEXT_REL.
7855 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7856 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7857 Add new macros 'UNSPEC_TEXT',
7858 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7859 + exclude function calls from 'UNSPEC_PLT' in case of data text
7861 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7862 new target hook for generating address diff vector tables in case of
7864 * doc/tm.texi : Regenerate.
7865 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7866 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7867 of addr diff vector generation.
7868 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7869 target hook definition.
7870 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7871 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7872 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7873 option for data text relative.
7875 2018-04-30 Richard Biener <rguenther@suse.de>
7877 * tree-chrec.h (evolution_function_is_constant_p): Remove
7879 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7881 2018-04-30 Richard Biener <rguenther@suse.de>
7884 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7886 2018-04-30 Richard Biener <rguenther@suse.de>
7888 PR tree-optimization/28364
7889 PR tree-optimization/85275
7890 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7891 copying first exit test.
7893 2018-04-28 Mark Wielaard <mark@klomp.org>
7895 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7897 (dwarf_AT): Handle DW_AT_addr_base.
7898 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7900 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7903 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7904 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7905 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7906 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7908 2018-04-28 Richard Biener <rguenther@suse.de>
7910 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7911 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7912 to reflect use. Only add interesting stmts.
7914 2018-04-27 Martin Jambor <mjambor@suse.cz>
7917 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7918 the jump function allows for passing through aggregate values.
7920 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7922 * input.h (in_system_header_at): Convert from macro to inline
7924 (from_macro_expansion_at): Likewise.
7925 (from_macro_definition_at): Likewise.
7927 2018-04-27 Jeff Law <law@redhat.com>
7929 * config.gcc: Mark tile* targets as deprecated/obsolete.
7931 2018-04-27 Richard Biener <rguenther@suse.de>
7933 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7936 2018-04-27 Richard Biener <rguenther@suse.de>
7938 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7940 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7942 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7943 with Yd constraint. Set "preferred_for_speed" attribute from
7944 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7946 (*movdi_internal): Ditto.
7947 (movti_interunit splitters): Remove
7948 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7949 (movdi_interunit splitters): Ditto.
7950 * config/i386/constraints.md (Ye): Remove.
7951 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7953 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7956 * config/aarch64/constraints.md (Usg): Limit to 31.
7959 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7961 PR tree-optimization/85529
7962 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7963 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7964 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7965 zero extension or masking of the MSB bit.
7966 (optimize_range_tests): Add FIRST_BB argument, pass it through
7967 to optimize_range_tests_var_bound.
7968 (maybe_optimize_range_tests, reassociate_bb): Adjust
7969 optimize_range_tests callers.
7971 2018-04-26 Richard Biener <rguenther@suse.de>
7972 Jakub Jelinek <jakub@redhat.com>
7974 * cgraph.h (symbol_table): Just declare debug method here.
7975 * symtab.c (symbol_table::debug): Define.
7977 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7979 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7981 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7983 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7984 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7985 (*movdi_internal): Substitute Yi and Yj constraint with x
7986 and Ym and Yn constraint with y constraint. Update "isa"
7987 attribute and set "preferred_for_speed" attribute from
7988 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7989 (*movsi_internal): Ditto.
7990 (*movdf_internal): Ditto.
7991 (*movsf_internal): Ditto.
7992 (*zero_extendsidi2): Ditto.
7993 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7994 (sse2_loadld): Ditto.
7995 (*vec_extract<ssevecmodelower>_0): Ditto.
7996 (*vec_extractv4si_0_zext_sse4): Ditto.
7997 (vec_concatv2di): Ditto.
7998 (*vec_dup<mode>): Ditto.
7999 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8000 * config/i386/constraints.md (Yi): Remove.
8005 2018-04-26 Nathan Sidwell <nathan@acm.org>
8007 * dumpfile.c (dump_open): New.
8008 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8009 (dump_finish): Detect stdio/stderr by value not name.
8011 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8013 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8015 2018-04-26 Tom de Vries <tom@codesourcery.com>
8018 * config/nvptx/nvptx.c (verify_neutering_jumps)
8019 (verify_neutering_labels): New function
8020 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8022 2018-04-26 Tom de Vries <tom@codesourcery.com>
8025 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8026 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8029 2018-04-26 Richard Biener <rguenther@suse.de>
8030 Tom de Vries <tom@codesourcery.com>
8033 * lto-streamer-out.c (output_function): Fixup loops if required to match
8034 discovery done in the reader.
8036 2018-04-26 Richard Biener <rguenther@suse.de>
8038 PR tree-optimization/85116
8039 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8040 have a loop exit from the single latch predecessor. Remove
8041 case of header with just condition.
8042 (ch_base::copy_headers): Exclude infinite loops from any
8044 (pass_ch::execute): Record exits.
8046 2018-04-26 Richard Biener <rguenther@suse.de>
8048 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8049 prologue cost vector and pass it to vect_get_load_cost.
8050 (vect_get_peeling_costs_all_drs): Likewise.
8051 (vect_peeling_hash_get_lowest_cost): Likewise.
8052 (vect_enhance_data_refs_alignment): Likewise.
8054 2018-04-26 Richard Biener <rguenther@suse.de>
8057 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8058 checking of integer<->pointer conversions.
8059 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8060 sign-/zero-extending pointer types.
8061 (expand_omp_for_static_chunk): Likewise.
8063 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8064 Jean Lee <xiaoyur347@gmail.com>
8066 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8067 (TARGET_ASAN_SHADOW_OFFSET): Define.
8068 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8069 true for -fsanitize=address.
8071 2018-04-25 Mark Wielaard <mark@klomp.org>
8073 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8076 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8078 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8079 than "alu", remove explicit "memory" and "imm_disp" attributes.
8080 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8083 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8084 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8085 gen_lowpart_no_emit.
8087 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8090 * config/i386/i386.c (ix86_expand_builtin): Change memory
8091 operand to XI, extend p0 to Pmode.
8092 * config/i386/i386.md: Change unspec volatile and operand
8093 1 mode to XI, change operand 0 mode to P.
8095 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8097 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8098 GET_MODE_MASK before any checking.
8099 (nds32_can_use_bset_p): Likewise.
8100 (nds32_can_use_btgl_p): Likewise.
8102 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8104 * config/nds32/nds32-doubleword.md: New define_split pattern for
8105 illegal register number.
8107 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8109 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8111 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8113 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8115 2018-04-25 Richard Biener <rguenther@suse.de>
8117 * lto-streamer.h (LTO_major_version): Bump to 8.
8119 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8121 * BASE-VER: Set to 9.0.0.
8123 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8125 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8126 in __abskf2 and __powikf2.
8128 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8131 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8132 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8133 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8134 Use the above on operand 2. Reindent.
8135 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8137 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8140 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8142 * config/i386/i386.opt (mcet): Removed.
8143 * doc/install.texi: Remove -mcet documentation.
8144 * doc/invoke.texi: Likewise.
8146 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8149 * doc/install.texi: Remove -mcet from bootstrap-cet.
8151 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8154 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8155 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8159 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8160 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8161 containing a CONST_VECTOR.
8163 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8165 * doc/install.texi: Update newlib dependency for nvptx.
8167 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8170 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8171 instead of INTVAL when shifting x left.
8173 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8175 PR tree-optimization/85478
8176 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8177 vect_grouped_store_supported for single element vectors.
8179 2018-04-24 Richard Biener <rguenther@suse.de>
8182 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8183 load cost increase to the case of non-constant step.
8185 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8188 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8189 destination if any_malformed_asm.
8191 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8194 * expr.c (store_field): In the bitfield case, if the value comes from
8195 a function call and is returned in registers by means of a PARALLEL,
8196 do not change the mode of the temporary unless BLKmode and VOIDmode.
8198 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8200 PR rtl-optimization/85423
8201 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8202 dependencies to debug insns when the previous insn is non-debug.
8204 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8206 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8207 enums into a single definition.
8208 (fls): Fix predicates and printing.
8211 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8213 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8214 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8215 and short u6 immediate.
8216 (check_if_valid_sleep_operand): Remove.
8217 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8219 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8221 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8222 flag_always_save_lp condition.
8223 * config/nds32/nds32.opt (malways-save-lp): New option.
8225 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8227 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8228 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8229 * config/nds32/nds32.h
8230 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8231 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8233 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8235 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8236 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8238 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8239 Chung-Ju Wu <jasonwucj@gmail.com>
8241 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8243 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8244 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8246 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8248 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8250 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8252 * config/nds32/nds32-protos.h (nds32_data_alignment,
8253 nds32_local_alignment): Declare.
8254 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8255 nds32_local_alignment): New functions.
8256 (TARGET_CONSTANT_ALIGNMENT): Define.
8257 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8259 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8261 * config/nds32/nds32.c
8262 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8263 (TARGET_MODES_TIEABLE_P): Likewise.
8265 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8267 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8270 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8271 Chung-Ju Wu <jasonwucj@gmail.com>
8273 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8274 for unaligned access.
8275 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8277 * config/nds32/nds32-intrinsic.md: Likewise.
8278 * config/nds32/nds32_intrinsic.h: Likewise.
8279 * config/nds32/nds32.h (nds32_builtins): Likewise.
8280 * config/nds32/nds32.opt (munaligned-access): New option.
8281 * config/nds32/nds32.c (nds32_asm_file_start): Display
8282 flag_unaligned_access status.
8284 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8286 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8287 -mno-relax is present.
8288 * config/riscv/linux.h (LINK_SPEC): Ditto.
8290 2018-04-20 Martin Sebor <msebor@redhat.com>
8293 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8295 (gimple_fold_builtin_stxcpy_chk): Same.
8296 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8298 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8301 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8302 __powikf2 when long double is IEEE 128-bit.
8304 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8306 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8307 step to make sure stack always aligned.
8309 2018-04-20 Carl Love <cel@us.ibm.com>
8312 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8313 size check for arg0.
8315 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8316 Tom de Vries <tom@codesourcery.com>
8319 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8320 Emit insns for calls too.
8321 (nvptx_find_par): Always look for worker-level predecessor insn.
8322 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8324 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8325 (nvptx_process_pars): Propagate frames for calls.
8327 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8330 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8332 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8333 (ix86_handle_option): Don't handle OPT_mibt.
8334 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8336 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8338 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8339 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8340 (ix86_target_macros): Define __CET__ with flag_cf_protection
8341 for -fcf-protection.
8342 * config/i386/i386.c (isa2_opts): Remove -mibt.
8343 * config/i386/i386.h (TARGET_IBT): Removed.
8344 (TARGET_IBT_P): Likewise.
8345 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8346 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8347 * config/i386/i386.opt (mcet): Update help message.
8350 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8351 -mcet as an alias for -mshstk.
8353 2018-04-20 Richard Biener <rguenther@suse.de>
8356 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8357 complexity by forcing a single use of the multiply operand.
8359 2018-04-20 Martin Jambor <mjambor@suse.cz>
8362 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8363 recursion dependency to only apply to non-clones.
8365 2018-04-20 Martin Jambor <mjambor@suse.cz>
8368 * ipa-cp.c (create_specialized_node): Check that clones of
8369 self-recursive edges exist during IPA-CP.
8371 2018-04-19 Toon Moene <toon@moene.org>
8373 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8376 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8378 PR tree-optimization/85467
8379 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8380 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8381 VECTOR_CST element to type.
8383 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8386 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8387 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8388 (builtin_longjmp): Likewise.
8389 (save_stack_nonlocal): New pattern.
8390 (restore_stack_nonlocal): Likewise.
8392 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8395 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8396 Replace ASM_OUTPUT_LABEL with fprintf.
8398 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8401 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8402 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8403 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8404 define __IBT__ and __SHSTK__ for -fcf-protection.
8405 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8407 (ix86_trampoline_init): Likewise.
8408 (x86_output_mi_thunk): Likewise.
8409 (ix86_notrack_prefixed_insn_p): Likewise.
8410 (ix86_option_override_internal): Don't disallow -fcf-protection.
8411 * config/i386/i386.md (rdssp<mode>): Also enable for
8413 (incssp<mode>): Likewise.
8414 (nop_endbr): Likewise.
8415 * config/i386/i386.opt (mcet): Change help message to built-in
8419 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8420 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8421 enable CET built-in functions.
8423 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8425 * common/config/i386/i386-common.c
8426 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8427 OPTION_MASK_ISA_MOVDIRI_UNSET,
8428 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8429 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8430 * config.gcc (movdirintrin.h): New header.
8431 * config/i386/cpuid.h (bit_MOVDIRI,
8432 bit_MOVDIR64B): New bits.
8433 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8435 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8436 (VOID, PVOID, PCVOID)): New function types.
8437 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8438 __builtin_ia32_directstoreu_u64,
8439 __builtin_ia32_movdir64b): New builtins.
8440 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8441 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8443 (ix86_valid_target_attribute_inner_p): Ditto.
8444 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8445 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8446 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8447 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8448 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8449 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8450 (movdiri<mode>, movdir64b_<mode>): New.
8451 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8452 * config/i386/immintrin.h: Include movdirintrin.h.
8453 * config/i386/movdirintrin.h: New file.
8454 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8456 2018-04-19 Richard Biener <rguenther@suse.de>
8459 * cfg.c (clear_bb_flags): When loop state says we have
8460 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8462 2018-04-19 Richard Biener <rguenther@suse.de>
8464 PR tree-optimization/84737
8465 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8466 copying restrict info.
8467 (vect_setup_realignment): Use it.
8468 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8469 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8470 the first DR to all generated stores.
8471 (vectorizable_load): Likewise for loads.
8473 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8475 PR tree-optimization/85446
8476 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8477 the integral and pointer types to have the same precision.
8479 * doc/install.texi: Document --disable-cet being the default and
8482 2018-04-18 Martin Liska <mliska@suse.cz>
8484 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8487 2018-04-18 Martin Liska <mliska@suse.cz>
8490 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8493 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8494 arguments if they are comparable.
8496 2018-04-18 Martin Liska <mliska@suse.cz>
8499 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8502 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8505 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8508 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8509 ENDBR after calling __morestack.
8511 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8514 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8515 by using gcc_base_ver to generate a gcc_driver_version, and use
8516 it when generating GCC_DRIVER_NAME.
8517 * configure: Regenerate.
8519 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8522 * config.gcc: Obsolete powerpc*-*-*spe*.
8524 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8527 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8528 (stabstr_D): Change type of unum from unsigned int to
8529 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8532 2018-04-17 Jim Wilson <jimw@sifive.com>
8535 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8536 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8537 Set arg_pointer_offset after using pretend_args_size.
8539 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8541 PR rtl-optimization/85431
8542 * dse.c (record_store): Ignore zero width stores.
8545 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8546 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8547 __builtin_stack_restore rather than after it.
8548 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8549 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8550 argument instead of virtual_dynamic_stack_rtx.
8552 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8554 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8556 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8557 Add note to error message to explain internal mapping of overloaded
8558 built-in function name to non-overloaded built-in function name.
8559 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8562 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8565 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8566 where the inputs overlap with the output.
8568 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8570 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8571 (=v, v) alternative and explicit "memory" attribute.
8572 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8573 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8575 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8576 "sselog1" type instead of "sselog".
8577 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8578 "sselog". Remove explicit "memory" attribute.
8579 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8580 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8582 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8583 "sselog1" type instead of "sselog". Remove explicit "memory"
8585 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8586 use "sselog1" type instead of "sselog". Remove explicit "memory"
8588 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8589 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8591 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8592 "sselog1" type instead of "sselog". Remove explicit "memory"
8594 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8595 use "sselog1" type instead of "sselog". Remove explicit "memory"
8599 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8602 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8605 2018-04-17 Martin Jambor <mjambor@suse.cz>
8608 * ipa-cp.c (create_specialized_node): Call
8609 expand_all_artificial_thunks if necessary.
8611 2018-04-17 Martin Liska <mliska@suse.cz>
8614 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8615 in message, remote space in between '_G' and '('.
8617 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8620 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8621 avx512f_vmcmp<mode>3<round_saeonly_name>,
8622 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8623 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8624 avx512f_rndscale<mode><round_saeonly_name>,
8625 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8626 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8627 Use %<iptr>2 instead of %2 for -masm=intel.
8628 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8629 avx512f_vcvttss2usi<round_saeonly_name>,
8630 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8632 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8633 avx512f_vcvttsd2usi<round_saeonly_name>,
8634 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8635 Use %q1 instead of %1 for -masm=intel.
8636 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8637 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8638 of %3 for -masm=intel.
8639 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8641 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8643 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8645 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8646 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8648 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8650 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8651 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8652 %g1 and one with %0 and %1.
8653 (avx512er_vmrcp28<mode><round_saeonly_name>,
8654 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8656 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8657 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8658 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8659 of %0 and %{%4%} for -masm=intel.
8660 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8661 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8662 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8663 order of %0 and %{%5%}%{z%} for -masm=intel.
8665 2018-04-17 Jan Hubicka <jh@suse.cz>
8668 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8670 2018-04-17 Martin Liska <mliska@suse.cz>
8673 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8674 for target_clone error message. Make default implementation
8675 clone to be a local declaration.
8676 (separate_attrs): Add new argument and check for an empty
8678 (expand_target_clones): Handle it.
8679 (ipa_target_clone): Make redirection just for target_clones
8682 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8683 Tom de Vries <tom@codesourcery.com>
8686 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8687 tiled basic blocks without omp continue statements.
8689 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8692 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8693 vec_extract expression as having side effects to make sure it gets
8696 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8699 * config/i386/i386.c (get_builtin_code_for_version): Check
8702 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8705 * gcc/config.gcc: Support "skylake".
8706 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8708 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8709 (processor_target_table): Add "skylake".
8710 (ix86_option_override_internal): Add "skylake".
8711 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8712 PROCESSOR_CANNONLAKE.
8713 (get_builtin_code_for_version): Fix priority for
8714 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8715 PROCESSOR_SKYLAKE-AVX512.
8716 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8717 (processor_type): Add PROCESSOR_SKYLAKE.
8719 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8720 Jason Merrill <jason@redhat.com>
8723 * convert.c (convert_to_integer_1): Use direct recursion for
8724 enumeral types and types with a precision less than the number
8725 of bits in their mode.
8727 2018-04-16 Julia Koval <julia.koval@intel.com>
8730 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8731 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8733 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8736 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8737 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8738 and -mno-direct-move.
8740 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8743 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8744 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8745 Ensure negative shifts result in {0}.
8747 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8749 PR rtl-optimization/79916
8750 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8751 regs (if any) to define how to gnerate SD moves when LRA is in
8754 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8756 PR rtl-optimization/85393
8757 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8758 * except.c (expand_dw2_landing_pad_for_region): Make static.
8759 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8760 a label and unconditional jump to old_bb, rather than
8761 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8764 PR rtl-optimization/85376
8765 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8766 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8767 instead of a specific value.
8769 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8770 Bin Cheng <bin.cheng@arm.com>
8772 PR tree-optimization/82965
8773 PR tree-optimization/83991
8774 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8775 by_profile_only parameter.
8776 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8777 information if the loop was predicted to iterate too many times.
8778 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8780 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8783 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8786 2018-04-13 Martin Liska <mliska@suse.cz>
8787 Jakub Jelinek <jakub@redhat.com>
8790 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8791 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8792 * builtins.c (expand_builtin_memory_copy_args): Use
8793 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8794 handle dest_addr == pc_rtx.
8796 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8799 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8800 asked to not generate direct moves.
8801 (fix_trunc<mode>si2_stfiwx): Similar.
8802 (fix_trunc<mode>si2_internal): Similar.
8804 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8807 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8808 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8809 lookup if dest in some wider mode is known to be const0_rtx and
8810 if so, record permanent equivalence for it to be ZERO_EXTEND of
8811 the narrower mode destination.
8813 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8815 * lto-streamer-out.c (output_function): Revert 259346.
8816 * omp-expand.c (expand_oacc_for): Likewise.
8818 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8820 PR rtl-optimization/85354
8821 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8822 * sel-sched.c (sel_global_init): ... here.
8824 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8827 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8828 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8829 mode for PE-COFF targets.
8830 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8831 (i386_pe_asm_lto_end): Likewise.
8832 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8833 (TARGET_ASM_LTO_END): Likewise.
8834 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8835 (i386_pe_asm_lto_start): New function.
8836 (i386_pe_asm_lto_end): Likewise.
8838 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8839 Richard Biener <rguenther@suse.de>
8842 * lto-streamer-out.c (output_function): Fix CFG loop state before
8844 * omp-expand.c (expand_oacc_for): Handle calls to internal
8845 functions like regular functions.
8847 2018-04-12 Richard Biener <rguenther@suse.de>
8850 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8851 for the early LTO debug to properly generate references to it
8852 during DIE emission. Do not re-use that for the skeleton for
8854 (dwarf2out_early_finish): Likewise.
8856 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8859 * config/i386/sse.md
8860 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8861 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8862 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8863 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8864 and output is a reg, avoid creating invalid lowpart subreg, but
8865 instead split into a 512-bit move. Don't split if not AVX512VL,
8866 input is xmm16+ reg and output is a mem.
8867 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8868 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8869 xmm16+ reg and output is a mem.
8871 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8873 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8874 also for flag_dwarf2_cfi_asm.
8876 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8878 PR rtl-optimization/85342
8879 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8880 a bool scalar var inside of the loop instead. Don't try to update
8881 recog_data.operand after failed apply_change_group.
8883 2018-04-12 Tom de Vries <tom@codesourcery.com>
8886 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8887 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8888 array with flexible array member as array without given dimension.
8889 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8890 argument for undefined param to true.
8892 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8895 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8896 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8897 from PowerPC section.
8898 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8899 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8900 help text that is too long.
8901 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8902 help text that is too long.
8903 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8904 help text that is too long.
8906 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8908 * config/alpha/alpha.md (stack_probe_internal): Rename
8909 from "probe_stack". Update all callers.
8911 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8913 PR rtl-optimization/84566
8914 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8915 sched_macro_fuse_insns.
8917 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8920 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8921 (compute_block_dependences): ... from here.
8923 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8925 PR tree-optimization/85331
8926 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8927 from int to HOST_WIDE_INT.
8929 2018-04-11 Martin Jambor <mjambor@suse.cz>
8932 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8933 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8934 not the same as the source val.
8935 (cgraph_edge_brings_value_p): New parameter.
8936 (gather_edges_for_value): Pass destination value to
8937 cgraph_edge_brings_value_p.
8938 (perhaps_add_new_callers): Likewise.
8939 (get_info_about_necessary_edges): Likewise and exclude values brought
8940 only by self-recursive edges.
8941 (create_specialized_node): Redirect only clones of self-calling edges.
8942 (+self_recursive_pass_through_p): New function.
8943 (find_more_scalar_values_for_callers_subset): Use it.
8944 (find_aggregate_values_for_callers_subset): Likewise.
8945 (known_aggs_to_agg_replacement_list): Removed.
8946 (decide_whether_version_node): Re-calculate known constants for all
8947 remaining context clones.
8949 2018-04-11 Richard Biener <rguenther@suse.de>
8952 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8953 from early DWARF output.
8954 (dwarf2out_early_finish): Output line info unconditionally into
8955 early DWARF and add reference to it.
8957 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8960 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8961 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8962 other than V2DFmode using iptr mode attribute.
8963 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8965 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8967 PR rtl-optimization/84659
8968 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8970 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8973 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8975 (output_loc_list): Pass address of a dummy size variable even in the
8976 locview handling loop.
8977 (index_location_lists): Add comment on why skip_loc_list_entry can't
8980 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8983 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8986 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8989 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8990 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8991 and -mstring-compare-inline-limit.
8993 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8996 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8997 for stack clash protection in a register whenever we need it to be in
9000 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9002 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9003 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9005 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9008 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9010 (mlong-double-): Ditto.
9011 * config/rs6000/sysv4.opt (msdata=): Ditto.
9012 (mtls-size=): Ditto.
9014 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9016 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9017 erroneous entries for
9018 "vector int vec_ldl (int, long int *)", and
9019 "vector unsigned int vec_ldl (int, unsigned long int *)".
9020 Add comments and entries for
9021 "vector bool char vec_ldl (int, bool char *)",
9022 "vector bool short vec_ldl (int, bool short *)",
9023 "vector bool int vec_ldl (int, bool int *)",
9024 "vector bool long long vec_ldl (int, bool long long *)",
9025 "vector pixel vec_ldl (int, pixel *)",
9026 "vector long long vec_ldl (int, long long *)",
9027 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9028 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9029 type tree bool_long_long_type_node and correct definition of
9030 bool_V2DI_type_node to make reference to this new type tree.
9031 (rs6000_mangle_type): Replace erroneous reference to
9032 bool_long_type_node with bool_long_long_type_node.
9033 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9034 comments to emphasize sign distinctions for char and int types and
9035 replace RS6000_BTI_bool_long constant with
9036 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9037 use of RS6000_BTI_pixel.
9038 (bool_long_type_node): Remove this macro definition.
9039 (bool_long_long_type_node): New macro definition
9041 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9043 PR rtl-optimization/85300
9044 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9045 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9046 simplify_unary_operation fails.
9048 2018-04-10 Martin Liska <mliska@suse.cz>
9050 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9051 cgraph_edge and ipa_ref.
9053 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9057 * config/i386/sse.md
9058 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9059 computation of the VEC_MERGE selector from mask.
9060 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9061 Fix decoding of the VEC_MERGE selector into mask.
9063 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9065 PR tree-optimization/85286
9066 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9068 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9070 * final.c (final_1): Set insn_last_address as well as
9071 insn_current_address.
9073 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9076 * explow.c (emit_stack_probe): Call validize_mem on memory location
9077 before passing it to gen_probe_stack. Create address operand and
9078 legitimize it for the probe_stack_address case.
9080 2018-04-09 Jan Hubicka <jh@suse.cz>
9083 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9084 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9085 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9087 (free_lang_data): Rebuild type inheritance graph.
9089 2018-04-09 Martin Sebor <msebor@redhat.com>
9091 * invoke.texi (-finline-small-functions): Mention other optimization
9093 (-findirect-inlining, -fpartial-inlining): Same.
9094 (-finline-functions-called-once): Same.
9095 (-freorder-blocks-and-partition): Same.
9097 2018-04-09 Jan Hubicka <jh@suse.cz>
9100 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9101 jumps; choose last target that matches the criteria (i.e.
9102 no partition changes for non-crossing jumps).
9103 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9104 support for redirecting crossing jumps to non-crossing.
9106 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9108 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9109 also for naked functions.
9111 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9113 * config/arc/arc.md (add_shift): New pattern.
9114 (add_shift2): Likewise.
9115 (sub_shift): Likewise.
9116 (sub_shift_cmp0_noout): Likewise.
9117 (compare_si_ashiftsi): Likewise.
9118 (xbfu_cmp0_noout): New combine pattern.
9119 (xbfu_cmp0"): Likewise.
9120 (movsi_set_cc_insn): Place the predicable variant first.
9121 (commutative_binary_cmp0_noout): Remove clobber.
9122 (commutative_binary_cmp0): New pattern.
9123 (noncommutative_binary_cmp0): Likewise.
9124 (noncommutative_binary_cmp0_noout): Likewise.
9125 (noncommutative_binary_comparison_result_used): Removed.
9126 (rsub_cmp0): New pattern.
9127 (rsub_cmp0_noout): Likewise.
9128 (extzvsi): Changed, keep only meaningful variants.
9129 (SQH, SEZ): New iterators.
9130 (SQH_postfix): New mode attribute.
9131 (SEZ_prefix): New code attribute.
9132 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9133 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9134 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9136 (noncommutative_operator): Check the availability of barrel
9139 2018-04-09 Richard Biener <rguenther@suse.de>
9141 PR tree-optimization/85284
9142 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9143 Only use the niter constraining form of simple_iv when the exit
9146 2018-04-09 Tom de Vries <tom@codesourcery.com>
9149 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9150 (define_expand "*memory_barrier"): New define_expand.
9151 (define_insn "memory_barrier"): New insn.
9153 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9155 PR rtl-optimization/80463
9156 PR rtl-optimization/83972
9157 PR rtl-optimization/83480
9159 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9160 correct producer for the insn.
9161 (tidy_control_flow): Fixup seqnos in case of debug insns.
9163 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9165 PR rtl-optimization/83913
9167 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9168 different sched-times when merging exprs.
9170 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9172 PR rtl-optimization/83962
9174 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9175 tidy_fallthru_edge and tidy_control_flow.
9177 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9179 PR rtl-optimization/83530
9181 * sel-sched.c (force_next_insn): New global variable.
9182 (remove_insn_for_debug): When force_next_insn is true, also leave only
9183 next insn in the ready list.
9184 (sel_sched_region): When the region wasn't scheduled, make another pass
9185 over it with force_next_insn set to 1.
9187 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9189 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9191 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9192 for interrupt control.
9193 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9194 functions for interrupt control.
9195 * config/nds32/nds32-intrinsic.md: Likewise.
9196 * config/nds32/nds32_intrinsic.h: Likewise.
9197 * config/nds32/nds32.h (nds32_builtins): Likewise.
9199 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9201 * config/nds32/nds32.c (nds32_init_machine_status,
9202 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9203 strict_aligned_p field.
9204 (nds32_expand_to_rtl_hook): New function.
9205 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9206 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9208 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9209 Chung-Ju Wu <jasonwucj@gmail.com>
9211 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9212 * config/nds32/nds32-n7.md: New file.
9213 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9214 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9216 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9217 * config/nds32/nds32.md (pipeline_model): Add n7.
9218 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9219 * config/nds32/pipelines.md: Include n7 settings.
9221 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9222 Chung-Ju Wu <jasonwucj@gmail.com>
9224 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9225 * config/nds32/nds32-e8.md: New file.
9226 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9227 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9229 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9230 * config/nds32/nds32.md (pipeline_model): Add e8.
9231 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9232 * config/nds32/pipelines.md: Include e8 settings.
9234 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9235 Chung-Ju Wu <jasonwucj@gmail.com>
9237 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9238 * config/nds32/nds32-n8.md: New file.
9239 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9240 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9242 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9243 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9244 * config/nds32/nds32.md (pipeline_model): Add n8.
9245 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9246 * config/nds32/pipelines.md: Include n8 settings.
9248 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9249 Chung-Ju Wu <jasonwucj@gmail.com>
9251 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9252 * config/nds32/nds32-n9-2r1w.md: New file.
9253 * config/nds32/nds32-n9-3r2w.md: New file.
9254 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9255 nds32_register_ports): New or modify for cpu n9.
9256 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9258 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9259 * config/nds32/nds32-utils.c: New file.
9260 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9261 TARGET_MUL_SLOW): Define.
9262 * config/nds32/nds32.md (pipeline_model): New attribute.
9263 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9264 New options that support cpu n9.
9265 * config/nds32/pipelines.md: Include n9 settings.
9266 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9268 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9270 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9271 information if necessary.
9272 (output_cond_branch_compare_zero): Likewise.
9273 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9274 (nds32_target_alignment): Refine for alignment.
9275 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9276 (FUNCTION_BOUNDARY): Modify.
9277 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9279 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9281 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9283 * config/nds32/constants.md (unspec_volatile_element): Add values for
9284 TLB operation and data prefetch.
9285 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9286 functions for TLB operation and data prefetch.
9287 * config/nds32/nds32-intrinsic.md: Likewise.
9288 * config/nds32/nds32_intrinsic.h: Likewise.
9289 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9290 (nds32_print_operand): Likewise.
9291 * config/nds32/nds32.h (nds32_builtins): Likewise.
9293 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9294 Andrew Pinski <pinsika@gcc.gnu.org>
9297 * match.pd: Use constant_boolean_node of correct type instead of
9298 boolean_true_node or boolean_false_node for simplifying
9299 pointer comparisons to zero.
9301 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9303 PR tree-optimization/80021
9304 * tree.c (verify_type_variant): Make error call in verify_variant_match
9305 translatable and remove final full stop.
9307 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9309 * config/nds32/constants.md (unspec_volatile_element): Add
9310 UNSPEC_VOLATILE_EH_RETURN.
9311 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9312 nds32_output_stack_pop): Support dwarf exception handling process.
9313 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9314 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9315 exception handling process.
9316 (nds32_compute_stack_frame): Likewise.
9317 (nds32_return_addr_rtx): Likewise.
9318 (nds32_initial_elimination_offset): Likewise.
9319 (nds32_expand_prologue): Likewise.
9320 (nds32_expand_epilogue): Likewise.
9321 (nds32_dynamic_chain_address): New function.
9322 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9324 (DYNAMIC_CHAIN_ADDRESS): Define.
9325 (EH_RETURN_DATA_REGNO): Define.
9326 (EH_RETURN_STACKADJ_RTX): Define.
9327 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9328 patterns for dwarf exception handling.
9330 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9332 * config/nds32/nds32.h: Clean up obsolete macros.
9334 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9336 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9337 Add enum values for particular instructions.
9338 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9339 particular intrinsic functions.
9340 * config/nds32/nds32-intrinsic.md: Likewise.
9341 * config/nds32/nds32_intrinsic.h: Likewise.
9342 * config/nds32/nds32.h (nds32_builtins): Likewise.
9343 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9344 (btst, ave): New patterns for particular instructions.
9346 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9348 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9349 Add enum values for atomic load/store and memory sync.
9350 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9352 * config/nds32/nds32-intrinsic.md: Likewise.
9353 * config/nds32/nds32_intrinsic.h: Likewise.
9354 * config/nds32/nds32.h (nds32_builtins): Likewise.
9356 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9358 PR tree-optimization/85257
9359 * fold-const.c (native_encode_vector): If not all elts could fit
9360 and off is -1, return 0 rather than offset.
9361 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9362 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9363 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9364 adjust buffer in native_interpret_expr call.
9366 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9368 * config/nds32/constants.md (unspec_volatile_element): Add cache
9369 control enum values.
9370 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9371 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9372 * config/nds32/nds32.c (nds32_cctl_names): New.
9373 (nds32_print_operand): Handle cache control register names.
9374 * config/nds32/nds32.h (nds32_builtins): New enum values.
9375 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9377 * config/nds32/nds32.md (type): Add mmu.
9378 * config/nds32/pipelines.md (simple_insn): Add mmu.
9380 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9382 * config/nds32/nds32.md (type): Remove call.
9383 * config/nds32/pipelines.md (simple_insn): Likewise.
9385 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9387 * config/nds32/constants.md (unspec_volatile_element): Add
9388 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9389 UNSPEC_VOLATILE_FMFCFG.
9390 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9391 description for fmfcfg and fmfcsr.
9392 (bdesc_1arg): Add fmtcsr.
9393 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9394 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9395 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9396 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9397 unspec_fmfcfg): New patterns.
9398 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9399 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9400 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9401 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9402 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9403 __nds32__fmfcfg): Define.
9405 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9407 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9408 intrinsic register names.
9409 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9410 intrinsic register enum values and macros.
9412 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9414 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9415 for load/store addressing form.
9416 (nds32_print_operand_address): Likewise.
9418 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9421 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9422 based on LABEL_REF. Remove useless assertion.
9423 (pic_address_needs_scratch): Fix formatting.
9424 (sparc_legitimize_pic_address): Minor tweaks.
9425 (sparc_delegitimize_address): Adjust assertion accordingly.
9426 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9427 into symbolic_operand.
9428 (movsi_high_pic_label_ref): Likewise.
9429 (movsi_lo_sum_pic_label_ref): Likewise.
9430 (movdi_pic_label_ref): Likewise.
9431 (movdi_high_pic_label_ref): Likewise.
9432 (movdi_lo_sum_pic_label_ref): Likewise.
9434 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9436 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9437 custom LIB_SPEC setup.
9439 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9440 Kito Cheng <kito.cheng@gmail.com>
9442 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9443 * config/riscv/freebsd.h: New.
9445 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9447 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9448 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9451 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9452 Kito Cheng <kito.cheng@gmail.com>
9454 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9455 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9456 * config/nds32/nds32-protos.h (nds32_output_call,
9457 nds32_output_return): Declare.
9458 * config/nds32/nds32.md: Refine all the call and return patterns.
9460 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9463 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9464 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9466 PR rtl-optimization/84872
9467 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9468 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9471 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9473 * expr.c (copy_blkmode_to_reg): Revert 254862.
9474 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9476 2018-04-06 Richard Biener <rguenther@suse.de>
9479 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9480 after seeing a component reference with an adjacent field. Treat
9481 refs to arrays at struct end of external decls similar to
9482 refs to unconstrained commons.
9484 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9487 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9488 look through SAVE_EXPRs with non-side-effects argument. Adjust
9490 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9493 2018-04-06 Richard Biener <rguenther@suse.de>
9496 * alias.c (find_base_term): New wrapper around find_base_term
9497 unwinding CSELIB_VAL_PTR changes.
9498 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9501 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9503 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9505 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9506 constant definitions.
9507 ("nop"): lr 0,0 -> nopr r0
9508 ("nop_lr0", "nop_lr1"): New insn definitions.
9510 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9512 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9513 NDS32_V3PUSH_AVAILABLE_P macro.
9515 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9516 Chung-Ju Wu <jasonwucj@gmail.com>
9518 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9519 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9520 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9521 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9522 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9523 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9524 * config/nds32/constraints.md: New constraints and checking for hard
9525 float configuration.
9526 * config/nds32/iterators.md: New mode iterator and attribute for hard
9527 float configuration.
9528 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9530 * config/nds32/nds32-fpu.md: New file.
9531 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9532 deal with hard float code generation.
9533 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9535 (abi_type, float_reg_number): New enum type.
9536 * config/nds32/nds32-predicates.c: New predicates for hard float.
9537 * config/nds32/nds32-protos.h: Declare functions for hard float.
9538 * config/nds32/nds32.c: Implementation for hard float configuration.
9539 * config/nds32/nds32.h: Definitions for hard float configuration.
9540 * config/nds32/nds32.md: Include hard float machine description and
9541 modify patterns for hard float configuration.
9542 * config/nds32/nds32.opt: New options for hard float configuration.
9543 * config/nds32/predicates.md: New predicates for hard float
9546 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9548 * common/config/nds32/nds32-common.c
9549 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9551 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9554 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9555 CONSTRUCTOR_ELT (ctor, ...)->value.
9557 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9560 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9562 2018-04-05 Tom de Vries <tom@codesourcery.com>
9565 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9568 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9569 Kito Cheng <kito.cheng@gmail.com>
9571 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9572 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9573 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9574 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9576 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9577 Kito Cheng <kito.cheng@gmail.com>
9579 * config/nds32/constraints.md (Ufe): New memory constraint.
9580 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9581 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9582 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9584 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9585 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9587 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9589 * config/nds32/nds32.md: Use optimize_size in the condition for
9590 alu-shift instructions.
9592 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9594 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9596 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9598 * config/nds32/nds32.md (negsi2): Refine pattern.
9600 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9601 Chung-Ju Wu <jasonwucj@gmail.com>
9603 * config/nds32/iterators.md (shift_rotate): New code iterator.
9604 (shift): New code attribute.
9605 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9606 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9607 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9608 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9609 bit-wise operations.
9610 (andsi3, *andsi3): Ditto.
9611 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9612 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9613 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9614 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9615 nds32_ior_operand, nds32_xor_operand): New predicates.
9617 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9619 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9620 (addsi3, subsi3): ... this.
9622 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9624 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9626 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9628 * config/nds32/nds32.md: Adjust indention.
9630 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9632 * config/nds32/nds32.md (feature): New attribute.
9634 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9636 * config/nds32/nds32.md (subtype): New attribute.
9638 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9641 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9642 expansion to perform a bitwise AND of the argument followed by a
9643 boolean negation of the result.
9645 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9647 PR rtl-optimization/84878
9648 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9649 the basic block. Assert the use reference is not artificial and that
9650 it has an associated insn.
9652 2018-04-04 Michael Matz <matz@suse.de>
9654 * builtins.c (compute_objsize): Pass correct operand
9655 to array_at_struct_end_p.
9657 2018-04-04 Richard Biener <rguenther@suse.de>
9660 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9661 from contexts for DINFO_LEVEL_TERSE and below.
9663 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9665 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9666 resiter_operand condition.
9667 * config/nds32/nds32.md (*move<mode>): Ditto.
9669 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9670 Monk Chiang <sh.chiang04@gmail.com>
9672 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9674 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9676 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9678 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9679 Kito Cheng <kito.cheng@gmail.com>
9681 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9682 nds32_cond_code_str, output_cond_branch,
9683 output_cond_branch_compare_zero, nds32_expand_cbranch,
9684 nds32_expand_cstore, nds32_expand_movcc,
9685 nds32_output_cbranchsi4_equality_zero,
9686 nds32_output_cbranchsi4_equality_reg,
9687 nds32_output_cbranchsi4_equality_reg_or_const_int,
9688 nds32_output_cbranchsi4_greater_less_zero: New functions.
9689 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9690 nds32_expand_cstore, nds32_expand_movcc,
9691 nds32_output_cbranchsi4_equality_zero,
9692 nds32_output_cbranchsi4_equality_reg,
9693 nds32_output_cbranchsi4_equality_reg_or_const_int,
9694 nds32_output_cbranchsi4_greater_less_zero): Declare.
9695 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9696 nds32_rimm11s_operand): New predicates.
9697 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9698 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9701 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9703 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9704 * config/nds32/nds32.md: Ditto.
9705 * config/nds32/pipelines.md: Ditto.
9707 2018-04-04 Richard Biener <rguenther@suse.de>
9709 PR tree-optimization/85168
9710 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9711 propagating abnormals.
9713 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9715 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9717 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9718 Kito Cheng <kito.cheng@gmail.com>
9720 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9721 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9722 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9723 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9724 * config/nds32/nds32.md (sibcall_internal): New.
9725 (sibcall_register): Remove.
9726 (sibcall_immediate): Remove.
9727 (sibcall_value_internal): New.
9728 (sibcall_value_register): Remove.
9729 (sibcall_value_immediate): Remove.
9730 * config/nds32/predicates.md (nds32_general_register_operand): New.
9731 (nds32_call_address_operand): New.
9733 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9735 PR rtl-optimization/85167
9736 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9737 bb_defs if *split_p, instead preinitialize it to NULL.
9739 PR tree-optimization/85156
9740 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9741 evaluating the argument multiple times.
9743 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9745 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9747 (_mm_cvtpd_ps): Likewise.
9748 (_mm_cvttpd_epi32): Likewise.
9749 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9750 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9751 vector, pixel, and bool following altivec.h include.
9753 2018-04-03 Martin Sebor <msebor@redhat.com>
9755 * doc/extend.texi (Common Function Attributes): Clarify.
9756 (const attribute): Likewise.
9757 (pure attribute): Likewise.
9759 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9762 * config/i386/i386.c (ix86_expand_vector_set): Use
9763 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9765 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9767 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9768 instructions when changing rounding bits to preserve precision bits
9769 in the x87 control word.
9771 2018-04-03 Martin Liska <mliska@suse.cz>
9773 PR tree-optimization/82491
9774 * rtl.h (strip_offset_and_add): Replace += suboffset with
9775 poly_uint64 () + suboffset.
9777 2018-03-29 Martin Liska <mliska@suse.cz>
9778 Martin Jambor <mjambor@suse.cz>
9781 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9782 param_type is not an integral or pointer type.
9784 2018-04-03 Richard Biener <rguenther@suse.de>
9786 * sese.h (recompute_all_dominators): Remove.
9788 2018-04-02 Martin Sebor <msebor@redhat.com>
9790 * doc/invoke.texi (-Wrestrict): Fix typos.
9792 2018-04-02 Jim Wilson <jimw@sifive.com>
9794 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9795 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9796 (<optab>di3, <optab>si3_extend): Likewise.
9797 (<optab>si3_mask, <optab>si3_mask_1): New.
9798 (<optab>di3_mask, <optab>di3_mask_1): New.
9799 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9800 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9801 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9803 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9805 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9808 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9810 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9811 (nds32_canonicalize_comparison): New function.
9813 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9814 Kito Cheng <kito.cheng@gmail.com>
9815 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9817 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9818 * config/nds32/constants.md (unspec_volatile_element): Add
9819 UNSPEC_VOLATILE_RELAX_GROUP.
9820 * config/nds32/nds32-relax-opt.c: New file.
9821 * config/nds32/nds32-predicates.c
9822 (nds32_symbol_load_store_p): New function.
9823 * config/nds32/nds32-protos.h
9824 (nds32_symbol_load_store_p): Declare function.
9825 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9826 * config/nds32/nds32.c
9827 (nds32_register_pass): New function to register pass.
9828 (nds32_register_passes): New function to register passes.
9829 * config/nds32/nds32.md (relax_group): New pattern.
9830 * config/nds32/nds32.opt (mrelax-hint): New option.
9831 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9833 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9835 * config/nds32/t-nds32: Modify files dependency.
9837 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9839 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9840 (PROFILE_HOOK): Define its implementation.
9842 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9844 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9845 type and 32-bit size.
9847 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9850 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9851 (V_128_256): New mode iterator.
9852 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9853 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9854 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9856 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9857 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9859 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9862 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9863 NaN inputs correctly.
9865 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9868 * config/rs6000/vsx.md (??r): New mode attribute.
9869 (*vsx_mov<mode>_64bit): Use it.
9870 (*vsx_mov<mode>_32bit): Likewise.
9872 2018-03-30 Martin Sebor <msebor@redhat.com>
9874 PR tree-optimization/84818
9875 * builtins.c (check_access): Use warning_n.
9877 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9880 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9882 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9885 2018-03-30 Julia Koval <julia.koval@intel.com>
9888 * x86-tune.def (movx, partial_reg_dependency): Enable for
9891 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9894 * lra-constraints.c (process_alt_operands): Move setting
9895 this_alternative_matches below.
9897 2018-03-29 Martin Liska <mliska@suse.cz>
9900 * doc/invoke.texi: Document how LTO works with debug info.
9901 Describe auto-load support of binutils. Mention 'x86-64'
9902 as valid option value of -march option.
9904 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9906 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9909 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9910 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9911 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9914 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9917 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9918 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9919 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9920 for RS6000_BTM_POWERPC64.
9921 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9922 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9923 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9928 2018-03-28 Carl Love <cel@us.ibm.com>
9931 2017-09-27 Carl Love <cel@us.ibm.com>
9933 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9934 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9935 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9938 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9940 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9941 instead of __vector bool.
9942 (_mm_max_pu8): Likewise.
9943 (_mm_min_pi16): Likewise.
9945 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9948 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9949 (DIVWEUO): Likewise.
9951 (DIVDEUO): Likewise.
9952 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9953 DIVWEUO and DIVDEUO.
9954 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9955 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9956 (div_extend): Likewise.
9957 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9959 (__builtin_divweuo): Likewise.
9960 (__builtin_divdeo): Likewise.
9961 (__builtin_divdeuo): Likewise.
9963 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9966 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9967 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9969 PR tree-optimization/82004
9970 * gimple-match-head.c (optimize_pow_to_exp): New function.
9971 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9972 Don't fold to exp if optimize_pow_to_exp is false.
9974 2018-03-28 Martin Liska <mliska@suse.cz>
9977 * calls.c (initialize_argument_information): Fix trailing space.
9978 * common.opt: Fix typo and provide better explanation for
9979 -fsanitize-coverage option.
9980 * config/i386/i386.opt: Fix typo.
9982 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9983 Martin Liska <mliska@suse.cz>
9986 * gimplify.c (asan_poison_variable): Don't do the check for
9987 gimplify_omp_ctxp here.
9988 (gimplify_decl_expr): Do it here.
9989 (gimplify_target_expr): Likewise.
9991 2018-03-28 Martin Liska <mliska@suse.cz>
9994 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9995 chkp_type_bounds_count if MPX is not enabled.
9997 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9999 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10001 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10004 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10005 function to create the function decl for complex long double
10006 multiply and divide for -mabi=ieeelongdouble.
10007 (init_float128_ieee): Call it.
10009 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10012 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10013 -fcf-protection=branch -mibt.
10014 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10016 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10019 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10020 arm_word_relocations.
10022 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10025 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10026 extern array declarations.
10028 2018-03-27 Richard Biener <rguenther@suse.de>
10030 PR middle-end/84067
10031 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10032 explicit single_use checks.
10034 2018-03-27 Richard Biener <rguenther@suse.de>
10036 PR tree-optimization/85082
10037 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10040 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10042 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10043 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10044 Turn on fasynchronous-unwind-tables and funwind-tables.
10046 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10049 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10050 (*bmi_blsr_<mode>_ccz): Ditto.
10052 2018-03-26 Tom de Vries <tom@codesourcery.com>
10054 PR tree-optimization/85063
10055 * omp-general.c (offloading_function_p): New function. Factor out
10057 * omp-offload.c (pass_omp_target_link::gate): ... here.
10058 * omp-general.h (offloading_function_p): Declare.
10059 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10060 with attribute omp declare target for offloading functions.
10062 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10064 PR tree-optimization/84005
10065 * tree-data-ref.h (get_base_for_alignment): Declare.
10066 * tree-data-ref.c (get_base_for_alignment_1): New function.
10067 (get_base_for_alignment): Likewise.
10068 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10069 get_base_for_alignment to find a suitable base object, instead
10070 of always using drb->base_address.
10072 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10074 PR inline-asm/85022
10075 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10076 known size by default.
10078 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10080 PR inline-asm/85030
10081 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10082 and non BLKmode operands.
10084 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10087 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10088 Clean up attributes.
10090 2018-03-23 Richard Biener <rguenther@suse.de>
10093 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10094 we are going to emit early debug for LTO.
10096 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10098 PR inline-asm/85034
10099 * function.c (match_asm_constraints_1): Don't optimize if input
10100 doesn't satisfy general_operand predicate for output's mode.
10102 PR inline-asm/85022
10103 * alias.c (write_dependence_p): Don't require for x_canonicalized
10104 non-VOIDmode if x has VOIDmode.
10107 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10108 just don't try to optimize it rather than assert it never happens.
10110 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10112 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10113 macro expansions for definition of ST_INTERNAL_<mode> and
10114 LD_INTERNAL_<mode> builtins.
10115 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10117 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10119 (altivec_expand_st_builtin): Likewise.
10120 (altivec_expand_builtin): Remove calls to deleted functions.
10121 (rs6000_address_for_altivec): Delete this function.
10122 * config/rs6000/vector.md: Remove expands for
10123 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10125 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10128 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10129 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10130 re-computing once computed.
10131 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10132 (arm_init_machine_status): Initialize
10133 machine->static_chain_stack_bytes.
10135 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10138 * doc/extend.texi: Add four new prototypes for vec_ld.
10139 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10140 definitions for more logical presentation.
10141 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10142 entries for V1TI variants of __builtin_altivec_ld builtin.
10143 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10144 handling of V1TI variant of LVX icode pattern.
10145 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10146 (rs6000_gimple_fold_builtin): Likewise.
10147 (altivec_init_builtins): Add code to define
10148 __builtin_altivec_lvx_v1ti function.
10150 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10152 PR inline-asm/84941
10153 * function.c (match_asm_constraints_1): Don't do the optimization
10154 if input isn't a REG, SUBREG, MEM or constant.
10156 2018-03-22 Tom de Vries <tom@codesourcery.com>
10158 PR tree-optimization/84956
10159 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10160 bb_has_abnormal_pred.
10162 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10165 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10166 DECL_INITIAL (decl) to decl at the end.
10167 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10168 adjust the comment.
10170 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10172 * doc/extend.texi (__builtin_tgmath): Document when complex
10173 integer types are treated as _Complex _Float64.
10175 2018-03-21 Tom de Vries <tom@codesourcery.com>
10177 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10179 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10181 PR tree-optimization/84960
10182 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10183 if it is ENTRY block, move them into single succ of ENTRY in that case.
10185 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10187 PR tree-optimization/84811
10188 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10189 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10190 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10191 (known_subrange_p): Update accordingly. Cast each value involved
10192 in the size comparison, rather than casting the result of the
10195 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10197 PR tree-optimization/84982
10198 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10199 by flipping the least significant bit rather than all bits from
10200 bitpos to bitpos + bitsize - 1.
10202 2018-03-21 Nathan Sidwell <nathan@acm.org>
10204 * doc/extend.texi (Deprecated Features): Remove mention of
10205 long-deleted deprecations.
10207 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10210 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10211 * configure: Regenerate.
10213 2018-03-21 Tom de Vries <tom@codesourcery.com>
10215 PR tree-optimization/83126
10216 * tree-parloops.c (num_phis): New function.
10217 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10219 2018-03-21 Nathan Sidwell <nathan@acm.org>
10221 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10222 mention anon-struct/union members and trailing attributes.
10224 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10226 PR tree-optimization/84969
10227 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10228 builtin memset partitions if they set different rhs values.
10230 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10232 PR rtl-optimization/84989
10233 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10234 VEC_DUPLICATE with scalar result mode.
10236 2018-03-21 Martin Liska <mliska@suse.cz>
10239 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10240 not intended return statement.
10242 2018-03-21 Martin Liska <mliska@suse.cz>
10245 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10246 (chkp_find_bound_slots_1): Limit number of iterations.
10248 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10251 * Minor grammar fixes for x86 options.
10253 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10256 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10257 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10259 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10262 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10263 (altivec_lvx_<mode>_1op): Likewise.
10264 (altivec_stvx_<mode>_2op): Likewise.
10265 (altivec_stvx_<mode>_1op): Likewise.
10266 (altivec_lvx_<VM2:mode>): New define_expand.
10267 (altivec_stvx_<VM2:mode>): Likewise.
10268 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10269 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10270 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10271 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10272 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10273 (rs6000_gen_lvx): Likewise.
10274 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10275 (altivec_expand_stv_builtin): Likewise.
10276 (altivec_expand_builtin): Likewise.
10277 * config/rs6000/vector.md: Likewise.
10279 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10282 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10285 2018-03-20 Richard Biener <rguenther@suse.de>
10288 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10289 sign-conversions as zero, fall back to standard scalar_stmt
10292 2018-03-20 Martin Liska <mliska@suse.cz>
10295 * predict.c (rebuild_frequencies): Handle case when we have
10296 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10298 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10301 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10302 flag_section_anchors.
10303 * varasm.c (use_blocks_for_decl_p): Remove hack for
10304 dw2_force_const_mem.
10307 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10309 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10310 be created, use lowpart_subreg of operands[0] rather than operands[0]
10312 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10313 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10314 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10315 and n constraint instead of aarch64_shift_imm_di and Usd.
10316 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10317 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10319 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10322 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10323 to favor GPR over NEON registers.
10324 (<shift>di3_neon): Likewise.
10326 2018-03-20 Tom de Vries <tom@codesourcery.com>
10329 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10330 (nvptx_process_pars): Emit bar.sync asap and alap.
10332 2018-03-20 Tom de Vries <tom@codesourcery.com>
10335 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10336 seen_label if seen_label is already set.
10338 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10341 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10342 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10343 Use 1U instead of 1. Formatting fixes.
10346 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10347 instead of TREE_TYPE (s1) for the return value.
10349 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10351 PR tree-optimization/84946
10352 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10353 bitsize + bitsize in poly_uint64 rather than poly_int64.
10356 * dwarf2asm.c: Include fold-const.c.
10357 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10358 of decl rather than decl itself.
10360 PR rtl-optimization/84643
10361 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10363 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10366 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10367 calling assemble_variable.
10369 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10372 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10373 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10375 2018-03-19 Jim Wilson <jimw@sifive.com>
10378 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10379 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10380 (riscv_first_stack_step): Likewise.
10381 (riscv_option_override): Use STACK_BOUNDARY instead of
10382 MIN_STACK_BOUNDARY.
10383 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10384 MIN_STACK_BOUNDARY.
10385 (BIGGEST_ALIGNMENT): Set to 128.
10386 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10387 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10390 2018-03-19 Richard Biener <rguenther@suse.de>
10392 PR tree-optimization/84933
10393 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10394 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10396 2018-03-19 Richard Biener <rguenther@suse.de>
10398 PR tree-optimization/84859
10399 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10400 (cond_if_else_store_replacement): Perform sinking operation on
10401 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10402 Generalize what a BB with a single eligible store is.
10404 2018-03-19 Richard Biener <rguenther@suse.de>
10406 PR tree-optimization/84929
10407 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10408 chrec_is_positive against non-chrec arg.
10410 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10413 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10415 2018-03-18 Martin Liska <mliska@suse.cz>
10417 PR rtl-optimization/84635
10418 * regrename.c (build_def_use): Use matches_mode only when
10421 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10423 PR tree-optimization/84913
10424 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10425 vectorize chains of COND_EXPRs.
10427 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10429 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10431 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10433 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10435 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10437 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10439 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10440 Kito Cheng <kito.cheng@gmail.com>
10442 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10443 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10444 (nds32_adjust_reg_alloc_order): New function.
10445 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10447 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10449 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10450 nds32_print_operand, nds32_print_operand_address): Use
10451 HOST_WIDE_INT_PRINT_DEC instead.
10453 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10455 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10457 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10460 * config/i386/i386.c (initial_ix86_tune_features,
10461 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10462 unsigned long long.
10463 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10464 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10465 rather than 1u << ix86_tune. Formatting fix.
10466 (ix86_option_override_internal): Change ix86_arch_mask from
10467 unsigned int to unsigned HOST_WIDE_INT, initialize to
10468 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10469 (ix86_function_specific_restore): Likewise.
10471 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10474 * postreload.c (reload_combine_recognize_pattern): Perform
10475 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10476 truncate_int_for_mode the result for the destination's mode.
10479 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10480 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10482 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10485 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10488 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10489 regno_allocno_class_array and sorted_pseudos.
10490 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10491 insns where regno is used.
10493 2018-03-16 Martin Liska <mliska@suse.cz>
10496 * multiple_target.c (create_dispatcher_calls): Redirect
10497 reference in the symbol table.
10499 2018-03-16 Martin Liska <mliska@suse.cz>
10502 * multiple_target.c (create_dispatcher_calls): Redirect also
10505 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10509 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10510 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10511 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10513 2018-03-16 Julia Koval <julia.koval@intel.com>
10515 * doc/invoke.texi (Skylake Server): Add CLWB.
10516 Cannonlake): Remove CLWB.
10518 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10520 PR tree-optimization/84841
10521 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10523 (FLOAT_ONE_CONST_TYPE): Define.
10524 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10525 (sort_by_operand_rank): Put entries with higher constant_type last
10526 rather than first to match comments.
10528 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10530 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10533 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10536 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10537 check and use address of target always.
10539 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10542 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10543 (indirect_thunk_bnd_needed): Likewise.
10544 (indirect_thunks_used): Likewise.
10545 (indirect_thunks_bnd_used): Likewise.
10546 (indirect_return_needed): New.
10547 (indirect_return_bnd_needed): Likewise.
10548 (output_indirect_thunk_function): Add a bool argument for
10550 (output_indirect_thunk_function): Don't generate alias for
10551 function return thunk.
10552 (ix86_code_end): Call output_indirect_thunk_function to generate
10553 function return thunks.
10554 (ix86_output_function_return): Set indirect_return_bnd_needed
10555 and indirect_return_needed instead of indirect_thunk_bnd_needed
10556 and indirect_thunk_needed.
10558 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10560 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10561 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10562 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10564 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10565 Paul Hua <paul.hua.gm@gmail.com>
10568 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10570 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10572 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10573 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10574 resp. SFmode cases.
10576 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10579 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10580 instead of GET_MODE_SIZE when comparing Units.
10582 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10585 * varasm.c (hash_section): Return an unchangeble hash value
10586 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10587 Return !aarch64_can_use_per_function_literal_pools_p ().
10589 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10592 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10593 rather than address of cmode as last argument to prepare_cmp_insn.
10595 2018-03-15 Julia Koval <julia.koval@intel.com>
10597 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10598 F_AVX512VNNI, F_AVX512BITALG): New.
10600 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10603 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10604 insn for floating-point loads and stores.
10606 2018-03-14 Carl Love <cel@us.ibm.com>
10608 * config/rs6000/rs6000-c.c: Add macro definitions for
10609 ALTIVEC_BUILTIN_VEC_PERMXOR.
10610 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10611 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10612 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10613 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10615 * config/doc/extend.texi: Add prototypes for vec_permxor.
10617 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10620 * diagnostic-show-locus.c (class layout_point): Convert m_line
10621 from int to linenum_type.
10622 (line_span::comparator): Use linenum "compare" function when
10623 comparing line numbers.
10624 (test_line_span): New function.
10625 (layout_range::contains_point): Convert param "row" from int to
10627 (layout_range::intersects_line_p): Likewise.
10628 (layout::will_show_line_p): Likewise.
10629 (layout::print_source_line): Likewise.
10630 (layout::should_print_annotation_line_p): Likewise.
10631 (layout::print_annotation_line): Likewise.
10632 (layout::print_leading_fixits): Likewise.
10633 (layout::annotation_line_showed_range_p): Likewise.
10634 (struct line_corrections): Likewise for field m_row.
10635 (line_corrections::line_corrections): Likewise for param "row".
10636 (layout::print_trailing_fixits): Likewise.
10637 (layout::get_state_at_point): Likewise.
10638 (layout::get_x_bound_for_row): Likewise.
10639 (layout::print_line): Likewise.
10640 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10641 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10642 * input.c (selftest::test_linenum_comparisons): New function.
10643 (selftest::input_c_tests): Call it.
10644 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10645 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10646 * selftest.h (ASSERT_GT): New macro.
10647 (ASSERT_GT_AT): New macro.
10648 (ASSERT_LT): New macro.
10649 (ASSERT_LT_AT): New macro.
10651 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10653 PR rtl-optimization/84780
10654 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10656 2018-03-14 Martin Liska <mliska@suse.cz>
10658 * tree.c (record_node_allocation_statistics): Use
10659 get_stats_node_kind.
10660 (get_stats_node_kind): New function extracted from
10661 record_node_allocation_statistics.
10662 (free_node): Use get_stats_node_kind.
10664 2018-03-14 Richard Biener <rguenther@suse.de>
10666 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10667 that the value-set of ANTIC_IN doesn't grow.
10670 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10672 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10673 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10675 2018-03-14 Julia Koval <julia.koval@intel.com>
10677 * config.gcc (icelake-client, icelake-server): New.
10679 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10680 (initial_ix86_arch_features): Ditto.
10681 (PTA_SKYLAKE): Add SGX.
10682 (PTA_ICELAKE): Remove.
10683 (PTA_ICELAKE_CLIENT): New.
10684 (PTA_ICELAKE_SERVER): New.
10685 (ix86_option_override_internal): Split up icelake on icelake client and
10687 (get_builtin_code_for_version): Ditto.
10688 (fold_builtin_cpu): Ditto.
10689 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10690 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10691 * config/i386/i386.h (processor_type): Ditto.
10692 * doc/invoke.texi: Ditto.
10694 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10697 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10698 INTEGER_CST offset, add it together with bitpos / 8 and
10699 sign extend based on POINTER_SIZE.
10703 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10706 * config/i386/constraints.md (Yc): New register constraint.
10707 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10708 Use Yc constraint for alternative 2 of operand 0. Remove
10709 preferred_for_speed attribute.
10711 2018-03-14 Richard Biener <rguenther@suse.de>
10713 PR tree-optimization/84830
10714 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10715 with the old one to avoid oscillations.
10717 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10720 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10722 (assign_by_spills): Return a flag of reload assignment failure.
10723 Do not process the reload assignment failures. Do not spill other
10724 reload pseudos if they has the same reg class. Update n if
10726 (lra_assign): Add a return arg. Set up from the result of
10727 assign_by_spills call.
10728 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10729 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10730 usage_insns if it is not NULL.
10731 (spill_hard_reg_in_range): New function.
10732 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10733 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10734 function prototypes.
10735 (lra_assign): Change prototype.
10736 * lra.c (lra): Add code to deal with fails by splitting hard reg
10739 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10741 * config/riscv/riscv.opt (mrelax): New option.
10742 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10743 "norelax" when riscv_mrelax is disabled.
10744 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10746 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10749 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10750 reassociation for int modes.
10752 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10754 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10755 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10757 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10758 * config/aarch64/aarch64-sve.md
10759 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10760 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10761 (*extend<mode><Vwide>2): Rename to...
10762 (aarch64_sve_extend<mode><Vwide>2): ...this.
10763 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10764 renaming the old pattern to...
10765 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10767 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10768 define_expand, renaming the old pattern to...
10769 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10770 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10771 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10772 account when deciding which SVE instruction the optab should use.
10773 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10775 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10777 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10778 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10779 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10780 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10781 (tlsdesc_small_<mode>): Turn a define_expand and use
10782 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10783 (tlsdesc_small_advsimd_<mode>): ...this.
10784 (tlsdesc_small_sve_<mode>): New pattern.
10786 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10788 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10789 (UNSPEC_UMUL_HIGHPART): New constants.
10790 (MUL_HIGHPART): New int iteraor.
10791 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10792 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10794 (*<su>mul<mode>3_highpart): New define_insn.
10796 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10799 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10802 2018-03-13 Martin Liska <mliska@suse.cz>
10805 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10807 (sem_item_optimizer::~sem_item_optimizer): Release it.
10808 (sem_item_optimizer::merge_classes): Register variable aliases.
10809 (sem_item_optimizer::fixup_pt_set): New function.
10810 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10811 * ipa-icf.h: Declare new variables and functions.
10813 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10815 PR middle-end/84834
10816 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10817 integer_pow2p@2 and test integer_pow2p in condition.
10818 (A < 0 ? C : 0): Similarly for @1.
10820 PR middle-end/84831
10821 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10822 characters starting at p contain '\0' character, don't look beyond
10826 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10827 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10830 * reg-stack.c (change_stack): Change update_end var from int to
10831 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10832 also call set_block_for_insn on the newly added insns and rescan.
10835 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10836 on the last operand.
10839 * tree.c (stabilize_reference_1): Return save_expr (e) for
10840 STATEMENT_LIST even if it doesn't have side-effects.
10842 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10844 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10846 2018-03-12 Renlin Li <renlin.li@arm.com>
10848 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10849 aarch64_output_scalar_simd_mov_immediate.
10851 2018-03-12 Martin Sebor <msebor@redhat.com>
10853 PR tree-optimization/83456
10854 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10855 for perfectly overlapping calls to memcpy.
10856 (gimple_fold_builtin_memory_chk): Same.
10857 (gimple_fold_builtin_strcpy): Handle no-warning.
10858 (gimple_fold_builtin_stxcpy_chk): Same.
10859 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10861 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10863 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10864 parameter. Use it for SFmode.
10865 (rs6000_function_arg_advance_1): Adjust.
10866 (rs6000_function_arg): Adjust.
10867 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10869 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10871 PR rtl-optimization/84169
10872 PR rtl-optimization/84780
10873 * combine.c (can_combine_p): Check for a 2-insn combination whether
10874 the destination register is used between the two insns, too.
10876 2018-03-12 Richard Biener <rguenther@suse.de>
10878 PR tree-optimization/84803
10879 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10880 for refs DR analysis didn't process.
10882 2018-03-12 Richard Biener <rguenther@suse.de>
10884 PR tree-optimization/84777
10885 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10886 force-vectorize loops ignore whether we are optimizing for size.
10888 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10890 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10891 (TARGET_MD_ASM_ADJUST): Define.
10893 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10894 Kito Cheng <kito.cheng@gmail.com>
10895 Chung-Ju Wu <jasonwucj@gmail.com>
10897 * config/nds32/nds32.c (nds32_compute_stack_frame,
10898 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10899 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10900 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10901 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10902 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10903 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10904 * config/nds32/nds32.md (prologue, epilogue): Use macro
10905 NDS32_V3PUSH_AVAILABLE_P to do checking.
10907 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10910 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10911 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10912 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10913 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10914 addition of most attributes on !orig_type_die or the attribute not
10915 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10917 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10918 Chung-Ju Wu <jasonwucj@gmail.com>
10920 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10921 __NDS32_VH__ macro.
10922 * config/nds32/nds32.opt (mvh): New option.
10924 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10925 Chung-Ju Wu <jasonwucj@gmail.com>
10927 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10929 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10930 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10933 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10934 Chung-Ju Wu <jasonwucj@gmail.com>
10936 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10938 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10939 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10941 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10942 Kito Cheng <kito.cheng@gmail.com>
10943 Chung-Ju Wu <jasonwucj@gmail.com>
10945 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10946 UNSPEC_FFMISM and UNSPEC_FLMISM.
10947 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10948 for ffb, ffmism and flmism.
10949 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10950 (unspec_ffmism): Ditto.
10951 (unspec_flmism): Ditto.
10952 (nds32_expand_builtin_impl): Check if string extension is available.
10953 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10954 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10956 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10959 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10962 * lra-assigns.c (assign_by_spills): Return a flag of reload
10963 assignment failure. Do not process the reload assignment
10964 failures. Do not spill other reload pseudos if they has the same
10966 (lra_assign): Add a return arg. Set up from the result of
10967 assign_by_spills call.
10968 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10969 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10970 usage_insns if it is not NULL.
10971 (spill_hard_reg_in_range): New function.
10972 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10973 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10974 function prototypes.
10975 (lra_assign): Change prototype.
10976 * lra.c (lra): Add code to deal with fails by splitting hard reg
10979 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10982 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10984 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10987 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10988 (dw_val_node): Add val_symbolic_view.
10989 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10990 (symview_upper_bound): New.
10991 (new_line_info_table): Initialize symviews_since_reset.
10992 (dwarf2out_source_line): Count symviews_since_reset and set
10993 symview_upper_bound.
10994 (dw_val_equal_p): Handle symview.
10995 (add_AT_symview): New.
10996 (print_dw_val): Handle symview.
10997 (attr_checksum, attr_checksum_ordered): Likewise.
10998 (same_dw_val_p, size_of_die): Likewise.
10999 (value_format, output_die): Likewise.
11000 (add_high_low_attributes): Use add_AT_symview for entry_view.
11001 (dwarf2out_finish): Reset symview_upper_bound, clear
11004 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11007 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11008 Add strict argument and use it.
11009 (rs6000_split_multireg_move): Update for new strict argument.
11010 (mem_operand_gpr): Disallow all non-offsettable addresses.
11011 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11013 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11016 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11017 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11018 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11021 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11022 decl, use remap_type if we want to use the type.
11024 2018-03-09 Martin Sebor <msebor@redhat.com>
11026 PR tree-optimization/84526
11027 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11029 (builtin_access::generic_overlap): Be prepared to handle non-array
11032 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11034 PR rtl-optimization/84682
11035 * lra-constraints.c (process_address_1): Check is_address flag
11036 for address constraints.
11037 (process_alt_operands): Likewise.
11038 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11039 preprocess_constraints.
11040 * recog.h (preprocess_constraints): Add oploc parameter.
11042 * recog.c (preprocess_constraints): Test address_operand for
11043 CT_ADDRESS constraints.
11045 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11048 * lra-assigns.c (assign_by_spills): Return a flag of reload
11049 assignment failure. Do not process the reload assignment
11050 failures. Do not spill other reload pseudos if they has the same
11052 (lra_assign): Add a return arg. Set up from the result of
11053 assign_by_spills call.
11054 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11055 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11056 usage_insns if it is not NULL.
11057 (spill_hard_reg_in_range): New function.
11058 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11059 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11060 function prototypes.
11061 (lra_assign): Change prototype.
11062 * lra.c (lra): Add code to deal with fails by splitting hard reg
11065 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11068 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11069 Accept complain bool parameter. Only emit errors if it is true.
11070 (arm_parse_cpu_option_name): Likewise.
11071 (arm_target_thumb_only): Adjust callers of the above.
11072 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11073 prototype to take a default true bool parameter.
11074 (arm_parse_arch_option_name): Likewise.
11076 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11077 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11081 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11082 * configure: Regenerate.
11083 * configure.ac ("linker --version-script option"): New.
11084 ("linker soname option"): New.
11086 2018-03-09 Richard Biener <rguenther@suse.de>
11088 PR tree-optimization/84775
11089 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11090 immediate uses of predicate stmts and mark them modified.
11093 PR tree-optimization/84178
11094 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11096 (version_loop_for_if_conversion): Delay update_ssa call.
11097 (tree_if_conversion): Delay update_ssa until after predicate
11100 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11103 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11104 when the function accesses prior frames.
11106 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11109 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11110 gen_llsym, otherwise call maybe_gen_llsym.
11112 PR inline-asm/84742
11113 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11114 has ',' character inside of it.
11116 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11119 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11120 as clobbering CC_REGNUM.
11122 2018-03-08 Richard Biener <rguenther@suse.de>
11124 PR middle-end/84552
11125 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11126 (follow_copies_to_constant): Do not follow SSA names registered
11129 2018-03-08 Richard Biener <rguenther@suse.de>
11131 PR tree-optimization/84178
11132 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11134 (version_loop_for_if_conversion): Delay update_ssa call.
11135 (tree_if_conversion): Delay update_ssa until after predicate
11138 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11140 PR tree-optimization/84178
11141 * tree-if-conv.c (release_bb_predicate): Remove the
11142 the assertion that the stmts have NULL use_ops.
11143 Discard the statements, asserting that they haven't
11144 yet been added to a BB.
11146 2018-03-08 Richard Biener <rguenther@suse.de>
11148 PR tree-optimization/84746
11149 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11150 (phi_translate): Pass in destination ANTIC_OUT set.
11151 (phi_translate_1): Likewise. For a simplified result lookup
11152 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11153 (phi_translate_set): Adjust.
11154 (do_pre_regular_insertion): Likewise.
11155 (do_pre_partial_partial_insertion): Likewise.
11157 2018-03-08 Martin Liska <mliska@suse.cz>
11159 PR gcov-profile/84735
11160 * doc/gcov.texi: Document usage of profile files.
11161 * gcov-io.h: Document changes in the format.
11163 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11167 * dwarf2out.c (struct dw_line_info_table): Update comments for
11169 (FORCE_RESET_NEXT_VIEW): New.
11170 (FORCE_RESETTING_VIEW_P): New.
11171 (RESETTING_VIEW_P): Check for -1 too.
11172 (ZERO_VIEW_P): Likewise.
11173 (new_line_info_table): Force-reset next view.
11174 (dwarf2out_begin_function): Likewise.
11175 (dwarf2out_source_line): Simplify zero_view_p initialization.
11176 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11177 view directly. Omit view when omitting .loc at line 0.
11179 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11181 PR tree-optimization/84740
11182 * tree-switch-conversion.c (process_switch): Call build_constructors
11183 only if info.phi_count is non-zero.
11185 PR tree-optimization/84739
11186 * tree-tailcall.c (find_tail_calls): Check call arguments against
11187 DECL_ARGUMENTS (current_function_decl) rather than
11188 DECL_ARGUMENTS (func) when checking for tail recursion.
11190 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11192 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11193 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11194 Volker Reichelt's entry and add entries for people that perform
11195 GCC fuzzy testing and report numerous bugs.
11197 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11200 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11201 readonly data in sdata, if that is disabled.
11202 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11203 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11204 -mreadonly-in-sdata option.
11206 2018-03-07 Martin Sebor <msebor@redhat.com>
11208 PR tree-optimization/84468
11209 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11210 basic block when looking for nul assignment.
11212 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11215 * except.h (output_function_exception_table): Adjust prototype.
11216 * except.c (output_function_exception_table): Remove FNNAME parameter
11217 and add SECTION parameter. Ouput one part of the table at a time.
11218 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11219 the first part of the exception table and emit unwind directives.
11220 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11221 (i386_pe_seh_cold_init): Likewise.
11222 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11223 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11224 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11225 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11226 * config/i386/winnt.c: Include except.h.
11227 (struct seh_frame_state): Add reg_offset, after_prologue and
11228 in_cold_section fields.
11229 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11230 (i386_pe_seh_cold_init): New function.
11231 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11232 to seh->in_cold_section.
11233 (seh_emit_push): Record the offset of the push.
11234 (seh_emit_save): Record the offet of the save.
11235 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11236 Test seh->after_prologue to disregard the epilogue.
11237 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11238 (i386_pe_end_cold_function): New function.
11240 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11243 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11244 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11247 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11249 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11250 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11252 PR middle-end/84723
11253 * multiple_target.c: Include tree-inline.h and intl.h.
11254 (expand_target_clones): Diagnose and fail if node->definition and
11255 !tree_versionable_function_p (node->decl).
11257 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11259 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11261 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11262 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11263 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11265 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11268 * combine.c (try_combine): Use reg_or_subregno instead of handling
11269 just paradoxical SUBREGs and REGs.
11271 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11273 * config/arc/arc.c (arc_finalize_pic): Remove function.
11274 (arc_must_save_register): We use single base PIC register, remove
11275 checks to save/restore the PIC register.
11276 (arc_expand_prologue): Likewise.
11277 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11279 (arc_verify_short): Likewise.
11280 (arc_attr_type): Likewise.
11281 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11282 (walk_stores): Likewise.
11283 (arc_address_cost): Make it static.
11284 (arc_verify_short): Likewise.
11285 (branch_dest): Likewise.
11286 (arc_attr_type): Likewise.
11287 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11288 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11289 (arc_final_prescan_insn): Remove inserting the nops due to
11290 hardware hazards. It is done in reorg step.
11291 (insn_length_variant_t): Remove.
11292 (insn_length_parameters_t): Likewise.
11293 (arc_insn_length_parameters): Likewise.
11294 (arc_get_insn_variants): Likewise.
11295 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11297 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11299 PR inline-asm/84683
11300 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11303 PR tree-optimization/84687
11304 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11306 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11308 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11310 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11311 Rename to ppc_speculation_barrier.
11312 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11313 __builtin_ppc_speculation_barrier.
11315 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11318 * combine.c (combine_simplify_rtx): Don't try to simplify if
11319 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11322 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11324 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11325 to 32 bytes when compiling for POWER9.
11327 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11330 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11331 regparm >= 3 with no arg reg available also for calls with
11332 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11335 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11337 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11339 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11342 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11344 2018-03-05 Richard Biener <rguenther@suse.de>
11346 PR tree-optimization/84486
11347 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11348 When inserting a __builtin_assume_aligned call set the LHS
11349 SSA name alignment info accordingly.
11351 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11353 PR tree-optimization/84114
11354 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11355 Avoid reassociation of FLOAT_MODE addition.
11357 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11359 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11360 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11361 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11362 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11363 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11364 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11365 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11367 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11368 __builtin_ia32_wbinvd): New builtins.
11369 (SPECIAL_ARGS2): New.
11370 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11371 (SPECIAL_ARGS2): New.
11372 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11373 (ix86_valid_target_attribute_inner_p): Ditto.
11374 (ix86_init_mmx_sse_builtins): Add special_args2.
11375 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11376 TARGET_WBNOINVD_P): New.
11377 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11378 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11379 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11380 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11381 * config/i386/pconfigintrin.h: New file.
11382 * config/i386/wbnoinvdintrin.h: Ditto.
11383 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11385 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11387 2018-03-05 Richard Biener <rguenther@suse.de>
11389 PR tree-optimization/84670
11390 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11392 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11393 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11394 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11395 doesn't grow if all successors (recursively) were visited at least
11398 2018-03-05 Richard Biener <rguenther@suse.de>
11400 PR tree-optimization/84650
11401 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11402 if executed in the loop pipeline.
11404 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11406 * doc/configfiles.texi (Configuration Files): Move info about
11407 conditionalizing $target-protos.h to...
11408 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11409 differs from $target-protos.h.
11411 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11412 Chung-Ju Wu <jasonwucj@gmail.com>
11414 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11415 * config/nds32/nds32-multiple.md (setmemsi): Define.
11416 * config/nds32/nds32-memory-manipulation.c
11417 (nds32_gen_dup_4_byte_to_word_value): New.
11418 (emit_setmem_word_loop): New.
11419 (emit_setmem_byte_loop): New.
11420 (nds32_expand_setmem_loop): New.
11421 (nds32_expand_setmem_loop_v3m): New.
11422 (nds32_expand_setmem_unroll): New.
11423 (nds32_expand_setmem): New.
11425 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11426 Chung-Ju Wu <jasonwucj@gmail.com>
11428 * config/nds32/nds32-memory-manipulation.c
11429 (nds32_emit_load_store): New.
11430 (nds32_emit_post_inc_load_store): New.
11431 (nds32_emit_mem_move): New.
11432 (nds32_emit_mem_move_block): New.
11433 (nds32_expand_movmemsi_loop_unknown_size): New.
11434 (nds32_expand_movmemsi_loop_known_size): New.
11435 (nds32_expand_movmemsi_loop): New.
11436 (nds32_expand_movmemsi_unroll): New.
11437 (nds32_expand_movmemqi): Rename ...
11438 (nds32_expand_movmemsi): ... to this.
11439 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11440 (movmemsi): ... to this.
11441 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11442 (nds32_expand_movmemsi): ... to this.
11444 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11445 Monk Chiang <sh.chiang04@gmail.com>
11446 Chung-Ju Wu <jasonwucj@gmail.com>
11448 * config/nds32/nds32-protos.h
11449 (nds32_expand_load_multiple): New arguments.
11450 (nds32_expand_store_multiple): Ditto.
11451 (nds32_valid_multiple_load_store): Rename ...
11452 (nds32_valid_multiple_load_store_p): ... to this.
11453 * config/nds32/nds32-memory-manipulation.c
11454 (nds32_expand_load_multiple): Refine implementation.
11455 (nds32_expand_store_multiple): Ditto.
11456 * config/nds32/nds32-multiple.md
11457 (load_multiple): Update nds32_expand_load_multiple interface.
11458 (store_multiple): Update nds32_expand_store_multiple interface.
11459 * config/nds32/nds32-predicates.c
11460 (nds32_valid_multiple_load_store): Rename ...
11461 (nds32_valid_multiple_load_store_p): ... to this and refine
11463 * config/nds32/predicates.md
11464 (nds32_load_multiple_and_update_address_operation): New predicate.
11465 (nds32_store_multiple_and_update_address_operation): New predicate.
11467 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11468 Chung-Ju Wu <jasonwucj@gmail.com>
11470 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11471 (combo): New attribute.
11472 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11474 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11476 * config/nds32/nds32.opt: Change -mcmodel= default value.
11478 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11479 Monk Chiang <sh.chiang04@gmail.com>
11480 Chung-Ju Wu <jasonwucj@gmail.com>
11482 * config/nds32/constants.md (unspec_element): New enum.
11483 * config/nds32/constraints.md (Umw): New constraint.
11484 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11485 * config/nds32/nds32-intrinsic.md: Likewise.
11486 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11487 (nds32_valid_smw_lwm_base_p): New.
11488 (nds32_output_smw_single_word): New.
11489 (nds32_output_lmw_single_word): New.
11490 (nds32_expand_unaligned_load): New.
11491 (nds32_expand_unaligned_store): New.
11492 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11493 (nds32_output_smw_single_word): Declare.
11494 (nds32_output_lmw_single_word): Declare.
11495 (nds32_expand_unaligned_load): Declare.
11496 (nds32_expand_unaligned_store): Declare.
11497 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11498 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11499 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11500 NDS32_BUILTIN_UASTORE_DW.
11501 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11504 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11505 Kito Cheng <kito.cheng@gmail.com>
11506 Chung-Ju Wu <jasonwucj@gmail.com>
11508 * config/nds32/nds32-intrinsic.c
11509 (nds32_expand_builtin_null_ftype_reg): Delete.
11510 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11511 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11512 (nds32_read_argument): New.
11513 (nds32_legitimize_target): Ditto.
11514 (nds32_legitimize_argument): Ditto.
11515 (nds32_check_constant_argument): Ditto.
11516 (nds32_expand_unop_builtin): Ditto.
11517 (nds32_expand_unopimm_builtin): Ditto.
11518 (nds32_expand_binop_builtin): Ditto.
11519 (nds32_builtin_decl_impl): Ditto.
11520 (builtin_description): Ditto.
11521 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11522 (nds32_init_builtins_impl): Ditto.
11523 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11524 (nds32_builtin_decl): New.
11525 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11526 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11528 2018-03-02 Jeff Law <law@redhat.com>
11530 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11531 (redundant_insn, fill_simple_delay_slots): Likewise.
11532 (fill_slots_from_thread): Likewise.
11533 * resource.c (mark_referenced_resources): Likewise.
11534 (mark_set_resources, find_dead_or_set_registers): Likewise.
11536 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11538 * substring-locations.h (format_warning_va): Formatting fix for
11539 ATTRIBUTE_GCC_DIAG.
11540 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11542 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11543 * substring-locations.c: Include intl.h.
11544 (format_warning_va): Turned into small wrapper around
11545 format_warning_n_va, renamed to ...
11546 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11547 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11549 (format_warning_at_substring_n): New function.
11550 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11551 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11552 format_warning_at_substring with just a shorter name instead of
11553 const function pointer.
11554 (fmtwarn_n): New function.
11555 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11556 appropriate, get rid of all the fmtstr temporaries, move conditionals
11557 with G_() wrapped string literals directly into fmtwarn arguments,
11558 cast dir.len to (int), formatting fixes.
11560 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11562 * doc/invoke.texi: Remove "Cilk Plus" references.
11564 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11565 Richard Biener <rguenther@suse.de>
11568 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11569 for error or warning attributes if CALL_FROM_THUNK_P is set.
11572 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11575 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11576 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11579 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11580 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11582 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11583 instead of -1U in last predictors element's probability member.
11585 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11588 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11589 arguments if they are comparable.
11591 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11593 PR tree-optimization/84634
11594 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11595 masks and masked_loop_p with a single loop_masks, making sure it's
11596 null for bb vectorization.
11598 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11600 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11601 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11602 loop->force_vectorize to check whether there is no alias.
11604 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11607 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11609 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11611 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11612 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11613 instead of a loop around prev_real_insn.
11614 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11617 PR inline-asm/84625
11618 * config/i386/i386.c (ix86_print_operand): Use conditional
11619 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11622 2018-03-02 Richard Biener <rguenther@suse.de>
11624 PR tree-optimization/84427
11625 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11626 (bitmap_set_subtract_values): Rewrite to handle multiple
11629 (prune_clobbered_mems): Likewise.
11630 (phi_translate): Take edge instead of pred/phiblock.
11631 (phi_translate_1): Likewise.
11632 (phi_translate_set): Likewise. Insert all translated
11633 exprs for a value into the set, keeping possibly multiple
11634 expressions per value.
11635 (compute_antic_aux): Adjust for phi_translate changes.
11636 When intersecting union the expressions and prune those
11637 not in the final value set, keeping possibly multiple
11638 expressions per value. Do not use value-insertion
11639 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11640 all expressions. Add verification that the value-sets
11641 only shrink during iteration.
11642 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11643 (do_pre_regular_insertion): Likewise.
11644 (do_pre_partial_partial_insertion): Likewise.
11646 2018-03-02 Richard Biener <rguenther@suse.de>
11649 * config/darwin.c (saved_debug_info_level): New static global.
11650 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11651 (darwin_asm_lto_end): Restore debug info generation settings.
11653 2018-03-01 Martin Liska <mliska@suse.cz>
11656 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11657 volatile arguments.
11659 2018-03-01 Richard Biener <rguenther@suse.de>
11662 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11663 type annotation with LTO when debug was disabled at compile-time.
11665 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11667 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11669 (mips_final_postscan_insn): Likewise.
11671 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11673 PR rtl-optimization/84528
11674 * alias.c (init_alias_target): Add commentary.
11675 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11676 a unique base value if the frame pointer is not eliminated
11677 to the stack pointer.
11679 2018-03-01 Tom de Vries <tom@codesourcery.com>
11681 PR rtl-optimization/83327
11682 * lra-int.h (hard_regs_spilled_into): Declare.
11683 * lra.c (hard_regs_spilled_into): Define.
11684 (init_reg_info): Init hard_regs_spilled_into.
11685 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11686 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11687 (process_bb_lives): Handle hard_regs_spilled_into.
11688 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11689 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11691 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11693 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11694 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11695 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11696 * config/rs6000/aix72.h: New file.
11698 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11700 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11701 instead of warning_at with conditional singular and plural messages
11705 * stor-layout.c (update_alignment_for_field): For
11706 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11707 && !DECL_PACKED (field), do the alignment update, just use
11708 only desired_align instead of MAX (type_align, desired_align)
11710 (place_field): Don't do known_align < desired_align handling
11711 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11712 is non-NULL, instead do it after rli->prev_field handling and
11713 only if not within a bitfield word. For DECL_PACKED (field)
11714 use type_align of BITS_PER_UNIT.
11716 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11718 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11719 superfluous parentheses and trailing spaces.
11721 2018-02-28 Richard Biener <rguenther@suse.de>
11723 PR tree-optimization/84584
11724 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11725 SCoPs with fake exit edge.
11727 2018-02-28 Martin Liska <mliska@suse.cz>
11730 * timevar.c (timer::print): Fix format to properly print 100%
11733 2018-02-28 Richard Biener <rguenther@suse.de>
11735 PR middle-end/84607
11736 * genmatch.c (capture_info::walk_match): Do not mark
11737 captured expressions without operands as expr_p given
11738 they act more like predicates and should be subject to
11739 "lost tail" side-effect preserving.
11741 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11743 PR rtl-optimization/81611
11744 * auto-inc-dec.c (attempt_change): Move dead note from
11745 mem_insn if it's the next use of regno
11746 (find_address): Take address use of reg holding
11747 non-incremented value. Add parm to limit search to the named
11749 (merge_in_block): Attempt to use a mem insn that is the next
11750 use of the original regno.
11752 2018-02-27 Martin Sebor <msebor@redhat.com>
11755 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11756 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11758 2018-02-27 Martin Sebor <msebor@redhat.com>
11760 PR translation/84207
11761 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11762 n argument to unsigned HOST_WIDE_INT.
11763 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11764 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11765 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11766 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11768 2018-02-27 Richard Biener <rguenther@suse.de>
11770 PR tree-optimization/84512
11771 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11772 Do not use the estimate returned from record_stmt_cost for
11773 the scalar iteration cost but sum properly using add_stmt_cost.
11775 2018-02-27 Richard Biener <rguenther@suse.de>
11777 PR tree-optimization/84466
11778 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11779 Adjust last change to less strictly validate use operands.
11781 2018-02-27 Martin Liska <mliska@suse.cz>
11783 PR gcov-profile/84548
11784 * gcov.c (process_file): Allow partial overlap and consider it
11785 also as group functions.
11786 (output_lines): Properly calculate range of lines for a group.
11788 2018-02-27 Martin Liska <mliska@suse.cz>
11790 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11791 'ggc' suffixes. Change first column width.
11792 (timer::print): Fix formatting of the column.
11794 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11796 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11797 preserve inline entry blocks for the sake of debug inline
11798 entry point markers alone.
11799 (remove_unused_locals): Suggest in comments a better place to
11800 force the preservation of inline entry blocks that are
11801 otherwise unused, but do not preserve them.
11803 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11805 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11807 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11810 * config/i386/constraints.md (Bs): Replace
11811 ix86_indirect_branch_register with
11812 TARGET_INDIRECT_BRANCH_REGISTER.
11814 * config/i386/i386.md (indirect_jump): Likewise.
11815 (tablejump): Likewise.
11816 (*sibcall_memory): Likewise.
11817 (*sibcall_value_memory): Likewise.
11818 Peepholes of indirect call and jump via memory: Likewise.
11819 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11820 (*sibcall_value_GOT_32): Likewise.
11821 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11822 (GOT_memory_operand): Likewise.
11823 (call_insn_operand): Likewise.
11824 (sibcall_insn_operand): Likewise.
11825 (GOT32_symbol_operand): Likewise.
11826 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11828 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11830 PR rtl-optimization/83496
11831 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11832 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11833 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11834 redundant insn, if any.
11835 (relax_delay_slots): Likewise.
11836 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11838 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11840 PR tree-optimization/83965
11841 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11842 that grouped statements are part of a reduction chain. Return
11843 true if the statement is not marked as a reduction itself but
11844 is part of a group.
11845 (vect_recog_dot_prod_pattern): Don't check whether the statement
11846 is part of a group here.
11847 (vect_recog_sad_pattern): Likewise.
11848 (vect_recog_widen_sum_pattern): Likewise.
11850 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11853 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11855 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11858 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11860 (ix86_output_indirect_function_return): New prototype.
11861 (ix86_split_simple_return_pop_internal): Likewise.
11862 * config/i386/i386.c (indirect_return_via_cx): New.
11863 (indirect_return_via_cx_bnd): Likewise.
11864 (indirect_thunk_name): Handle return va CX_REG.
11865 (output_indirect_thunk_function): Create alias for
11866 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11867 (ix86_output_indirect_jmp): Remove the bool argument.
11868 (ix86_output_indirect_function_return): New function.
11869 (ix86_split_simple_return_pop_internal): Likewise.
11870 * config/i386/i386.md (*indirect_jump): Don't pass false
11871 to ix86_output_indirect_jmp.
11872 (*tablejump_1): Likewise.
11873 (simple_return_pop_internal): Change it to define_insn_and_split.
11874 Call ix86_split_simple_return_pop_internal to split it for
11875 -mfunction-return=.
11876 (simple_return_indirect_internal): Call
11877 ix86_output_indirect_function_return instead of
11878 ix86_output_indirect_jmp.
11880 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11883 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11884 memset and value initialization afterwards.
11886 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11888 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11890 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11893 * common/config/aarch64/aarch64-common.c
11894 (aarch_option_optimization_table[]): Switch
11895 off fomit-frame-pointer
11897 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11898 Chung-Ju Wu <jasonwucj@gmail.com>
11900 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11902 (store_multiple): Ditto.
11904 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11906 * config.gcc: Add --with-cpu support for nds32 target.
11907 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11908 * config/nds32/nds32.opt: Add -mcpu= option.
11910 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11912 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11913 isel=yes): Warn for these deprecated options.
11915 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11917 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11918 ISA_2_5_MASKS_EMBEDDED.
11920 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11922 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11923 p->max as pointers rather than using iterative_hash_expr.
11925 2018-02-23 Carl Love <cel@us.ibm.com>
11927 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11928 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11930 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11931 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11932 P8V_BUILTIN_VEC_VUNSIGNED2.
11934 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11937 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11938 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11940 * lra-constraints.c (curr_insn_transform): Set up
11941 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11943 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11944 * lra-eliminations.c (spill_pseudos): Ditto.
11945 (process_insn_for_elimination): Ditto.
11946 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11947 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11948 LRA_NON_CLOBBERED_ALT.
11950 2018-02-22 Martin Sebor <msebor@redhat.com>
11952 PR tree-optimization/84480
11953 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11954 to maybe_diag_stxncpy_trunc. Call it.
11955 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11956 from gimple_fold_builtin_strcpy. Print inlining stack.
11957 (handle_builtin_stxncpy): Print inlining stack.
11958 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11960 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11963 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11964 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11965 and -fcheck-pointer-bounds are used together.
11966 (indirect_thunk_prefix): New enum.
11967 (indirect_thunk_need_prefix): New function.
11968 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11969 "_nt" instead of "_bnd" for NOTRACK prefix.
11970 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11971 (output_indirect_thunk_function): Likewise.
11973 (ix86_code_end): Update output_indirect_thunk_function calls.
11974 (ix86_output_indirect_branch_via_reg): Replace
11975 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11976 (ix86_output_indirect_branch_via_push): Likewise.
11977 (ix86_output_function_return): Likewise.
11978 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11979 incompatible with -fcf-protection=branch and
11980 -fcheck-pointer-bounds.
11982 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11985 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11986 Change gcc_assert call to output_operand_lossage.
11988 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11990 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11992 2018-02-22 DJ Delorie <dj@redhat.com>
11993 Sebastian Perta <sebastian.perta@renesas.com>
11994 Oleg Endo <olegendo@gcc.gnu.org>
11996 * config/rx/rx.c (rx_rtx_costs): New function.
11997 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11999 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12001 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12003 2018-02-22 Martin Liska <mliska@suse.cz>
12006 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12007 Add "native" as a possible value.
12009 2018-02-22 Martin Liska <mliska@suse.cz>
12012 * config/i386/i386.c (ix86_option_override_internal):
12013 Add "native" as a possible value for -march and -mtune.
12015 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12018 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12019 to all type variants.
12021 PR tree-optimization/84503
12022 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12023 width as info->bitpos + info->bitsize - start.
12024 (merged_store_group::merge_overlapping): Simplify width computation.
12025 (check_no_overlap): New function.
12026 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12027 start + width and last_order of the group, fail if check_no_overlap
12029 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12030 to group if check_no_overlap fails.
12032 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12034 * config/rs6000/altivec.md: Delete contraint arguments to
12035 define_expand, define_split, and define_peephole2, and in
12036 define_insn_and_split if always unused.
12037 * config/rs6000/darwin.md: Ditto.
12038 * config/rs6000/dfp.md: Ditto.
12039 * config/rs6000/rs6000.md: Ditto.
12040 * config/rs6000/sync.md: Ditto.
12041 * config/rs6000/vector.md: Ditto.
12042 * config/rs6000/vsx.md: Ditto.
12044 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12046 * config/rs6000/altivec.md: Write output control strings as braced
12047 blocks instead of double-quoted strings.
12048 * config/rs6000/darwin.md: Ditto.
12049 * config/rs6000/rs6000.md: Ditto.
12050 * config/rs6000/vector.md: Ditto.
12051 * config/rs6000/vsx.md: Ditto.
12053 2018-02-21 Jason Merrill <jason@redhat.com>
12055 PR c++/84314 - ICE with templates and fastcall attribute.
12056 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12058 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12060 * ipa-cp.c (determine_versionability): Fix comment typos.
12062 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12065 * ipa-cp.c (determine_versionability): Do not version functions caling
12068 2018-02-21 Martin Liska <mliska@suse.cz>
12071 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12072 Add "native" as a possible value.
12073 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12074 the macro when native cpu detection is available.
12076 2018-02-21 Martin Liska <mliska@suse.cz>
12079 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12080 Add "native" as a possible value.
12081 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12082 when native cpu detection is available.
12084 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12085 Martin Sebor <msebor@redhat.com>
12087 PR tree-optimization/84478
12088 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12090 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12091 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12092 support which is conservatively correct, for 2 only stay conservative
12093 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12094 argument to the 2 argument get_range_strlen, adjust 6 arg
12095 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12097 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12098 (gimple_fold_builtin_strlen): Pass true as last argument to
12101 2018-02-20 Martin Sebor <msebor@redhat.com>
12103 PR middle-end/84095
12104 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12105 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12106 (builtin_memref::builtin_memref): Factor out parts into
12107 set_base_and_offset and call it.
12109 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12111 PR middle-end/84406
12112 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12113 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12114 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12115 search at the associated MODE_INT.
12117 2018-02-20 Jeff Law <law@redhat.com>
12119 PR middle-end/82123
12120 PR tree-optimization/81592
12121 PR middle-end/79257
12122 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12123 for range data rather than using global data.
12124 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12125 range data rather than using global data.
12126 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12127 pass it to children as needed.
12128 (struct directive::fmtresult): Similarly.
12129 (struct directive::set_width): Similarly.
12130 (struct directive::set_precision): Similarly.
12131 (format_integer, format_directive, parse_directive): Similarly.
12132 (format_none): Accept unnamed vr_values parameter.
12133 (format_percent, format_floating, format_character): Similarly.
12134 (format_string, format_plain): Similarly.
12135 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12136 the EVRP range analyzer for range data rather than using global data.
12137 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12138 gimple-ssa-evrp-analyze.h
12139 (class sprintf_dom_walker): Add after_dom_children member function.
12140 Add evrp_range_analyzer member.
12141 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12142 range analyzer as needed.
12143 (sprintf_dom_walker::after_dom_children): New member function.
12144 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12146 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12147 (evrp_range_analyzer::pop_to_marker): Likewise.
12149 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12151 PR tree-optimization/84419
12152 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12153 with the required type if its current type is compatible but
12156 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12158 PR middle-end/82004
12159 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12160 after vectorization.
12162 2018-02-20 Martin Liska <mliska@suse.cz>
12165 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12166 possible values if we don't have a hint.
12168 2018-02-20 Martin Liska <mliska@suse.cz>
12172 * final.c (shorten_branches): Build align_tab array with one
12174 * opts.c (finish_options): Add alignment option limit check.
12175 (MAX_CODE_ALIGN): Likewise.
12176 (MAX_CODE_ALIGN_VALUE): Likewise.
12177 * doc/invoke.texi: Document maximum allowed option value for
12178 all -falign-* options.
12180 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12183 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12184 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12185 * var-tracking.c (emit_note_insn_var_location): Remove all references
12186 to NOTE_INSN_CALL_ARG_LOCATION.
12187 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12188 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12189 Use copy_rtx_if_shared.
12190 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12191 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12192 (dwarf2out_var_location): Remove handling of
12193 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12195 * final.c (final_scan_insn): Remove all references to
12196 NOTE_INSN_CALL_ARG_LOCATION.
12197 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12198 before dumping final insns.
12199 * except.c (emit_note_eh_region_end): Remove all references to
12200 NOTE_INSN_CALL_ARG_LOCATION.
12201 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12202 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12203 * config/arc/arc.c (hwloop_optimize): Likewise.
12204 * config/arm/arm.c (create_fix_barrier): Likewise.
12205 * config/s390/s390.c (s390_chunkify_start): Likewise.
12206 * config/sh/sh.c (find_barrier): Likewise.
12207 * config/i386/i386.c (rest_of_insert_endbranch,
12208 ix86_seh_fixup_eh_fallthru): Likewise.
12209 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12210 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12211 * config/frv/frv.c (frv_function_prologue): Likewise.
12212 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12214 (note_outside_basic_block_p): Remove all references to
12215 NOTE_INSN_CALL_ARG_LOCATION.
12216 * gengtype.c (adjust_field_rtx_def): Likewise.
12217 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12219 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12220 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12223 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12226 PR tree-optimization/84452
12227 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12228 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12231 2018-02-19 Martin Liska <mliska@suse.cz>
12234 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12236 2018-02-19 Martin Liska <mliska@suse.cz>
12237 Richard Sandiford <richard.sandiford@linaro.org>
12239 PR tree-optimization/82491
12240 * gimple-fold.c (get_base_constructor): Make earlier bail out
12243 2018-02-19 Carl Love <cel@us.ibm.com>
12245 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12246 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12248 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12249 P8V_BUILTIN_VEC_NEG.
12251 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12253 * config/rl78/rl78.md (movdf): New define expand.
12255 2018-02-19 Martin Liska <mliska@suse.cz>
12258 * doc/invoke.texi: Fix typo.
12259 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12261 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12263 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12264 handle rs6000_single_float and rs6000_double_float specially for
12267 2018-02-16 Jeff Law <law@redhat.com>
12269 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12270 (pop_regs): Use it.
12272 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12275 * ipa-inline.c (inline_small_functions): Fix a typo.
12277 2018-02-16 Nathan Sidwell <nathan@acm.org>
12279 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12281 2018-02-16 Carl Love <cel@us.ibm.com>
12283 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12284 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12285 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12286 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12287 expansion to P8V_BUILTIN_VEC_FLOAT2.
12289 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12291 PR rtl-optimization/70023
12292 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12293 src_regno into account.
12295 2018-02-16 Carl Love <cel@us.ibm.com>
12297 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12298 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12299 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12300 * config/rs6000/rs6000.c: Remove case statements for
12301 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12302 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12303 and P9V_BUILTIN_VEC_VINSERT4B.
12304 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12305 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12306 * config/rs6000/vsx.md:
12307 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12310 2018-02-16 Carl Love <cel@us.ibm.com>
12312 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12314 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12316 * config/rs6000/rs6000-c.c: Add the definitions for
12317 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12318 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12319 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12320 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12321 definition for insert4b and define insn *insert3b_internal.
12322 * doc/extend.texi: Add documentation for vec_extract4b.
12324 2018-02-16 Nathan Sidwell <nathan@acm.org>
12326 * doc/extend.texi (Backwards Compatibility): Mention friend
12327 injection. Note for-scope is deprecated.
12328 * doc/invoke.texi (-ffriend-injection): Deprecate.
12330 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12332 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12333 that moved to I2, also allow destinations that are a paradoxical
12334 subreg (instead of a normal reg).
12336 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12339 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12342 2018-02-16 Richard Biener <rguenther@suse.de>
12344 PR tree-optimization/84037
12345 PR tree-optimization/84016
12347 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12348 Adjust vec_construct for the fact we need additional higher latency
12349 128bit inserts for AVX256 and AVX512 vector builds.
12350 (ix86_add_stmt_cost): Scale vector construction cost for
12353 2018-02-16 Richard Biener <rguenther@suse.de>
12355 PR tree-optimization/84417
12356 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12357 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12358 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12360 2018-02-16 Martin Liska <mliska@suse.cz>
12363 * internal-fn.def (ASAN_CHECK): Set proper flags.
12364 (ASAN_MARK): Likewise.
12366 2018-02-16 Julia Koval <julia.koval@intel.com>
12368 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12369 from PTA_CANNONLAKE.
12371 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12374 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12375 Use ++iter rather than iter++ for std::list iterators.
12376 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12377 defer deleting them until all nodes in the forest are processed. Do
12378 free even leaf nodes. Change to_process into auto_vec.
12381 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12382 * vec.h (vec_default_construct): Use memset instead of placement new
12383 if BROKEN_VALUE_INITIALIZATION is defined.
12384 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12385 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12388 PR rtl-optimization/83723
12389 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12390 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12391 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12393 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12395 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12397 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12399 PR rtl-optimization/81443
12400 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12401 from inner REGs to paradoxical SUBREGs.
12403 2018-02-16 Richard Biener <rguenther@suse.de>
12405 PR tree-optimization/84399
12406 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12407 For operands we can analyze at their definition make sure we can
12408 analyze them at each use as well.
12410 2018-02-16 Richard Biener <rguenther@suse.de>
12412 PR tree-optimization/84190
12413 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12414 volatile accesses if the decl isn't volatile.
12416 2018-02-15 Jason Merrill <jason@redhat.com>
12418 PR c++/84314 - ICE with templates and fastcall attribute.
12419 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12420 TYPE_CANONICAL on an existing type.
12422 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12424 PR tree-optimization/84383
12425 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12426 dstoff nor call operand_equal_p if dstbase is NULL.
12428 PR tree-optimization/84334
12429 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12430 also a CONSTANT_CLASS_P, punt.
12432 2018-02-14 Jim Wilson <jimw@sifive.com>
12434 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12435 first SMALL_OPERAND check. New local min_second_step. Move assert
12436 to where locals are set. Add TARGET_RVC support.
12437 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12439 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12441 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12443 2018-02-14 Martin Sebor <msebor@redhat.com>
12445 PR tree-optimization/83698
12446 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12447 arrays constrain the offset range to their bounds.
12448 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12449 (builtin_access::overlap): Avoid setting the size of overlap if it's
12451 (maybe_diag_overlap): Also consider arrays when deciding what values
12452 of offsets to include in diagnostics.
12454 2018-02-14 Martin Sebor <msebor@redhat.com>
12457 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12458 that correspond to the kind of a declaration.
12460 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12463 * config/pa/pa.md: Load address of PIC label using the linkage table
12464 if the label is nonlocal.
12466 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12468 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12469 warning message if user requests -maltivec=be.
12470 * doc/invoke.texi: Document deprecation of -maltivec=be.
12472 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12475 * config/rs6000/rs6000-c.c: Update definitions for
12476 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12477 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12479 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12482 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12483 add _get_ssp intrinsics. Remove argument from
12484 __builtin_ia32_rdssp[d|q].
12485 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12486 * config/i386/i386-builtin.def: Remove argument from
12487 __builtin_ia32_rdssp[d|q].
12488 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12489 ix86_expand_special_args_builtin for _rdssp[d|q].
12490 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12491 Clear register before usage.
12492 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12493 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12495 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12497 PR tree-optimization/84357
12498 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12499 operand 1 of an ARRAY_REF too.
12501 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12504 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12505 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12507 (set_of_reg): New struct.
12508 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12509 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12510 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12512 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12513 Split into bitclr, bitset, bitinvert patterns if appropriate.
12514 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12515 use rx_fuse_in_memory_bitop.
12516 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12517 to named insn, correct maximum insn length.
12519 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12522 * machmode.def: Define a complex mode for PARTIAL_INT.
12523 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12525 * doc/rtl.texi: Document CSPImode.
12526 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12528 (msp430_hard_regno_nregs_with_padding): Likewise.
12530 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12533 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12535 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12537 PR rtl-optimization/84169
12538 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12539 we generated a parallel as new i3 and we split that to new i2 and i3
12540 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12541 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12542 those to i2, not i1. Partially rewrite this scan code.
12544 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12547 * stor-layout.c (place_field): For variable length fields, adjust
12548 offset_align afterwards not just based on the field's alignment,
12549 but also on the size.
12551 PR middle-end/84309
12552 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12553 of exps and logs in the use_exp2 case.
12555 2018-02-13 Jeff Law <law@redhat.com>
12557 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12558 entry for "vector".
12560 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12563 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12567 * common.opt (gas-loc-support, gas-locview-support): New.
12568 (ginline-points, ginternal-reset-location-views): New.
12569 * doc/invoke.texi: Document them. Use @itemx where intended.
12570 (gvariable-location-views): Adjust.
12571 * target.def (reset_location_view): New.
12572 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12573 (TARGET_RESET_LOCATION_VIEW): New.
12574 * doc/tm.texi: Rebuilt.
12575 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12576 (dwarf2out_default_as_locview_support): New.
12577 (output_asm_line_debug_info): Use option variables.
12578 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12579 (output_loc_list): Likewise.
12580 (add_high_low_attributes): Check option variables.
12581 Don't output entry view attribute in strict mode.
12582 (gen_inlined_subroutine_die): Check option variables.
12583 (dwarf2out_inline_entry): Likewise.
12584 (init_sections_and_labels): Likewise.
12585 (dwarf2out_early_finish): Likewise.
12586 (maybe_reset_location_view): New, from...
12587 (dwarf2out_var_location): ... here. Call it.
12588 * debug.h (dwarf2out_default_as_loc_support): Declare.
12589 (dwarf2out_default_as_locview_support): Declare.
12590 * hooks.c (hook_int_rtx_insn_0): New.
12591 * hooks.h (hook_int_rtx_insn_0): Declare.
12592 * toplev.c (process_options): Take -gas-loc-support and
12593 -gas-locview-support from dwarf2out. Enable
12594 -gvariable-location-views by default only with locview
12595 assembler support. Enable -ginternal-reset-location-views by
12596 default only if the target defines the corresponding hook.
12597 Enable -ginline-points by default if location views are
12598 enabled; force it disabled if statement frontiers are
12600 * tree-inline.c (expand_call_inline): Check option variables.
12601 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12603 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12605 PR tree-optimization/84321
12606 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12607 handling. Also check whether the anti-range contains any values
12608 that satisfy the mask; switch to a VR_RANGE if not.
12610 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12613 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12615 2018-02-13 Martin Jambor <mjambor@suse.cz>
12618 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12619 of call statements, also set location of a load to a temporary.
12621 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12623 * config/rl78/rl78.c (add_vector_labels): New function.
12624 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12625 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12626 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12627 which checks that no arguments are passed.
12628 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12629 * doc/extend.texi: Documentation for the new attribute.
12631 2018-02-13 Andreas Schwab <schwab@suse.de>
12633 * config/riscv/linux.h (CPP_SPEC): Define.
12635 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12638 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12639 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12640 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12641 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12642 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12643 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12644 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12645 temporarily for AES and PCLMUL builtins.
12647 PR tree-optimization/84339
12648 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12649 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12652 PR middle-end/84309
12653 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12654 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12655 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12657 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12659 * omp-simd-clone.h: New file.
12660 * omp-simd-clone.c: Include omp-simd-clone.h.
12661 (expand_simd_clones): No longer static.
12662 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12663 cgraph.h and omp-simd-clone.h.
12664 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12665 (vect_recog_widen_shift_pattern): Formatting fix.
12666 (vect_pattern_recog_1): Don't check optab for calls.
12669 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12670 operands[2] into a REG before using gen_lowpart on it.
12672 2018-02-12 Jeff Law <law@redhat.com>
12675 * config/sh/sh.c (find_barrier): Consider a sibling call
12678 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12679 successfully back substituting a reg.
12681 2018-02-12 Richard Biener <rguenther@suse.de>
12683 PR tree-optimization/84037
12684 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12685 parameter, move visited init to caller.
12686 (vect_slp_analyze_operations): Separate cost from validity
12687 check, initialize visited once for all instances.
12688 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12690 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12692 (vect_model_promotion_demotion_cost): Likewise.
12693 (vectorizable_bswap): Guard cost modeling with !slp_node
12694 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12696 (vectorizable_call): Likewise.
12697 (vectorizable_conversion): Likewise.
12698 (vectorizable_assignment): Likewise.
12699 (vectorizable_shift): Likewise.
12700 (vectorizable_operation): Likewise.
12701 (vectorizable_store): Likewise.
12702 (vectorizable_load): Likewise.
12703 (vectorizable_condition): Likewise.
12704 (vectorizable_comparison): Likewise.
12706 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12709 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12710 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12711 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12713 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12715 PR middle-end/83665
12716 * params.def (inline-min-speedup): Increase from 8 to 15.
12717 (max-inline-insns-auto): Decrease from 40 to 30.
12718 * ipa-split.c (consider_split): Add some buffer for function to
12719 be considered inlining candidate.
12720 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12723 2018-02-12 Richard Biener <rguenther@suse.de>
12725 PR tree-optimization/84037
12726 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12727 matched stmts if we cannot swap the non-matched ones.
12729 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12731 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12732 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12733 _mm_maskz_scalef_round_ss): New intrinsics.
12734 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12735 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12736 __builtin_ia32_scalefss_round): Remove.
12737 (__builtin_ia32_scalefsd_mask_round,
12738 __builtin_ia32_scalefss_mask_round): New intrinsics.
12739 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12740 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12741 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12742 "<round_constraint>")): Changed to ...
12743 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12744 "<round_scalar_constraint>")): ... this.
12745 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12746 %0, %1, %2<round_op3>}"): Changed to ...
12747 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12748 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12749 %2<round_scalar_mask_op3>}"): ... this.
12750 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12752 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12754 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12755 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12756 (_mm_maskz_sqrt_round_ss): New intrinsics.
12757 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12758 (__builtin_ia32_sqrtsd_mask_round)
12759 (__builtin_ia32_sqrtss_mask_round): New builtins.
12760 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12761 (__builtin_ia32_sqrtss_round): Remove.
12762 (__builtin_ia32_sqrtsd_mask_round)
12763 (__builtin_ia32_sqrtss_mask_round): New builtins.
12764 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12765 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12766 ((match_operand:VF_128 1 "vector_operand"
12767 "xBm,<round_constraint>")): Changed to ...
12768 ((match_operand:VF_128 1 "vector_operand"
12769 "xBm,<round_scalar_constraint>")): ... this.
12770 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12771 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12772 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12773 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12774 %<iptr>1<round_scalar_mask_op3>}): ... this.
12775 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12776 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12778 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12781 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12782 Cast vec_cmpeq result to correct type.
12783 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12784 Cast vec_cmpgt result to correct type.
12786 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12788 * final.c (final_scan_insn_1): Renamed from...
12789 (final_scan_insn): ... this. New wrapper, to recover
12790 seen from the outermost call in recursive ones.
12791 * config/sparc/sparc.c (output_return): Drop seen from call.
12792 (output_sibcall): Likewise.
12793 * config/visium/visium.c (output_branch): Likewise.
12795 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12797 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12800 2018-02-10 Alan Modra <amodra@gmail.com>
12803 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12804 Specify LR as an input.
12806 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12809 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12810 remove_member_access_dummy_vars): New functions.
12811 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12812 lower_omp_1, execute_lower_omp): Use them.
12814 PR rtl-optimization/84308
12815 * shrink-wrap.c (spread_components): Release todo vector.
12817 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12819 PR rtl-optimization/57193
12820 * ira-color.c (struct allocno_color_data): Add member
12821 conflict_allocno_hard_prefs.
12822 (update_conflict_allocno_hard_prefs): New.
12823 (bucket_allocno_compare_func): Add a preference based on
12824 conflict_allocno_hard_prefs.
12825 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12826 (color_allocnos): Remove a dead code. Initiate
12827 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12829 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12832 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12833 constraint from =wa to wa. Avoid a subreg on the output operand,
12834 instead use a pseudo and subreg it in a move.
12835 (p9_xxbrd_<mode>): Changed to ...
12836 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12837 (p9_xxbrd_v2df): New expander.
12838 (p9_xxbrw_<mode>): Changed to ...
12839 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12840 (p9_xxbrw_v4sf): New expander.
12842 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12844 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12846 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12849 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12850 multiply in 32-bit mode.
12851 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12852 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12855 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12857 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12858 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12859 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12860 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12862 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12865 * dwarf2out.c (is_trivial_indirect_ref): New function.
12866 (dwarf2out_late_global_decl): Do not generate a location
12867 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12868 and that are not defined in the current unit.
12870 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12872 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12873 instead of a libcall for UNORDERED.
12875 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12878 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12879 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12881 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12884 * config/s390/s390.c (s390_set_current_function): Invoke
12885 s390_indirect_branch_settings also if fndecl didn't change.
12887 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12889 * config/rs6000/rs6000.md (blockage): Set length to zero.
12891 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12893 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12895 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12898 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12899 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12903 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12904 PARALLEL incoming that failed vt_get_decl_and_offset check.
12906 PR middle-end/84237
12907 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12908 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12910 (get_variable_section): For decls in named .bss* sections pass true as
12911 second argument to bss_initializer_p.
12913 2018-02-09 Marek Polacek <polacek@redhat.com>
12914 Jakub Jelinek <jakub@redhat.com>
12917 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12918 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12919 Sync some changes from cxx_fold_indirect_ref.
12921 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12923 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12925 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12926 (BLOCK_INLINE_ENTRY_LABEL): New.
12927 (dwarf2out_var_location): Disregard inline entry markers.
12928 (inline_entry_data): New struct.
12929 (inline_entry_data_hasher): New hashtable type.
12930 (inline_entry_data_hasher::hash): New.
12931 (inline_entry_data_hasher::equal): New.
12932 (inline_entry_data_table): New variable.
12933 (add_high_low_attributes): Add DW_AT_entry_pc and
12934 DW_AT_GNU_entry_view attributes if a pending entry is found
12935 in inline_entry_data_table. Add old entry_pc attribute only
12936 if debug nonbinding markers are disabled.
12937 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12938 markers are enabled.
12939 (block_within_block_p, dwarf2out_inline_entry): New.
12940 (dwarf2out_finish): Check that no entries remained in
12941 inline_entry_data_table.
12942 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12943 (final_scan_insn, notice_source_line): Likewise.
12944 (rest_of_clean_state): Skip inline entry markers.
12945 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12947 * gimple.c (gimple_build_debug_inline_entry): New.
12948 * gimple.h (enum gimple_debug_subcode): Add
12949 GIMPLE_DEBUG_INLINE_ENTRY.
12950 (gimple_build_debug_inline_entry): Declare.
12951 (gimple_debug_inline_entry_p): New.
12952 (gimple_debug_nonbind_marker_p): Adjust.
12953 * insn-notes.def (INLINE_ENTRY): New.
12954 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12955 inline entry marker notes.
12956 (print_insn): Likewise.
12957 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12958 (INSN_DEBUG_MARKER_KIND): Likewise.
12959 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12960 * tree-inline.c (expand_call_inline): Build and insert
12961 debug_inline_entry stmt.
12962 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12963 inline entry blocks early, if nonbind markers are enabled.
12964 (dump_scope_block): Dump fragment info.
12965 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12966 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12967 (gimple_build_debug_inline_entry): New.
12968 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12969 Enable/disable inline entry points too.
12970 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12971 (DEBUG_INSN): Describe inline entry markers.
12973 * common.opt (gvariable-location-views): New.
12974 (gvariable-location-views=incompat5): New.
12975 * config.in: Rebuilt.
12976 * configure: Rebuilt.
12977 * configure.ac: Test assembler for view support.
12978 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12979 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12980 * dwarf2out.c (var_loc_view): New typedef.
12981 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12982 (dwarf2out_locviews_in_attribute): New.
12983 (dwarf2out_locviews_in_loclist): New.
12984 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12985 (enum dw_line_info_opcode): Add LI_adv_address.
12986 (struct dw_line_info_table): Add view.
12987 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12988 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12989 (zero_view_p): New variable.
12990 (ZERO_VIEW_P): New macro.
12991 (output_asm_line_debug_info): New.
12992 (struct var_loc_node): Add view.
12993 (add_AT_view_list, AT_loc_list): New.
12994 (add_var_loc_to_decl): Add view param. Test it against last.
12995 (new_loc_list): Add view params. Record them.
12996 (AT_loc_list_ptr): Handle loc and view lists.
12997 (view_list_to_loc_list_val_node): New.
12998 (print_dw_val): Handle dw_val_class_view_list.
12999 (size_of_die): Likewise.
13000 (value_format): Likewise.
13001 (loc_list_has_views): New.
13002 (gen_llsym): Set vl_symbol too.
13003 (maybe_gen_llsym, skip_loc_list_entry): New.
13004 (dwarf2out_maybe_output_loclist_view_pair): New.
13005 (output_loc_list): Output view list or entries too.
13006 (output_view_list_offset): New.
13007 (output_die): Handle dw_val_class_view_list.
13008 (output_dwarf_version): New.
13009 (output_compilation_unit_header): Use it.
13010 (output_skeleton_debug_sections): Likewise.
13011 (output_rnglists, output_line_info): Likewise.
13012 (output_pubnames, output_aranges): Update version comments.
13013 (output_one_line_info_table): Output view numbers in asm comments.
13014 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13015 Call maybe_gen_llsym.
13016 (loc_list_from_tree_1): Adjust.
13017 (add_AT_location_description): Create view list attribute if
13018 needed, check it's absent otherwise.
13019 (convert_cfa_to_fb_loc_list): Adjust.
13020 (maybe_emit_file): Call output_asm_line_debug_info for test.
13021 (dwarf2out_var_location): Reset views as needed. Precompute
13022 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13023 attribute. Set view.
13024 (new_line_info_table): Reset next view.
13025 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13026 (dwarf2out_source_line): Likewise. Output view resets and labels to
13027 the assembler, or select appropriate line info opcodes.
13028 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13029 (optimize_string_length): Catch it. Adjust.
13030 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13031 dw_val_class_view_list, and remove it if no longer needed.
13032 (hash_loc_list): Hash view numbers.
13033 (loc_list_hasher::equal): Compare them.
13034 (optimize_location_lists): Check whether a view list symbol is
13035 needed, and whether the locview attribute is present, and
13036 whether they match. Remove the locview attribute if no longer
13038 (index_location_lists): Call skip_loc_list_entry for test.
13039 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13040 Use output_dwarf_version.
13041 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13042 (struct dw_val_node): Add val_view_list.
13043 * final.c (SEEN_NEXT_VIEW): New.
13044 (set_next_view_needed): New.
13045 (clear_next_view_needed): New.
13046 (maybe_output_next_view): New.
13047 (final_start_function): Rename to...
13048 (final_start_function_1): ... this. Take pointer to FIRST,
13049 add SEEN parameter. Emit param bindings in the initial view.
13050 (final_start_function): Reintroduce SEEN-less interface.
13051 (final): Rename to...
13052 (final_1): ... this. Take SEEN parameter. Output final pending
13053 next view at the end.
13054 (final): Reintroduce seen-less interface.
13055 (final_scan_insn): Output pending next view before switching
13056 sections or ending a block. Mark the next view as needed when
13057 outputting variable locations. Notify debug backend of section
13058 changes, and of location view changes.
13059 (rest_of_handle_final): Adjust.
13060 * toplev.c (process_options): Autodetect value for debug variable
13061 location views option. Warn on incompat5 without -gdwarf-5.
13062 * doc/invoke.texi (gvariable-location-views): New.
13063 (gvariable-location-views=incompat5): New.
13064 (gno-variable-location-views): New.
13066 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13068 PR tree-optimization/84136
13069 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13070 that the result of find_edge is non-NULL.
13072 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13075 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13076 storing integer register in SImode. Fix cost of 256 and 512
13077 byte aligned SSE register store.
13079 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13081 * config/i386/i386.c (ix86_multiplication_cost): Fix
13082 multiplication cost for TARGET_AVX512DQ.
13084 2018-02-08 Marek Polacek <polacek@redhat.com>
13086 PR tree-optimization/84238
13087 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13090 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13092 PR tree-optimization/84265
13093 * tree-vect-stmts.c (vectorizable_store): Don't treat
13094 VMAT_CONTIGUOUS accesses as grouped.
13095 (vectorizable_load): Likewise.
13097 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13099 PR tree-optimization/81635
13100 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13101 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13102 (test_round_for_mask): New functions.
13103 (wide_int_cc_tests): Call test_round_for_mask.
13104 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13105 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13106 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13107 range returned by get_range_info.
13109 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13112 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13113 * symtab.c: Include builtins.h
13114 (symtab_node::output_to_lto_symbol_table_p): Move here
13115 from lto-streamer-out.c:output_symbol_p.
13116 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13117 (output_symbol_p): Move all logic to symtab.c
13118 (produce_symtab): Update.
13120 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13122 * config/s390/s390-opts.h (enum indirect_branch): Define.
13123 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13124 (s390_indirect_branch_via_thunk)
13125 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13126 (enum s390_indirect_branch_type): Define.
13127 * config/s390/s390.c (struct s390_frame_layout, struct
13128 machine_function): Remove.
13129 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13130 (indirect_branch_table_label_no, indirect_branch_table_name):
13132 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13133 (enum s390_indirect_branch_option): Define.
13134 (s390_return_addr_from_memory): New function.
13135 (s390_handle_string_attribute): New function.
13136 (s390_attribute_table): Add new attribute handler.
13137 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13138 (s390_indirect_branch_via_thunk): New function.
13139 (s390_indirect_branch_via_inline_thunk): New function.
13140 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13141 sibling call optimization for non z10 compiles.
13142 (s390_emit_call): Force indirect branch target to be a single
13143 register. Add r1 clobber for non-z10 compiles.
13144 (s390_emit_epilogue): Emit return jump via return_use expander.
13145 (s390_reorg): Handle JUMP_INSNs as execute targets.
13146 (s390_option_override_internal): Perform validity checks for the
13147 new command line options.
13148 (s390_indirect_branch_attrvalue): New function.
13149 (s390_indirect_branch_settings): New function.
13150 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13151 (s390_output_indirect_thunk_function): New function.
13152 (s390_code_end): Implement target hook.
13153 (s390_case_values_threshold): Implement target hook.
13154 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13156 * config/s390/s390.h (struct s390_frame_layout)
13157 (struct machine_function): Move here from s390.c.
13158 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13159 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13160 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13161 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13162 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13163 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13164 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13165 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13166 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13167 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13168 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13169 (mnemonic attribute): Add values which aren't recognized
13171 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13172 pattern for branch conversion. Fix mnemonic attribute.
13173 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13174 indirect branch via thunk if requested.
13175 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13176 ("*indirect_jump"): Disable for branch conversion using out of
13178 ("indirect_jump_via_thunk<mode>_z10")
13179 ("indirect_jump_via_thunk<mode>")
13180 ("indirect_jump_via_inlinethunk<mode>_z10")
13181 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13182 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13183 ("casesi_jump_via_inlinethunk<mode>_z10")
13184 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13185 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13186 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13187 ("*indirect2_jump"): Disable for branch conversion.
13188 ("casesi_jump"): Turn into expander and expand patterns for branch
13190 ("return_use"): New expander.
13191 ("*return"): Emit return via thunk and rename it to ...
13192 ("*return<mode>"): ... this one.
13193 * config/s390/s390.opt: Add new options and and enum for the
13196 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13198 * lra-constraints.c (match_reload): Unconditionally use
13199 gen_lowpart_SUBREG, rather than selecting between that
13200 and equivalent gen_rtx_SUBREG code.
13202 2018-02-08 Richard Biener <rguenther@suse.de>
13204 PR tree-optimization/84233
13205 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13206 changed flag instead of boguously re-using phi_inserted.
13208 2018-02-08 Martin Jambor <mjambor@suse.cz>
13210 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13211 static local variables.
13213 2018-02-08 Richard Biener <rguenther@suse.de>
13215 PR tree-optimization/84278
13216 * tree-vect-stmts.c (vectorizable_store): When looking for
13217 smaller vector types to perform grouped strided loads/stores
13218 make sure the mode is supported by the target.
13219 (vectorizable_load): Likewise.
13221 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13223 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13224 Increase LDP/STP opportunities by adding adjacent callee-saves.
13226 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13228 PR rtl-optimization/84068
13229 PR rtl-optimization/83459
13230 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13232 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13234 PR tree-optimization/84224
13235 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13236 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13237 non-zero arguments.
13239 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13242 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13243 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13244 count, remove one USE.
13246 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13248 * doc/install.texi (Configuration): Document the
13249 --with-long-double-format={ibm,ieee} PowerPC configuration
13253 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13254 Convert from define_expand to be define_insn_and_split. Rework
13255 float/double/_Float128 conversions to QI/HI/SImode to work with
13256 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13257 conversions to QI/HImode types did a store and then a load to
13258 truncate the value. For conversions to VSX registers, don't split
13259 the insn, instead emit the code directly. Use the code iterator
13260 any_fix to combine signed and unsigned conversions.
13261 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13262 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13263 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13264 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13265 (fix_<mode>di2_hw): Likewise.
13266 (fixuns_<mode>di2_hw): Likewise.
13267 (fix_<mode>si2_hw): Likewise.
13268 (fixuns_<mode>si2_hw): Likewise.
13269 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13270 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13271 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13272 fix<uns>_trunc<SFDF:mode>si2_p8.
13273 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13274 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13275 (fix<uns>_<mode>_mem): Likewise.
13276 (fctiw<u>z_<mode>_mem): Likewise.
13277 (fix<uns>_<mode>_mem): Likewise.
13278 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13279 the register allocator from doing a direct move to the GPRs to do
13280 a store, and instead use the ISA 3.0 store byte/half-word from
13281 vector register instruction. For IEEE 128-bit floating point,
13282 also optimize stores of 32-bit ints.
13283 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13285 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13287 * genextract.c (push_pathstr_operand): New function to support
13289 (walk_rtx): Call push_pathstr_operand.
13290 (print_path): Support [a-zA-Z].
13292 2018-02-07 Richard Biener <rguenther@suse.de>
13294 PR tree-optimization/84037
13295 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13296 (cse_and_gimplify_to_preheader): Declare.
13297 (vect_get_place_in_interleaving_chain): Likewise.
13298 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13300 (_loop_vec_info::~_loop_vec_info): Delete it.
13301 (cse_and_gimplify_to_preheader): New function.
13302 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13303 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13304 (vectorizable_load): Likewise. For grouped stores always base
13305 the IV on the first element.
13306 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13307 condition before gimplifying.
13309 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13311 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13312 *DIV_EXPR and *MOD_EXPR.
13314 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13317 * config/i386/i386.c (ix86_option_override_internal): Mask out
13318 the CF_SET bit when checking -fcf-protection.
13320 2018-02-07 Tom de Vries <tom@codesourcery.com>
13323 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13326 2018-02-07 Richard Biener <rguenther@suse.de>
13328 PR tree-optimization/84204
13329 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13332 PR tree-optimization/84205
13333 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13334 special-case isl_ast_op_zdiv_r.
13336 PR tree-optimization/84223
13337 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13338 Only add conditions from within the region.
13339 (gather_bbs::after_dom_children): Adjust.
13341 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13344 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13345 * config/avr/avr.md: Only post-reload split REG-REG moves if
13346 either register is GENERAL_REG_P.
13348 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13350 PR tree-optimization/84235
13351 * tree-ssa-scopedtables.c
13352 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13353 if the subtraction is performed in floating point type where NaNs are
13354 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13355 build 1. Formatting fix.
13357 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13360 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13361 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13362 and skip it regardless of bb boundaries. Use CALL_P macro,
13363 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13364 unnecessarily, formatting fix.
13366 2018-02-06 Michael Collison <michael.collison@arm.com>
13368 * config/arm/thumb2.md:
13369 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13370 (*thumb_mov_notscc): Ditto.
13372 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13375 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13376 unsigned_fix, not "s".
13378 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13380 * configure.ac (gcc_fn_eh_frame_ro): New function.
13381 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13382 correct .eh_frame permissions.
13383 * configure: Regenerate.
13385 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13387 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13388 irrelevant options.
13390 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13392 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13393 Display warning message for -mno-speculate-indirect-jumps.
13395 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13397 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13399 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13401 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13403 PR tree-optimization/84225
13404 * tree-eh.c (find_trapping_overflow): Only call
13405 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13407 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13410 * config/i386/i386.c: Reimplement the check of possible options
13411 -mibt/-mshstk conbination. Change error messages.
13412 * doc/invoke.texi: Fix a typo: remove extra '='.
13414 2018-02-06 Marek Polacek <polacek@redhat.com>
13416 PR tree-optimization/84228
13417 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13419 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13422 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13423 emitted arch directives.
13424 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13425 __ARM_FEATURE_COPROC before changing architectures.
13427 2018-02-06 Richard Biener <rguenther@suse.de>
13429 * config/i386/i386.c (print_reg): Fix typo.
13430 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13432 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13434 * configure: Regenerate.
13436 2018-02-05 Martin Sebor <msebor@redhat.com>
13438 PR tree-optimization/83369
13439 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13442 2018-02-05 Martin Liska <mliska@suse.cz>
13444 * doc/invoke.texi: Cherry-pick upstream r323995.
13446 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13448 * ira.c (ira_init_register_move_cost): Adjust comment.
13450 2018-02-05 Martin Liska <mliska@suse.cz>
13452 PR gcov-profile/84137
13453 * doc/gcov.texi: Fix typo in documentation.
13455 2018-02-05 Martin Liska <mliska@suse.cz>
13457 PR gcov-profile/83879
13458 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13459 using dlopen functionality.
13461 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13463 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13464 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13465 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13466 _mm_maskz_range_round_ss): New intrinsics.
13467 (__builtin_ia32_rangesd128_round)
13468 (__builtin_ia32_rangess128_round): Remove.
13469 (__builtin_ia32_rangesd128_mask_round,
13470 __builtin_ia32_rangess128_mask_round): New builtins.
13471 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13472 __builtin_ia32_rangess128_round): Remove.
13473 (__builtin_ia32_rangesd128_mask_round,
13474 __builtin_ia32_rangess128_mask_round): New builtins.
13475 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13476 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13477 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13478 "<round_saeonly_constraint>")): Changed to ...
13479 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13480 "<round_saeonly_scalar_constraint>")): ... this.
13481 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13482 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13483 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13484 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13485 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13487 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13489 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13491 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13492 Remove all values except native, 8540 and 8548.
13494 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13496 * config/i386/i386.c (ix86_output_function_return): Pass
13497 INVALID_REGNUM, instead of -1, as invalid register number to
13498 indirect_thunk_name and output_indirect_thunk.
13500 2018-02-02 Julia Koval <julia.koval@intel.com>
13502 * config.gcc: Add -march=icelake.
13503 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13504 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13505 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13506 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13507 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13508 (processor_target_table): Add icelake.
13509 (ix86_option_override_internal): Handle new PTAs.
13510 (get_builtin_code_for_version): Handle icelake.
13511 (M_INTEL_COREI7_ICELAKE): New.
13512 (fold_builtin_cpu): Handle icelake.
13513 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13514 * doc/invoke.texi: Add -march=icelake.
13516 2018-02-02 Julia Koval <julia.koval@intel.com>
13518 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13519 to wide_int_bitmask.
13520 * wide-int-bitmask.h: New.
13522 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13525 * config/i386/i386.md: Replace Pmode with word_mode in
13526 builtin_setjmp_setup and builtin_longjmp to support x32.
13528 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13532 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13534 (rs6000_supported_cpu_names): New static variable.
13535 (linux_cpu_translation_table): Likewise.
13536 (elf_platform) <cpu>: Define new static variable and use it.
13537 Translate kernel AT_PLATFORM name to canonical name if needed.
13538 Error if platform name is unknown.
13540 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13543 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13545 2018-02-01 Jeff Law <law@redhat.com>
13548 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13549 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13550 the scratch if RELEASE_VIA_POP is false.
13551 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13552 If we have to save a temporary register, decrement SIZE appropriately.
13553 Pass new arguments to release_scratch_register_on_entry.
13554 (ix86_adjust_stack_and_probe): Likewise.
13555 (ix86_emit_probe_stack_range): Pass new arguments to
13556 release_scratch_register_on_entry.
13558 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13560 PR rtl-optimization/84157
13561 * combine.c (change_zero_ext): Use REG_P predicate in
13562 front of HARD_REGISTER_P predicate.
13564 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13566 * config/avr/avr.c (avr_option_override): Move disabling of
13567 -fdelete-null-pointer-checks to...
13568 * common/config/avr/avr-common.c (avr_option_optimization_table):
13571 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13573 PR tree-optimization/81635
13574 * tree-data-ref.c (split_constant_offset_1): For types that
13575 wrap on overflow, try to use range info to prove that wrapping
13578 2018-02-01 Renlin Li <renlin.li@arm.com>
13581 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13582 TAILCALL_ADDR_REGS.
13583 (aarch64_register_move_cost): Likewise.
13584 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13585 TAILCALL_ADDR_REGS.
13586 (REG_CLASS_NAMES): Likewise.
13587 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13588 TAILCALL_ADDR_REGS. Remove IP registers.
13589 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13591 2018-02-01 Richard Biener <rguenther@suse.de>
13593 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13594 for specifying RPO order and allow NULL for that.
13595 * domwalk.c (dom_walker::dom_walker): Likewise.
13596 (dom_walker::walk): Handle NULL RPO order.
13597 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13599 (rewrite_update_dom_walker): Likewise.
13600 (mark_def_dom_walker): Likewise.
13602 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13604 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13605 (aarch64_maybe_expand_sve_subreg_move): Declare.
13606 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13607 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13609 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13610 that are semantically a reverse operation.
13611 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13612 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13613 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13615 (aarch64_can_change_mode_class): For big-endian, forbid changes
13616 between two SVE modes if they have different element sizes.
13618 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13620 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13621 the TImode handling for big-endian targets.
13623 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13625 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13626 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13628 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13629 Remove BSWAP handing for big-endian targets and use the form of
13630 LD1RQ appropariate for the mode.
13632 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13634 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13635 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13636 duplicated element.
13638 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13641 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13642 check for operands that need to go through aarch64_sve_reload_be.
13644 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13646 PR tree-optimization/81661
13647 PR tree-optimization/84117
13648 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13649 * tree-eh.c: Include gimplify.h.
13650 (find_trapping_overflow, replace_trapping_overflow,
13651 rewrite_to_non_trapping_overflow): New functions.
13652 * tree-vect-loop.c: Include tree-eh.h.
13653 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13654 * tree-data-ref.c: Include tree-eh.h.
13655 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13657 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13659 PR rtl-optimization/84123
13660 * combine.c (change_zero_ext): Check if hard register satisfies
13661 can_change_dest_mode before calling gen_lowpart_SUBREG.
13663 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13666 * ira.c (ira_init_register_move_cost): Remove assert.
13668 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13670 PR rtl-optimization/84071
13671 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13672 * doc/tm.texi: Regenerate.
13674 2018-01-31 Richard Biener <rguenther@suse.de>
13676 PR tree-optimization/84132
13677 * tree-data-ref.c (analyze_miv_subscript): Properly
13678 check whether evolution_function_is_affine_multivariate_p
13679 before calling gcd_of_steps_may_divide_p.
13681 2018-01-31 Julia Koval <julia.koval@intel.com>
13684 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13685 * config/i386/i386.md (rdpid_rex64) New.
13686 (rdpid): Make 32bit only.
13688 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13691 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13692 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13694 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13697 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13699 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13701 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13703 PR rtl-optimization/84071
13704 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13705 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13707 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13709 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13710 (arc_attribute_table): Add 'aux' attribute.
13711 (arc_in_small_data_p): Consider aux like variables.
13712 (arc_is_aux_reg_p): New function.
13713 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13714 (arc_get_aux_arg): New function.
13715 (prepare_move_operands): Handle aux-register access.
13716 (arc_handle_aux_attribute): New function.
13717 * doc/extend.texi (ARC Variable attributes): Add subsection.
13719 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13721 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13722 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13723 (arc_attribute_table): Add 'uncached' attribute.
13724 (arc_print_operand): Print '.di' flag for uncached memory
13726 (arc_in_small_data_p): Do not consider for small data the uncached
13728 (arc_is_uncached_mem_p): New function.
13729 * config/arc/predicates.md (compact_store_memory_operand): Check
13730 for uncached memory accesses.
13731 (nonvol_nonimm_operand): Likewise.
13732 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13734 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13737 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13738 falign-loops=): Add Optimization flag.
13740 2018-01-30 Jeff Law <law@redhat.com>
13743 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13744 INT_REGISTERS_SAVED. Check it prior to calling
13745 get_scratch_register_on_entry.
13746 (ix86_adjust_stack_and_probe): Similarly.
13747 (ix86_emit_probe_stack_range): Similarly.
13748 (ix86_expand_prologue): Corresponding changes.
13750 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13753 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13754 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13756 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13759 * lra-constraints.c (curr_insn_transform): Process AND in the
13762 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13764 PR rtl-optimization/83986
13765 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13766 dependence against last_pending_memory_flush in addition to
13767 pending_jump_insns.
13769 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13771 PR tree-optimization/81611
13772 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13775 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13778 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13781 2018-01-30 Richard Biener <rguenther@suse.de>
13782 Jakub Jelinek <jakub@redhat.com>
13784 PR tree-optimization/84111
13785 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13786 inner loops added during recursion, as they don't have up-to-date
13789 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13792 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13793 (can_inline_edge_by_limits_p): ... here.
13794 (can_early_inline_edge_p, check_callers,
13795 update_caller_keys, update_callee_keys, recursive_inlining,
13796 add_new_edges_to_heap, speculation_useful_p,
13797 inline_small_functions,
13798 inline_small_functions, flatten_function,
13799 inline_to_all_callers_1): Update.
13801 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13803 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13806 2018-01-30 Richard Biener <rguenther@suse.de>
13808 PR tree-optimization/83008
13809 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13810 invariant and constant vector uses in stmts when they need
13811 more than one stmt.
13813 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13816 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13817 * configure: Regenerate.
13819 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13821 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13823 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13824 Use gen_rtx_REG rather than gen_lowpart.
13826 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13828 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13829 rather than 0 when creating partial subregs.
13831 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13833 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13836 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13839 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13840 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13841 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13842 flags. This restores the settings used before the 2017-07-24.
13843 Turning off pre increment/decrement/modify allows IVOPTS to
13844 optimize DF/SF loops where the index is an int.
13846 2018-01-29 Richard Biener <rguenther@suse.de>
13847 Kelvin Nilsen <kelvin@gcc.gnu.org>
13850 * tree-vect-stmts.c (vectorizable_call): Don't call
13851 targetm.vectorize_builtin_md_vectorized_function if callee is
13854 2018-01-22 Carl Love <cel@us.ibm.com>
13856 * doc/extend.tex: Fix typo in second arg in
13857 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13859 2018-01-29 Richard Biener <rguenther@suse.de>
13861 PR tree-optimization/84086
13862 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13863 (flush_ssaname_freelist): When SSA names were released reset
13864 the SCEV hash table.
13866 2018-01-29 Richard Biener <rguenther@suse.de>
13868 PR tree-optimization/84057
13869 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13870 removed paths when removing edges.
13872 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13874 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13875 -mfunction-return=@var{choice}.
13877 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13879 PR diagnostic/84034
13880 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13881 Handle CR like TAB.
13882 (layout::print_source_line): Likewise.
13883 (test_get_line_width_without_trailing_whitespace): Add test cases.
13885 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13887 PR middle-end/84040
13888 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13891 2018-01-26 Jim Wilson <jimw@sifive.com>
13893 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13895 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13898 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13900 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13901 and CMP + SUB-immediate -> SUBS.
13903 2018-01-26 Martin Sebor <msebor@redhat.com>
13905 PR tree-optimization/83896
13906 * tree-ssa-strlen.c (get_string_len): Rename...
13907 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13908 Avoid assuming length is constant.
13909 (handle_char_store): Use HOST_WIDE_INT for string length.
13911 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13914 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13915 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13917 2018-01-26 Richard Biener <rguenther@suse.de>
13919 PR rtl-optimization/84003
13920 * dse.c (record_store): Only record redundant stores when
13921 the earlier store aliases at least all accesses the later one does.
13923 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13925 PR rtl-optimization/83985
13926 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13927 REG_CFA_RESTORE insns.
13928 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13929 REG_CFA_RESTORE insns here.
13932 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13933 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13935 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13937 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13938 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13939 (arc_init): Likewise.
13940 (arc_override_options): Likewise.
13941 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13943 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13945 * config/arc/arc.h (TARGET_DBNZ): Define.
13946 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13947 properly set the tune attribute.
13948 (dbnz): Use TARGET_DBNZ guard.
13949 * config/arc/arc.opt (mtune): Add core3 option.
13951 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13953 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13954 recognize new pic like addresses.
13955 (arc_delegitimize_address): Clean up.
13957 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13959 * config/arc/arc-arches.def: Option mrf16 valid for all
13961 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13962 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13963 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13964 * config/arc/arc-tables.opt: Regenerate.
13965 * config/arc/arc.c (arc_conditional_register_usage): Handle
13966 reduced register file case.
13967 (arc_file_start): Set must have build attributes.
13968 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13969 mrf16 option value.
13970 * config/arc/arc.opt (mrf16): Add new option.
13971 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13972 * config/arc/genmultilib.awk: Handle new mrf16 option.
13973 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13974 * config/arc/t-multilib: Regenerate.
13975 * doc/invoke.texi (ARC Options): Document mrf16 option.
13977 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13979 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13980 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13981 (arc_attribute_table): Add 'secure_call' attribute.
13982 (arc_print_operand): Print secure call operand.
13983 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13985 (arc_is_secure_call_p): New function. * config/arc/arc.md
13986 (call_i): Add support for sjli instruction.
13987 (call_value_i): Likewise.
13988 * config/arc/constraints.md (Csc): New constraint.
13990 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13991 John Eric Martin <John.Martin@emmicro-us.com>
13993 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13994 * config/arc/arc.c (_arc_jli_section): New struct.
13995 (arc_jli_section): New type.
13996 (rc_jli_sections): New static variable.
13997 (arc_handle_jli_attribute): New function.
13998 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13999 (arc_file_end): New function.
14000 (TARGET_ASM_FILE_END): Define.
14001 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14002 (arc_add_jli_section): New function.
14003 (jli_call_scan): Likewise.
14004 (arc_reorg): Call jli_call_scan.
14005 (arc_output_addsi): Remove 'S' from printing asm operand.
14006 (arc_is_jli_call_p): New function.
14007 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14009 (movhi_insn): Likewise.
14010 (movsi_insn): Likewise.
14011 (movsi_set_cc_insn): Likewise.
14012 (loadqi_update): Likewise.
14013 (load_zeroextendqisi_update): Likewise.
14014 (load_signextendqisi_update): Likewise.
14015 (loadhi_update): Likewise.
14016 (load_zeroextendhisi_update): Likewise.
14017 (load_signextendhisi_update): Likewise.
14018 (loadsi_update): Likewise.
14019 (loadsf_update): Likewise.
14020 (movsicc_insn): Likewise.
14021 (bset_insn): Likewise.
14022 (bxor_insn): Likewise.
14023 (bclr_insn): Likewise.
14024 (bmsk_insn): Likewise.
14025 (bicsi3_insn): Likewise.
14026 (cmpsi_cc_c_insn): Likewise.
14027 (movsi_ne): Likewise.
14028 (movsi_cond_exec): Likewise.
14029 (clrsbsi2): Likewise.
14030 (norm_f): Likewise.
14040 (call_i): Remove 'S' asm letter, add jli instruction.
14041 (call_value_i): Likewise.
14042 * config/arc/arc.op (mjli-always): New option.
14043 * config/arc/constraints.md (Cji): New constraint.
14044 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14046 (subsf3_fpx): Likewise.
14047 (mulsf3_fpx): Likewise.
14048 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14050 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14051 function attrbutes.
14052 * doc/invoke.texi (ARC): Document mjli-always option.
14054 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14056 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14057 avoid addition with 0 and use incw and decw where possible.
14059 2018-01-26 Richard Biener <rguenther@suse.de>
14061 PR tree-optimization/81082
14062 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14063 association if it requires casting to unsigned.
14064 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14065 from fold_plusminus_mult_expr to catch important cases late when
14066 range info is available.
14068 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14070 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14071 * configure.ac (hidden_linkonce): New test.
14072 * configure: Regenerate.
14073 * config.in: Regenerate.
14075 2018-01-26 Julia Koval <julia.koval@intel.com>
14077 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14078 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14079 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14080 _mm_mask_bitshuffle_epi64_mask): Fix type.
14081 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14082 USI_FTYPE_V4DI_V4DI_USI): Remove.
14083 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14084 __builtin_ia32_vpshufbitqmb256_mask,
14085 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14086 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14087 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14089 2018-01-26 Alan Modra <amodra@gmail.com>
14092 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14093 UNSPEC_VBPERMQ. Sort other unspecs.
14095 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14097 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14099 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14101 PR middle-end/83055
14102 * predict.c (drop_profile): Do not push/pop cfun; update also
14104 (handle_missing_profiles): Fix logic looking for zero profiles.
14106 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14108 PR middle-end/83977
14109 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14110 on functions with #pragma omp declare simd or functions with simd
14112 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14113 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14114 Remove trailing \n from warning_at calls.
14116 2018-01-25 Tom de Vries <tom@codesourcery.com>
14119 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14120 for neutered workers.
14122 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14125 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14126 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14128 2018-01-24 Jeff Law <law@redhat.com>
14131 * i386.c (get_probe_interval): Move to earlier point.
14132 (ix86_compute_frame_layout): If -fstack-clash-protection and
14133 the frame is larger than the probe interval, then use pushes
14134 to save registers rather than reg->mem moves.
14135 (ix86_expand_prologue): Remove conditional for int_registers_saved
14138 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14141 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14142 min/max for never referenced object.
14144 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14146 PR middle-end/83977
14147 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14149 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14150 attributes from DECL_ATTRIBUTES (decl) without affecting
14151 DECL_ATTRIBUTES (current_function_decl).
14152 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14153 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14155 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14157 PR tree-optimization/83979
14158 * fold-const.c (fold_comparison): Use constant_boolean_node
14159 instead of boolean_{true,false}_node.
14161 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14163 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14166 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14168 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14169 Simplify the clause that sets the length attribute.
14170 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14171 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14172 clause that sets the length attribute.
14173 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14175 2018-01-24 Tom de Vries <tom@codesourcery.com>
14178 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14179 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14180 Add strict parameter.
14181 (prevent_branch_around_nothing): Insert dummy insn between branch to
14182 label and label with no ptx insn inbetween.
14183 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14185 2018-01-24 Tom de Vries <tom@codesourcery.com>
14188 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14189 for neutered threads in warp.
14190 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14192 2018-01-24 Richard Biener <rguenther@suse.de>
14194 PR tree-optimization/83176
14195 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14198 2018-01-24 Richard Biener <rguenther@suse.de>
14200 PR tree-optimization/82819
14201 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14202 code generating pluses that are no-ops in the target precision.
14204 2018-01-24 Richard Biener <rguenther@suse.de>
14206 PR middle-end/84000
14207 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14209 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14211 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14212 to merge probabilities.
14213 * predict.c (probably_never_executed): Also mark as cold functions
14214 with global 0 profile and guessed local profile.
14215 * profile-count.c (profile_probability::combine_with_count): New
14217 * profile-count.h (profile_probability::operator*,
14218 profile_probability::operator*=, profile_probability::operator/,
14219 profile_probability::operator/=): Reduce precision to adjusted
14220 and set value to guessed on contradictory divisions.
14221 (profile_probability::combine_with_freq): Remove.
14222 (profile_probability::combine_wiht_count): Declare.
14223 (profile_count::force_nonzero):: Set to adjusted.
14224 (profile_count::probability_in):: Set quality to adjusted.
14225 * tree-ssa-tail-merge.c (replace_block_by): Use
14226 combine_with_count.
14228 2018-01-23 Andrew Waterman <andrew@sifive.com>
14229 Jim Wilson <jimw@sifive.com>
14231 * config/riscv/riscv.c (riscv_stack_boundary): New.
14232 (riscv_option_override): Set riscv_stack_boundary. Handle
14233 riscv_preferred_stack_boundary_arg.
14234 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14235 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14236 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14237 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14238 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14239 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14241 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14244 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14245 of struct ix86_frame.
14246 (ix86_expand_epilogue): Likewise. Add a local variable for
14247 the reg_save_offset field in struct ix86_frame.
14249 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14251 PR tree-optimization/82604
14252 * tree-loop-distribution.c (enum partition_kind): New enum item
14253 PKIND_PARTIAL_MEMSET.
14254 (partition_builtin_p): Support above new enum item.
14255 (generate_code_for_partition): Ditto.
14256 (compute_access_range): Differentiate cases that equality can be
14257 proven at all loops, the innermost loops or no loops.
14258 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14259 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14260 (finalize_partitions, distribute_loop): Don't fuse partition of
14261 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14262 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14263 parloop is enabled.
14265 2018-01-23 Martin Liska <mliska@suse.cz>
14267 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14268 order to ignore the predictor.
14269 (PRED_POLYMORPHIC_CALL): Likewise.
14270 (PRED_RECURSIVE_CALL): Likewise.
14272 2018-01-23 Martin Liska <mliska@suse.cz>
14274 * tree-profile.c (tree_profiling): Print function header to
14275 aware reader which function we are working on.
14276 * value-prof.c (gimple_find_values_to_profile): Do not print
14277 not interesting value histograms.
14279 2018-01-23 Martin Liska <mliska@suse.cz>
14281 * profile-count.h (enum profile_quality): Add
14282 profile_uninitialized as the first value. Do not number values
14283 as they are zero based.
14284 (profile_count::verify): Update sanity check.
14285 (profile_probability::verify): Likewise.
14287 2018-01-23 Nathan Sidwell <nathan@acm.org>
14289 * doc/invoke.texi (ffor-scope): Deprecate.
14291 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14293 PR tree-optimization/83510
14294 * domwalk.c (set_all_edges_as_executable): New function.
14295 (dom_walker::dom_walker): Convert bool param
14296 "skip_unreachable_blocks" to enum reachability. Move setup of
14297 edge flags to set_all_edges_as_executable and only do it when
14298 reachability is REACHABLE_BLOCKS.
14299 * domwalk.h (enum dom_walker::reachability): New enum.
14300 (dom_walker::dom_walker): Convert bool param
14301 "skip_unreachable_blocks" to enum reachability.
14302 (set_all_edges_as_executable): New decl.
14303 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14304 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14306 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14307 but converting true to REACHABLE_BLOCKS.
14308 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14310 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14311 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14312 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14314 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14315 if check_all_array_refs will be called.
14317 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14319 * tree.c (selftest::test_location_wrappers): Add more test
14322 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14324 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14325 (selftest::test_bit_in_range): Likewise.
14327 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14330 * doc/sourcebuild.texi (vect_float): Say that the selector
14331 only describes the situation when -funsafe-math-optimizations is on.
14332 (vect_float_strict): Document.
14334 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14336 PR tree-optimization/83965
14337 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14338 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14339 instead of checking only for a reduction.
14340 (vect_recog_widen_sum_pattern): Likewise.
14342 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14344 * predict.c (probably_never_executed): Only use precise profile info.
14345 (compute_function_frequency): Skip after inlining hack since we now
14346 have quality checking.
14348 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14350 * profile-count.h (profile_probability::very_unlikely,
14351 profile_probability::unlikely, profile_probability::even): Set
14352 precision to guessed.
14354 2018-01-23 Richard Biener <rguenther@suse.de>
14356 PR tree-optimization/83963
14357 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14358 Properly terminate dominator walk when crossing the exit edge not
14359 when visiting its source block.
14361 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14364 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14365 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14367 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14369 PR tree-optimization/83957
14370 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14371 semicolon after for body surrounded by braces.
14373 PR tree-optimization/83081
14374 * profile-count.h (profile_probability::split): New method.
14375 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14376 Use profile_probability::split.
14377 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14378 when splitting a single conditional jump into 2.
14380 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14382 PR tree-optimization/69452
14383 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14386 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14388 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14389 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14390 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14392 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14394 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14396 * config/rl78/rl78.md (movdi): New define_expand.
14397 * config/rl78/rl78.c (rl78_split_movdi): New function.
14399 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14402 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14404 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14405 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14406 128-bit to produce an UNSPEC move to get the double word with the
14407 signbit and then a shift directly to do signbit.
14408 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14409 implementation with a new version that just does either a direct
14410 move or a regular move. Move memory interface to separate insns.
14411 Move insns so they are next to the expander.
14412 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14413 with signbit move. Split big and little endian case.
14414 (signbit<mode>2_dm_mem_le): Likewise.
14415 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14416 (signbit<mode>2_dm2): Likewise.
14418 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14420 * config/rl78/rl78.md (anddi3): New define_expand.
14422 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14424 * config/rl78/rl78.md (umindi3): New define_expand.
14426 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14428 * config/rl78/rl78.md (smindi3): New define_expand.
14430 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14432 * config/rl78/rl78.md (smaxdi3): New define_expand.
14434 2018-01-22 Carl Love <cel@us.ibm.com>
14436 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14437 LVX_V1TI): Add macro expansion.
14438 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14439 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14440 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14441 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14442 Change check to determine if the instruction is a byte reversing
14443 entry. Fix typo in comment.
14444 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14445 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14446 Add def_builtin calls for new builtins.
14447 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14448 Add define_insn expansion.
14450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14452 * config/rl78/rl78.md (umaxdi3): New define_expand.
14454 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14456 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14457 for non-QImode registers.
14459 2018-01-22 Richard Biener <rguenther@suse.de>
14461 PR tree-optimization/83963
14462 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14463 including the loop exit block.
14464 (scop_detection::merge_sese): Likewise.
14465 (scop_detection::add_scop): Do it here instead.
14467 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14469 * doc/sourcebuild.texi (arm_softfloat): Document.
14471 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14474 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14475 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14476 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14478 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14479 David Edelsohn <dje.gcc@gmail.com>
14482 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14483 Change "crset eq" to "crset 2".
14484 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14485 (*call_indirect_aix<mode>_nospec): Likewise.
14486 (*call_value_indirect_aix<mode>_nospec): Likewise.
14487 (*call_indirect_elfv2<mode>_nospec): Likewise.
14488 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14489 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14490 change assembly output from . to $.
14491 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14492 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14493 (*tablejump<mode>_internal1_nospec): Likewise.
14495 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14498 * config/sh/sh_optimize_sett_clrt.cc:
14499 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14501 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14503 PR tree-optimization/83940
14504 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14505 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14506 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14507 use it to pass back the definition type.
14508 (vect_check_store_rhs): Likewise rhs_dt_out.
14509 (vect_build_gather_load_calls): Add a mask_dt argument and use
14510 it instead of a call to vect_is_simple_use.
14511 (vectorizable_store): Update calls to vect_check_load_store_mask
14512 and vect_check_store_rhs. Use the dt returned by the latter instead
14513 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14514 instead of calls to vect_is_simple_use. Pass the scalar rather
14515 than the vector operand to vect_is_simple_use when handling
14516 second and subsequent copies of an rhs value.
14517 (vectorizable_load): Update calls to vect_check_load_store_mask
14518 and vect_build_gather_load_calls. Use the cached mask_dt and
14519 gs_info.offset_dt instead of calls to vect_is_simple_use.
14521 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14523 PR middle-end/83945
14524 * tree-emutls.c: Include gimplify.h.
14525 (lower_emutls_2): New function.
14526 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14527 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14528 it before further processing.
14531 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14532 UINTVAL (trueop1) instead of INTVAL (op1).
14534 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14538 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14539 INCOMING_FRAME_SP_OFFSET if not defined.
14540 (scan_trace): Add ENTRY argument. If true and
14541 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14542 emit a note to adjust the CFA offset.
14543 (create_cfi_notes): Adjust scan_trace callers.
14544 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14545 INCOMING_FRAME_SP_OFFSET in the CIE.
14546 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14547 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14549 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14550 * doc/tm.texi: Regenerated.
14552 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14554 PR rtl-optimization/83147
14555 * lra-constraints.c (remove_inheritance_pseudos): Use
14556 lra_substitute_pseudo_within_insn.
14558 2018-01-19 Tom de Vries <tom@codesourcery.com>
14559 Cesar Philippidis <cesar@codesourcery.com>
14562 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14564 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14567 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14568 spaces for function labels.
14570 2018-01-19 Martin Liska <mliska@suse.cz>
14572 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14573 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14574 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14575 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14576 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14577 (PRED_CONST_RETURN): Change from 69 to 65.
14578 (PRED_NULL_RETURN): Change from 91 to 71.
14579 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14580 (PRED_LOOP_GUARD): Change from 66 to 73.
14582 2018-01-19 Martin Liska <mliska@suse.cz>
14584 * predict.c (predict_insn_def): Add new assert.
14585 (struct branch_predictor): Change type to signed integer.
14586 (test_prediction_value_range): Amend test to cover
14587 PROB_UNINITIALIZED.
14588 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14589 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14590 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14591 (PRED_LOOP_IV_COMPARE): Likewise.
14592 * predict.h (PROB_UNINITIALIZED): Define new constant.
14594 2018-01-19 Martin Liska <mliska@suse.cz>
14596 * predict.c (dump_prediction): Add new format for
14597 analyze_brprob.py script which is enabled with -details
14599 * profile-count.h (precise_p): New function.
14601 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14603 PR tree-optimization/83922
14604 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14605 there are no statements that need masking.
14606 (vect_active_double_reduction_p): New function.
14607 (vect_analyze_loop_operations): Use it when handling phis that
14608 are not in the loop header.
14610 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14612 PR tree-optimization/83914
14613 * tree-vect-loop.c (vectorizable_induction): Don't convert
14614 init_expr or apply the peeling adjustment for inductions
14615 that are nested within the vectorized loop.
14617 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14619 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14622 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14626 * function.h (gimplify_parameters): Add gimple_seq * argument.
14627 * function.c: Include gimple.h and options.h.
14628 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14629 for the added local temporaries if needed.
14630 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14631 if there are any parameter cleanups, wrap whole body into a
14632 try/finally with the cleanups.
14634 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14637 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14638 Use GET_MODE_CLASS for scalar floating point.
14640 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14644 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14645 Fix call of call_cgraph_insertion_hooks.
14647 2018-01-18 Martin Sebor <msebor@redhat.com>
14649 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14651 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14654 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14657 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14660 * common.opt: (-ffile-prefix-map): New option.
14661 * opts.c (common_handle_option): Defer it.
14662 * opts-global.c (handle_common_deferred_options): Handle it.
14663 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14664 * file-prefix-map.h: New file.
14665 (remap_debug_filename, add_debug_prefix_map): ...here.
14666 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14667 * final.c (debug_prefix_map, add_debug_prefix_map
14668 remap_debug_filename): Move to...
14669 * file-prefix-map.c: New file.
14670 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14671 generalize, get rid of alloca(), use strrchr() instead of strchr().
14672 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14673 Implement in terms of add_prefix_map().
14674 (remap_macro_filename, remap_debug_filename): Implement in term of
14676 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14677 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14678 * dbxout.c: Include file-prefix-map.h.
14679 * varasm.c: Likewise.
14680 * vmsdbgout.c: Likewise.
14681 * xcoffout.c: Likewise.
14682 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14683 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14684 * doc/invoke.texi (-ffile-prefix-map): Document.
14685 (-fdebug-prefix-map): Update description.
14687 2018-01-18 Martin Liska <mliska@suse.cz>
14689 * config/i386/i386.c (indirect_thunk_name): Document that also
14691 (output_indirect_thunk): Document why both instructions
14692 (pause and lfence) are generated.
14694 2018-01-18 Richard Biener <rguenther@suse.de>
14696 PR tree-optimization/83887
14697 * graphite-scop-detection.c
14698 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14699 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14700 (scop_detection::merge_sese): Re-implement with a flood-fill
14701 algorithm that properly finds a SESE region if it exists.
14703 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14706 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14707 pointer_diff optimizations use view_convert instead of convert.
14709 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14711 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14712 Generate different code for -mno-speculate-indirect-jumps.
14713 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14714 (*call_indirect_aix<mode>): Disable for
14715 -mno-speculate-indirect-jumps.
14716 (*call_indirect_aix<mode>_nospec): New define_insn.
14717 (*call_value_indirect_aix<mode>): Disable for
14718 -mno-speculate-indirect-jumps.
14719 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14720 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14721 -mno-speculate-indirect-jumps.
14722 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14724 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14726 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14727 long double type, set the flags for noting the default long double
14728 type, even if we don't pass or return a long double type.
14730 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14733 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14736 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14738 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14739 support for merge[hl].
14740 (fold_mergehl_helper): New helper function.
14741 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14742 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14743 (altivec_vmrglw_direct): Add xxmrglw insn.
14745 2018-01-17 Andrew Waterman <andrew@sifive.com>
14747 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14748 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14750 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14753 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14754 call the lto_location_cache before reading the
14755 DECL_SOURCE_LOCATION of the types.
14757 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14758 Richard Sandiford <richard.sandiford@linaro.org>
14760 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14761 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14762 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14763 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14764 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14766 * config/aarch64/constraints.md (aarch64_movti_operand):
14768 * config/aarch64/predicates.md (Uti): Add new constraint.
14770 2018-01-17 Carl Love <cel@us.ibm.com>
14772 * config/rs6000/vsx.md (define_expand xl_len_r,
14773 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14774 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14776 (define_expand, define_insn): Move the shift left from the
14777 define_insn to the define_expand for lxvl and stxvl instructions.
14778 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14779 and XL_LEN_R definitions to PURE.
14781 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14783 * config/i386/i386.c (indirect_thunk_name): Declare regno
14784 as unsigned int. Compare regno with INVALID_REGNUM.
14785 (output_indirect_thunk): Ditto.
14786 (output_indirect_thunk_function): Ditto.
14787 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14788 in the call to output_indirect_thunk_function.
14790 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14792 PR middle-end/83884
14793 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14794 rather than the size of inner_type to determine the stack slot size
14795 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14797 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14800 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14803 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14805 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14806 endian Linux systems to optionally enable multilibs for selecting
14807 the long double type if the user configured an explicit type.
14808 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14809 have no long double multilibs if not defined.
14810 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14811 warn if the user used -mabi={ieee,ibm}longdouble and we built
14812 multilibs for long double.
14813 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14814 appropriate multilib option.
14815 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14817 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14818 for building long double multilibs.
14819 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14821 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14823 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14826 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14828 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14831 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14834 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14837 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14839 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14840 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14842 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14844 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14845 different rtl trees depending on TARGET_64BIT.
14846 (rs6000_gen_lvx): Likewise.
14848 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14850 * config/visium/visium.md (nop): Tweak comment.
14851 (hazard_nop): Likewise.
14853 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14855 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14856 -mspeculate-indirect-jumps.
14857 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14858 for -mno-speculate-indirect-jumps.
14859 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14860 (*call_value_indirect_elfv2<mode>): Disable for
14861 -mno-speculate-indirect-jumps.
14862 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14863 (indirect_jump): Emit different RTL for
14864 -mno-speculate-indirect-jumps.
14865 (*indirect_jump<mode>): Disable for
14866 -mno-speculate-indirect-jumps.
14867 (*indirect_jump<mode>_nospec): New define_insn.
14868 (tablejump): Emit different RTL for
14869 -mno-speculate-indirect-jumps.
14870 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14871 (tablejumpsi_nospec): New define_expand.
14872 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14873 (tablejumpdi_nospec): New define_expand.
14874 (*tablejump<mode>_internal1): Disable for
14875 -mno-speculate-indirect-jumps.
14876 (*tablejump<mode>_internal1_nospec): New define_insn.
14877 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14880 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14882 * caller-save.c (insert_save): Drop unnecessary parameter. All
14885 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14886 Richard Biener <rguenth@suse.de>
14889 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14890 return early, inline manually is_gimple_sizepos. Make sure if we
14891 call gimplify_expr we don't end up with a gimple constant.
14892 * tree.c (variably_modified_type_p): Don't return true for
14893 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14894 * gimplify.h (is_gimple_sizepos): Remove.
14896 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14898 PR tree-optimization/83857
14899 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14900 vectorizable_live_operation for pure SLP statements.
14901 (vectorizable_live_operation): Handle PHIs.
14903 2018-01-16 Richard Biener <rguenther@suse.de>
14905 PR tree-optimization/83867
14906 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14907 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14909 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14912 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14913 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14914 If off is not INTEGER_CST, issue a may not be aligned warning
14915 rather than isn't aligned. Use isn%'t rather than isn't.
14916 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14918 <case MULT_EXPR>: Improve the case when bottom and one of the
14919 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14920 operand, in that case check if the other operand is multiple of
14921 bottom divided by the INTEGER_CST operand.
14923 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14926 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14927 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14928 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14929 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14930 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14931 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14932 (pa_function_arg_size): New function.
14934 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14936 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14937 in a separate statement.
14939 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14941 PR tree-optimization/83847
14942 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14943 group gathers and scatters.
14945 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14947 PR rtl-optimization/86620
14948 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14950 PR rtl-optimization/83213
14951 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14952 to last if both are JUMP_INSNs.
14954 PR tree-optimization/83843
14955 * gimple-ssa-store-merging.c
14956 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14957 store_immediate_info for bswap/nop orig_stores.
14959 2018-01-15 Andrew Waterman <andrew@sifive.com>
14961 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14963 <UDIV>: Increase cost if !TARGET_DIV.
14965 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14967 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14968 (define_attr "cr_logical_3op"): New.
14969 (cceq_ior_compare): Adjust.
14970 (cceq_ior_compare_complement): Adjust.
14971 (*cceq_rev_compare): Adjust.
14972 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14973 (is_cracked_insn): Adjust.
14974 (insn_must_be_first_in_group): Adjust.
14975 * config/rs6000/40x.md: Adjust.
14976 * config/rs6000/440.md: Adjust.
14977 * config/rs6000/476.md: Adjust.
14978 * config/rs6000/601.md: Adjust.
14979 * config/rs6000/603.md: Adjust.
14980 * config/rs6000/6xx.md: Adjust.
14981 * config/rs6000/7450.md: Adjust.
14982 * config/rs6000/7xx.md: Adjust.
14983 * config/rs6000/8540.md: Adjust.
14984 * config/rs6000/cell.md: Adjust.
14985 * config/rs6000/e300c2c3.md: Adjust.
14986 * config/rs6000/e500mc.md: Adjust.
14987 * config/rs6000/e500mc64.md: Adjust.
14988 * config/rs6000/e5500.md: Adjust.
14989 * config/rs6000/e6500.md: Adjust.
14990 * config/rs6000/mpc.md: Adjust.
14991 * config/rs6000/power4.md: Adjust.
14992 * config/rs6000/power5.md: Adjust.
14993 * config/rs6000/power6.md: Adjust.
14994 * config/rs6000/power7.md: Adjust.
14995 * config/rs6000/power8.md: Adjust.
14996 * config/rs6000/power9.md: Adjust.
14997 * config/rs6000/rs64.md: Adjust.
14998 * config/rs6000/titan.md: Adjust.
15000 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15002 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15003 ix86_indirect_branch_register logic.
15005 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15007 * config/i386/constraints.md (Bs): Update
15008 ix86_indirect_branch_register check. Don't check
15009 ix86_indirect_branch_register with GOT_memory_operand.
15011 * config/i386/predicates.md (GOT_memory_operand): Don't check
15012 ix86_indirect_branch_register here.
15013 (GOT32_symbol_operand): Likewise.
15015 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15017 * config/i386/predicates.md (constant_call_address_operand):
15018 Rewrite ix86_indirect_branch_register logic.
15019 (sibcall_insn_operand): Likewise.
15021 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15023 * config/i386/constraints.md (Bs): Replace
15024 ix86_indirect_branch_thunk_register with
15025 ix86_indirect_branch_register.
15027 * config/i386/i386.md (indirect_jump): Likewise.
15028 (tablejump): Likewise.
15029 (*sibcall_memory): Likewise.
15030 (*sibcall_value_memory): Likewise.
15031 Peepholes of indirect call and jump via memory: Likewise.
15032 * config/i386/i386.opt: Likewise.
15033 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15034 (GOT_memory_operand): Likewise.
15035 (call_insn_operand): Likewise.
15036 (sibcall_insn_operand): Likewise.
15037 (GOT32_symbol_operand): Likewise.
15039 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15041 PR middle-end/83837
15042 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15043 type rather than type addr's type points to.
15044 (expand_omp_atomic_mutex): Likewise.
15045 (expand_omp_atomic): Likewise.
15047 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15050 * config/i386/i386.c (output_indirect_thunk_function): Use
15051 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15052 for __x86_return_thunk.
15054 2018-01-15 Richard Biener <rguenther@suse.de>
15056 PR middle-end/83850
15057 * expmed.c (extract_bit_field_1): Fix typo.
15059 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15062 * config/arm/iterators.md (VF): New mode iterator.
15063 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15064 Remove integer-related logic from pattern.
15065 (neon_vabd<mode>_3): Likewise.
15067 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15069 PR middle-end/82694
15070 * common.opt (fstrict-overflow): No longer an alias.
15071 (fwrapv-pointer): New option.
15072 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15073 also for pointer types based on flag_wrapv_pointer.
15074 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15075 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15076 opts->x_flag_wrapv got set.
15077 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15078 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15079 POINTER_TYPE_OVERFLOW_UNDEFINED.
15080 * match.pd: Likewise in address comparison pattern.
15081 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15083 2018-01-15 Richard Biener <rguenther@suse.de>
15086 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15087 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15088 Reset type names to their identifier if their TYPE_DECL doesn't
15089 have linkage (and thus is used for ODR and devirt).
15090 (save_debug_info_for_decl): Remove.
15091 (save_debug_info_for_type): Likewise.
15092 (add_tree_to_fld_list): Adjust.
15093 * tree-pretty-print.c (dump_generic_node): Make dumping of
15094 type names more robust.
15096 2018-01-15 Richard Biener <rguenther@suse.de>
15098 * BASE-VER: Bump to 8.0.1.
15100 2018-01-14 Martin Sebor <msebor@redhat.com>
15103 * builtins.c (check_access): Avoid warning when the no-warning bit
15106 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15108 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15109 * ira-color (allocno_hard_regs_compare): Likewise.
15111 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15114 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15115 Use .pushsection/.popsection.
15117 2018-01-14 Martin Sebor <msebor@redhat.com>
15120 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15122 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15124 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15125 entry from extra_headers.
15126 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15127 extra_headers, make the list bitwise identical to the i?86-*-* one.
15129 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15131 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15132 -mcmodel=large with -mindirect-branch=thunk,
15133 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15134 -mfunction-return=thunk-extern.
15135 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15136 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15137 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15139 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15141 * config/i386/i386.c (print_reg): Print the name of the full
15142 integer register without '%'.
15143 (ix86_print_operand): Handle 'V'.
15144 * doc/extend.texi: Document 'V' modifier.
15146 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15148 * config/i386/constraints.md (Bs): Disallow memory operand for
15149 -mindirect-branch-register.
15151 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15152 (GOT_memory_operand): Likewise.
15153 (call_insn_operand): Likewise.
15154 (sibcall_insn_operand): Likewise.
15155 (GOT32_symbol_operand): Likewise.
15156 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15157 for -mindirect-branch-register.
15158 (tablejump): Likewise.
15159 (*sibcall_memory): Likewise.
15160 (*sibcall_value_memory): Likewise.
15161 Disallow peepholes of indirect call and jump via memory for
15162 -mindirect-branch-register.
15163 (*call_pop): Replace m with Bw.
15164 (*call_value_pop): Likewise.
15165 (*sibcall_pop_memory): Replace m with Bs.
15166 * config/i386/i386.opt (mindirect-branch-register): New option.
15167 * doc/invoke.texi: Document -mindirect-branch-register option.
15169 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15171 * config/i386/i386-protos.h (ix86_output_function_return): New.
15172 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15173 set function_return_type.
15174 (indirect_thunk_name): Add ret_p to indicate thunk for function
15176 (output_indirect_thunk_function): Pass false to
15177 indirect_thunk_name.
15178 (ix86_output_indirect_branch_via_reg): Likewise.
15179 (ix86_output_indirect_branch_via_push): Likewise.
15180 (output_indirect_thunk_function): Create alias for function
15181 return thunk if regno < 0.
15182 (ix86_output_function_return): New function.
15183 (ix86_handle_fndecl_attribute): Handle function_return.
15184 (ix86_attribute_table): Add function_return.
15185 * config/i386/i386.h (machine_function): Add
15186 function_return_type.
15187 * config/i386/i386.md (simple_return_internal): Use
15188 ix86_output_function_return.
15189 (simple_return_internal_long): Likewise.
15190 * config/i386/i386.opt (mfunction-return=): New option.
15191 (indirect_branch): Mention -mfunction-return=.
15192 * doc/extend.texi: Document function_return function attribute.
15193 * doc/invoke.texi: Document -mfunction-return= option.
15195 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15197 * config/i386/i386-opts.h (indirect_branch): New.
15198 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15199 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15200 with local indirect jump when converting indirect call and jump.
15201 (ix86_set_indirect_branch_type): New.
15202 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15203 (indirectlabelno): New.
15204 (indirect_thunk_needed): Likewise.
15205 (indirect_thunk_bnd_needed): Likewise.
15206 (indirect_thunks_used): Likewise.
15207 (indirect_thunks_bnd_used): Likewise.
15208 (INDIRECT_LABEL): Likewise.
15209 (indirect_thunk_name): Likewise.
15210 (output_indirect_thunk): Likewise.
15211 (output_indirect_thunk_function): Likewise.
15212 (ix86_output_indirect_branch_via_reg): Likewise.
15213 (ix86_output_indirect_branch_via_push): Likewise.
15214 (ix86_output_indirect_branch): Likewise.
15215 (ix86_output_indirect_jmp): Likewise.
15216 (ix86_code_end): Call output_indirect_thunk_function if needed.
15217 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15219 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15220 (ix86_attribute_table): Add indirect_branch.
15221 * config/i386/i386.h (machine_function): Add indirect_branch_type
15222 and has_local_indirect_jump.
15223 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15225 (tablejump): Likewise.
15226 (*indirect_jump): Use ix86_output_indirect_jmp.
15227 (*tablejump_1): Likewise.
15228 (simple_return_indirect_internal): Likewise.
15229 * config/i386/i386.opt (mindirect-branch=): New option.
15230 (indirect_branch): New.
15233 (thunk-inline): Likewise.
15234 (thunk-extern): Likewise.
15235 * doc/extend.texi: Document indirect_branch function attribute.
15236 * doc/invoke.texi: Document -mindirect-branch= option.
15238 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15241 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15243 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15245 * ipa-inline.c (want_inline_small_function_p): Return false if
15246 inlining has already failed with CIF_FINAL_ERROR.
15247 (update_caller_keys): Call want_inline_small_function_p before
15249 (update_callee_keys): Likewise.
15251 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15253 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15255 (rs6000_quadword_masked_address_p): Likewise.
15256 (quad_aligned_load_p): Likewise.
15257 (quad_aligned_store_p): Likewise.
15258 (const_load_sequence_p): Add comment to describe the outer-most loop.
15259 (mimic_memory_attributes_and_flags): New function.
15260 (rs6000_gen_stvx): Likewise.
15261 (replace_swapped_aligned_store): Likewise.
15262 (rs6000_gen_lvx): Likewise.
15263 (replace_swapped_aligned_load): Likewise.
15264 (replace_swapped_load_constant): Capitalize argument name in
15265 comment describing this function.
15266 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15267 and stores that access quad-word aligned addresses and replace
15268 with stvx or lvx instructions when appropriate.
15269 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15270 New function prototype.
15271 (rs6000_quadword_masked_address_p): Likewise.
15272 (rs6000_gen_lvx): Likewise.
15273 (rs6000_gen_stvx): Likewise.
15274 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15275 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15276 when memory address is aligned.
15277 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15278 this split to select lvx instruction when memory address is aligned.
15279 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15280 instruction when memory address is aligned.
15281 (*vsx_le_perm_load_v16qi): Likewise.
15282 (four unnamed splitters): Modify to select the stvx instruction
15283 when memory is aligned.
15285 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15287 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15288 which appears in the queue multiple times.
15290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15291 Alan Hayward <alan.hayward@arm.com>
15292 David Sherwood <david.sherwood@arm.com>
15294 * tree-vectorizer.h (vec_lower_bound): New structure.
15295 (_loop_vec_info): Add check_nonzero and lower_bounds.
15296 (LOOP_VINFO_CHECK_NONZERO): New macro.
15297 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15298 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15299 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15300 fields. Make seg_len the distance travelled, not including the
15302 (dr_direction_indicator): Declare.
15303 (dr_zero_step_indicator): Likewise.
15304 (dr_known_forward_stride_p): Likewise.
15305 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15307 (runtime_alias_check_p): Allow runtime alias checks with
15309 (operator ==): Compare access_size and align.
15310 (prune_runtime_alias_test_list): Rework for new distinction between
15311 the access_size and seg_len.
15312 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15314 (get_segment_min_max): New function.
15315 (create_intersect_range_checks): Use it.
15316 (dr_step_indicator): New function.
15317 (dr_direction_indicator): Likewise.
15318 (dr_zero_step_indicator): Likewise.
15319 (dr_known_forward_stride_p): Likewise.
15320 * tree-loop-distribution.c (data_ref_segment_size): Return
15321 DR_STEP * (niters - 1).
15322 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15324 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15325 (vect_preserves_scalar_order_p): New function, split out from...
15326 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15327 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15328 (vect_vfa_access_size): New function.
15329 (vect_vfa_align): Likewise.
15330 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15331 (dump_lower_bound): New function.
15332 (vect_check_lower_bound): Likewise.
15333 (vect_small_gap_p): Likewise.
15334 (vectorizable_with_step_bound_p): Likewise.
15335 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15336 depencies if the vectorization factor is 1. Convert the checks
15337 for nonzero steps into checks on the bounds of DR_STEP. Try using
15338 a bunds check for variable steps if the minimum required step is
15339 relatively small. Update calls to the dr_with_seg_len
15340 constructor and to vect_compile_time_alias.
15341 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15343 (vect_loop_versioning): Call it.
15344 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15346 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15349 Alan Hayward <alan.hayward@arm.com>
15350 David Sherwood <david.sherwood@arm.com>
15352 * doc/sourcebuild.texi (vect_scatter_store): Document.
15353 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15355 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15357 * genopinit.c (main): Add supports_vec_scatter_store and
15358 supports_vec_scatter_store_cached to target_optabs.
15359 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15360 IFN_MASK_SCATTER_STORE.
15361 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15363 * internal-fn.h (internal_store_fn_p): Declare.
15364 (internal_fn_stored_value_index): Likewise.
15365 * internal-fn.c (scatter_store_direct): New macro.
15366 (expand_scatter_store_optab_fn): New function.
15367 (direct_scatter_store_optab_supported_p): New macro.
15368 (internal_store_fn_p): New function.
15369 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15370 IFN_MASK_SCATTER_STORE.
15371 (internal_fn_mask_index): Likewise.
15372 (internal_fn_stored_value_index): New function.
15373 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15374 for scatter stores.
15375 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15376 * optabs-query.c (supports_vec_scatter_store_p): New function.
15377 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15378 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15379 true for scatter stores.
15380 (vect_gather_scatter_fn_p): Handle scatter stores too.
15381 (vect_check_gather_scatter): Consider using scatter stores if
15382 supports_vec_scatter_store_p.
15383 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15384 scatter stores too.
15385 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15386 internal_fn_stored_value_index.
15387 (check_load_store_masking): Handle scatter stores too.
15388 (vect_get_store_rhs): Make public.
15389 (vectorizable_call): Use internal_store_fn_p.
15390 (vectorizable_store): Handle scatter store internal functions.
15391 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15392 when deciding whether the end of the group has been reached.
15393 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15394 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15395 (mask_scatter_store<mode>): New insns.
15397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15398 Alan Hayward <alan.hayward@arm.com>
15399 David Sherwood <david.sherwood@arm.com>
15401 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15402 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15403 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15405 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15406 Use vect_truncate_gather_scatter_offset if we can't treat the
15407 operation as a normal gather load or scatter store.
15408 (get_group_load_store_type): Take the gather_scatter_info
15409 as argument. Try using a gather load or scatter store for
15410 single-element groups.
15411 (get_load_store_type): Update calls to get_group_load_store_type
15412 and vect_use_strided_gather_scatters_p.
15414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15415 Alan Hayward <alan.hayward@arm.com>
15416 David Sherwood <david.sherwood@arm.com>
15418 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15419 optional tree argument.
15420 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15422 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15423 but continue to use the current value as a fallback.
15424 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15425 to compare the updates.
15426 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15427 (get_load_store_type): Use it when handling a strided access.
15428 (vect_get_strided_load_store_ops): New function.
15429 (vect_get_data_ptr_increment): Likewise.
15430 (vectorizable_load): Handle strided gather loads. Always pass
15431 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15434 Alan Hayward <alan.hayward@arm.com>
15435 David Sherwood <david.sherwood@arm.com>
15437 * doc/md.texi (gather_load@var{m}): Document.
15438 (mask_gather_load@var{m}): Likewise.
15439 * genopinit.c (main): Add supports_vec_gather_load and
15440 supports_vec_gather_load_cached to target_optabs.
15441 * optabs-tree.c (init_tree_optimization_optabs): Use
15442 ggc_cleared_alloc to allocate target_optabs.
15443 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15444 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15446 * internal-fn.h (internal_load_fn_p): Declare.
15447 (internal_gather_scatter_fn_p): Likewise.
15448 (internal_fn_mask_index): Likewise.
15449 (internal_gather_scatter_fn_supported_p): Likewise.
15450 * internal-fn.c (gather_load_direct): New macro.
15451 (expand_gather_load_optab_fn): New function.
15452 (direct_gather_load_optab_supported_p): New macro.
15453 (direct_internal_fn_optab): New function.
15454 (internal_load_fn_p): Likewise.
15455 (internal_gather_scatter_fn_p): Likewise.
15456 (internal_fn_mask_index): Likewise.
15457 (internal_gather_scatter_fn_supported_p): Likewise.
15458 * optabs-query.c (supports_at_least_one_mode_p): New function.
15459 (supports_vec_gather_load_p): Likewise.
15460 * optabs-query.h (supports_vec_gather_load_p): Declare.
15461 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15462 and memory_type field.
15463 (NUM_PATTERNS): Bump to 15.
15464 * tree-vect-data-refs.c: Include internal-fn.h.
15465 (vect_gather_scatter_fn_p): New function.
15466 (vect_describe_gather_scatter_call): Likewise.
15467 (vect_check_gather_scatter): Try using internal functions for
15468 gather loads. Recognize existing calls to a gather load function.
15469 (vect_analyze_data_refs): Consider using gather loads if
15470 supports_vec_gather_load_p.
15471 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15472 (vect_get_gather_scatter_offset_type): Likewise.
15473 (vect_convert_mask_for_vectype): Likewise.
15474 (vect_add_conversion_to_patterm): Likewise.
15475 (vect_try_gather_scatter_pattern): Likewise.
15476 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15477 (vect_vect_recog_func_ptrs): Add it.
15478 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15479 internal_fn_mask_index and internal_gather_scatter_fn_p.
15480 (check_load_store_masking): Take the gather_scatter_info as an
15481 argument and handle gather loads.
15482 (vect_get_gather_scatter_ops): New function.
15483 (vectorizable_call): Check internal_load_fn_p.
15484 (vectorizable_load): Likewise. Handle gather load internal
15486 (vectorizable_store): Update call to check_load_store_masking.
15487 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15488 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15489 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15490 (aarch64_gather_scale_operand_d): New predicates.
15491 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15492 (mask_gather_load<mode>): New insns.
15494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15495 Alan Hayward <alan.hayward@arm.com>
15496 David Sherwood <david.sherwood@arm.com>
15498 * optabs.def (fold_left_plus_optab): New optab.
15499 * doc/md.texi (fold_left_plus_@var{m}): Document.
15500 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15501 * internal-fn.c (fold_left_direct): Define.
15502 (expand_fold_left_optab_fn): Likewise.
15503 (direct_fold_left_optab_supported_p): Likewise.
15504 * fold-const-call.c (fold_const_fold_left): New function.
15505 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15506 * tree-parloops.c (valid_reduction_p): New function.
15507 (gather_scalar_reductions): Use it.
15508 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15509 (vect_finish_replace_stmt): Declare.
15510 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15511 (needs_fold_left_reduction_p): New function, split out from...
15512 (vect_is_simple_reduction): ...here. Accept reductions that
15513 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15514 (vect_force_simple_reduction): Also store the reduction type in
15515 the assignment's STMT_VINFO_REDUC_TYPE.
15516 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15517 (merge_with_identity): New function.
15518 (vect_expand_fold_left): Likewise.
15519 (vectorize_fold_left_reduction): Likewise.
15520 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15521 scalar phi in place for it. Check for target support and reject
15522 cases that would reassociate the operation. Defer the transform
15523 phase to vectorize_fold_left_reduction.
15524 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15525 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15526 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15530 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15531 call to ifc_temp_var.
15533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15534 Alan Hayward <alan.hayward@arm.com>
15535 David Sherwood <david.sherwood@arm.com>
15537 * target.def (legitimize_address_displacement): Take the original
15538 offset as a poly_int.
15539 * targhooks.h (default_legitimize_address_displacement): Update
15541 * targhooks.c (default_legitimize_address_displacement): Likewise.
15542 * doc/tm.texi: Regenerate.
15543 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15544 as an argument, moving assert of ad->disp == ad->disp_term to...
15545 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15546 Try calling targetm.legitimize_address_displacement before expanding
15547 the address rather than afterwards, and adjust for the new interface.
15548 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15549 Match the new hook interface. Handle SVE addresses.
15550 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15551 new hook interface.
15553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15555 * Makefile.in (OBJS): Add early-remat.o.
15556 * target.def (select_early_remat_modes): New hook.
15557 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15558 * doc/tm.texi: Regenerate.
15559 * targhooks.h (default_select_early_remat_modes): Declare.
15560 * targhooks.c (default_select_early_remat_modes): New function.
15561 * timevar.def (TV_EARLY_REMAT): New timevar.
15562 * passes.def (pass_early_remat): New pass.
15563 * tree-pass.h (make_pass_early_remat): Declare.
15564 * early-remat.c: New file.
15565 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15567 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15570 Alan Hayward <alan.hayward@arm.com>
15571 David Sherwood <david.sherwood@arm.com>
15573 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15574 vfm1 with a bound_epilog parameter.
15575 (vect_do_peeling): Update calls accordingly, and move the prologue
15576 call earlier in the function. Treat the base bound_epilog as 0 for
15577 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15578 this base when peeling for gaps.
15579 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15580 with fully-masked loops.
15581 (vect_estimate_min_profitable_iters): Handle the single peeled
15582 iteration in that case.
15584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15585 Alan Hayward <alan.hayward@arm.com>
15586 David Sherwood <david.sherwood@arm.com>
15588 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15589 single-element interleaving even if the size is not a power of 2.
15590 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15591 accesses for single-element interleaving if the group size is
15594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15595 Alan Hayward <alan.hayward@arm.com>
15596 David Sherwood <david.sherwood@arm.com>
15598 * doc/md.texi (fold_extract_last_@var{m}): Document.
15599 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15600 * optabs.def (fold_extract_last_optab): New optab.
15601 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15602 * internal-fn.c (fold_extract_direct): New macro.
15603 (expand_fold_extract_optab_fn): Likewise.
15604 (direct_fold_extract_optab_supported_p): Likewise.
15605 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15606 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15607 EXTRACT_LAST_REDUCTION.
15608 (get_initial_def_for_reduction): Do not create an initial vector
15609 for EXTRACT_LAST_REDUCTION reductions.
15610 (vectorizable_reduction): Leave the scalar phi in place for
15611 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15612 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15613 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15614 transform phase to vectorizable_condition.
15615 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15617 (vect_finish_stmt_generation): ...here.
15618 (vect_finish_replace_stmt): New function.
15619 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15620 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15622 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15625 Alan Hayward <alan.hayward@arm.com>
15626 David Sherwood <david.sherwood@arm.com>
15628 * doc/md.texi (extract_last_@var{m}): Document.
15629 * optabs.def (extract_last_optab): New optab.
15630 * internal-fn.def (EXTRACT_LAST): New internal function.
15631 * internal-fn.c (cond_unary_direct): New macro.
15632 (expand_cond_unary_optab_fn): Likewise.
15633 (direct_cond_unary_optab_supported_p): Likewise.
15634 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15635 loops using EXTRACT_LAST.
15636 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15637 (extract_last_<mode>): ...this optab.
15638 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
15645 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15646 * doc/tm.texi: Regenerate.
15647 * targhooks.h (default_empty_mask_is_expensive): Declare.
15648 * targhooks.c (default_empty_mask_is_expensive): New function.
15649 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15650 if the target says that empty masks are expensive.
15651 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15653 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15656 Alan Hayward <alan.hayward@arm.com>
15657 David Sherwood <david.sherwood@arm.com>
15659 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15660 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15661 (vect_use_loop_mask_for_alignment_p): New function.
15662 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15663 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15664 niters_skip argument. Make sure that the first niters_skip elements
15665 of the first iteration are inactive.
15666 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15667 Update call to vect_set_loop_masks_directly.
15668 (get_misalign_in_elems): New function, split out from...
15669 (vect_gen_prolog_loop_niters): ...here.
15670 (vect_update_init_of_dr): Take a code argument that specifies whether
15671 the adjustment should be added or subtracted.
15672 (vect_update_init_of_drs): Likewise.
15673 (vect_prepare_for_masked_peels): New function.
15674 (vect_do_peeling): Skip prologue peeling if we're using a mask
15675 instead. Update call to vect_update_inits_of_drs.
15676 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15678 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15679 alignment. Do not include the number of peeled iterations in
15680 the minimum threshold in that case.
15681 (vectorizable_induction): Adjust the start value down by
15682 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15683 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15684 Take the number of skipped iterations into account when calculating
15686 * tree-vect-stmts.c (vect_gen_while_not): New function.
15688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15689 Alan Hayward <alan.hayward@arm.com>
15690 David Sherwood <david.sherwood@arm.com>
15692 * doc/sourcebuild.texi (vect_fully_masked): Document.
15693 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15694 default value to 0.
15695 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15697 (vect_analyze_loop_2): ...here. Don't check the vectorization
15698 factor against the number of loop iterations if the loop is
15701 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15702 Alan Hayward <alan.hayward@arm.com>
15703 David Sherwood <david.sherwood@arm.com>
15705 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15706 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15707 (dump_groups): Update accordingly.
15708 (iv_use::mem_type): New member variable.
15709 (address_p): New function.
15710 (record_use): Add a mem_type argument and initialize the new
15712 (record_group_use): Add a mem_type argument. Use address_p.
15713 Remove obsolete null checks of base_object. Update call to record_use.
15714 (find_interesting_uses_op): Update call to record_group_use.
15715 (find_interesting_uses_cond): Likewise.
15716 (find_interesting_uses_address): Likewise.
15717 (get_mem_type_for_internal_fn): New function.
15718 (find_address_like_use): Likewise.
15719 (find_interesting_uses_stmt): Try find_address_like_use before
15720 calling find_interesting_uses_op.
15721 (addr_offset_valid_p): Use the iv mem_type field as the type
15722 of the addressed memory.
15723 (add_autoinc_candidates): Likewise.
15724 (get_address_cost): Likewise.
15725 (split_small_address_groups_p): Use address_p.
15726 (split_address_groups): Likewise.
15727 (add_iv_candidate_for_use): Likewise.
15728 (autoinc_possible_for_pair): Likewise.
15729 (rewrite_groups): Likewise.
15730 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15731 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15732 (get_alias_ptr_type_for_ptr_address): New function.
15733 (rewrite_use_address): Rewrite address uses in calls that were
15734 identified by find_address_like_use.
15736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15737 Alan Hayward <alan.hayward@arm.com>
15738 David Sherwood <david.sherwood@arm.com>
15740 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15742 * gimple-expr.h (is_gimple_addressable: Likewise.
15743 * gimple-expr.c (is_gimple_address): Likewise.
15744 * internal-fn.c (expand_call_mem_ref): New function.
15745 (expand_mask_load_optab_fn): Use it.
15746 (expand_mask_store_optab_fn): Likewise.
15748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15749 Alan Hayward <alan.hayward@arm.com>
15750 David Sherwood <david.sherwood@arm.com>
15752 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15753 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15754 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15755 (cond_umax@var{mode}): Document.
15756 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15757 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15758 (cond_umin_optab, cond_umax_optab): New optabs.
15759 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15760 (COND_IOR, COND_XOR): New internal functions.
15761 * internal-fn.h (get_conditional_internal_fn): Declare.
15762 * internal-fn.c (cond_binary_direct): New macro.
15763 (expand_cond_binary_optab_fn): Likewise.
15764 (direct_cond_binary_optab_supported_p): Likewise.
15765 (get_conditional_internal_fn): New function.
15766 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15767 Cope with reduction statements that are vectorized as calls rather
15769 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15770 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15771 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15772 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15773 (UNSPEC_COND_EOR): New unspecs.
15774 (optab): Add mappings for them.
15775 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15776 (sve_int_op, sve_fp_op): New int attributes.
15778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15779 Alan Hayward <alan.hayward@arm.com>
15780 David Sherwood <david.sherwood@arm.com>
15782 * optabs.def (while_ult_optab): New optab.
15783 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15784 * internal-fn.def (WHILE_ULT): New internal function.
15785 * internal-fn.h (direct_internal_fn_supported_p): New override
15786 that takes two types as argument.
15787 * internal-fn.c (while_direct): New macro.
15788 (expand_while_optab_fn): New function.
15789 (convert_optab_supported_p): Likewise.
15790 (direct_while_optab_supported_p): New macro.
15791 * wide-int.h (wi::udiv_ceil): New function.
15792 * tree-vectorizer.h (rgroup_masks): New structure.
15793 (vec_loop_masks): New typedef.
15794 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15795 and fully_masked_p.
15796 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15797 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15798 (vect_max_vf): New function.
15799 (slpeel_make_loop_iterate_ntimes): Delete.
15800 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15801 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15802 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15803 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15804 internal-fn.h, stor-layout.h and optabs-query.h.
15805 (vect_set_loop_mask): New function.
15806 (add_preheader_seq): Likewise.
15807 (add_header_seq): Likewise.
15808 (interleave_supported_p): Likewise.
15809 (vect_maybe_permute_loop_masks): Likewise.
15810 (vect_set_loop_masks_directly): Likewise.
15811 (vect_set_loop_condition_masked): Likewise.
15812 (vect_set_loop_condition_unmasked): New function, split out from
15813 slpeel_make_loop_iterate_ntimes.
15814 (slpeel_make_loop_iterate_ntimes): Rename to..
15815 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15816 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15817 (vect_do_peeling): Update call accordingly.
15818 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15820 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15821 mask_compare_type, can_fully_mask_p and fully_masked_p.
15822 (release_vec_loop_masks): New function.
15823 (_loop_vec_info): Use it to free the loop masks.
15824 (can_produce_all_loop_masks_p): New function.
15825 (vect_get_max_nscalars_per_iter): Likewise.
15826 (vect_verify_full_masking): Likewise.
15827 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15828 retries, and free the mask rgroups before retrying. Check loop-wide
15829 reasons for disallowing fully-masked loops. Make the final decision
15830 about whether use a fully-masked loop or not.
15831 (vect_estimate_min_profitable_iters): Do not assume that peeling
15832 for the number of iterations will be needed for fully-masked loops.
15833 (vectorizable_reduction): Disable fully-masked loops.
15834 (vectorizable_live_operation): Likewise.
15835 (vect_halve_mask_nunits): New function.
15836 (vect_double_mask_nunits): Likewise.
15837 (vect_record_loop_mask): Likewise.
15838 (vect_get_loop_mask): Likewise.
15839 (vect_transform_loop): Handle the case in which the final loop
15840 iteration might handle a partial vector. Call vect_set_loop_condition
15841 instead of slpeel_make_loop_iterate_ntimes.
15842 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15843 (check_load_store_masking): New function.
15844 (prepare_load_store_mask): Likewise.
15845 (vectorizable_store): Handle fully-masked loops.
15846 (vectorizable_load): Likewise.
15847 (supportable_widening_operation): Use vect_halve_mask_nunits for
15849 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15850 (vect_gen_while): New function.
15851 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15852 (aarch64_uqdec<mode>): New insn.
15854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15855 Alan Hayward <alan.hayward@arm.com>
15856 David Sherwood <david.sherwood@arm.com>
15858 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15859 (reduc_xor_scal_optab): New optabs.
15860 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15861 (reduc_xor_scal_@var{m}): Document.
15862 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15863 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15864 internal functions.
15865 * fold-const-call.c (fold_const_call): Handle them.
15866 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15867 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15868 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15869 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15870 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15871 (UNSPEC_XORV): New unspecs.
15872 (optab): Add entries for them.
15873 (BITWISEV): New int iterator.
15874 (bit_reduc_op): New int attributes.
15876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15877 Alan Hayward <alan.hayward@arm.com>
15878 David Sherwood <david.sherwood@arm.com>
15880 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15881 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15882 * optabs.def (vec_shl_insert_optab): New optab.
15883 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15884 (duplicate_and_interleave): Likewise.
15885 * tree-vect-loop.c: Include internal-fn.h.
15886 (neutral_op_for_slp_reduction): New function, split out from
15887 get_initial_defs_for_reduction.
15888 (get_initial_def_for_reduction): Handle option 2 for variable-length
15889 vectors by loading the neutral value into a vector and then shifting
15890 the initial value into element 0.
15891 (get_initial_defs_for_reduction): Replace the code argument with
15892 the neutral value calculated by neutral_op_for_slp_reduction.
15893 Use gimple_build_vector for constant-length vectors.
15894 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15895 but the first group_size elements have a neutral value.
15896 Use duplicate_and_interleave otherwise.
15897 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15898 Update call to get_initial_defs_for_reduction. Handle SLP
15899 reductions for variable-length vectors by creating one vector
15900 result for each scalar result, with the elements associated
15901 with other scalar results stubbed out with the neutral value.
15902 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15903 Require IFN_VEC_SHL_INSERT for double reductions on
15904 variable-length vectors, or SLP reductions that have
15905 a neutral value. Require can_duplicate_and_interleave_p
15906 support for variable-length unchained SLP reductions if there
15907 is no neutral value, such as for MIN/MAX reductions. Also require
15908 the number of vector elements to be a multiple of the number of
15909 SLP statements when doing variable-length unchained SLP reductions.
15910 Update call to vect_create_epilog_for_reduction.
15911 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15912 and remove initial values.
15913 (duplicate_and_interleave): Make public.
15914 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15915 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15918 Alan Hayward <alan.hayward@arm.com>
15919 David Sherwood <david.sherwood@arm.com>
15921 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15922 (can_duplicate_and_interleave_p): New function.
15923 (vect_get_and_check_slp_defs): Take the vector of statements
15924 rather than just the current one. Remove excess parentheses.
15925 Restriction rejectinon of vect_constant_def and vect_external_def
15926 for variable-length vectors to boolean types, or types for which
15927 can_duplicate_and_interleave_p is false.
15928 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15929 (duplicate_and_interleave): New function.
15930 (vect_get_constant_vectors): Use gimple_build_vector for
15931 constant-length vectors and suitable variable-length constant
15932 vectors. Use duplicate_and_interleave for other variable-length
15933 vectors. Don't defer the update when inserting new statements.
15935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15936 Alan Hayward <alan.hayward@arm.com>
15937 David Sherwood <david.sherwood@arm.com>
15939 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15940 min_profitable_iters doesn't go negative.
15942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15943 Alan Hayward <alan.hayward@arm.com>
15944 David Sherwood <david.sherwood@arm.com>
15946 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15947 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15948 * optabs.def (vec_mask_load_lanes_optab): New optab.
15949 (vec_mask_store_lanes_optab): Likewise.
15950 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15951 (MASK_STORE_LANES): Likewise.
15952 * internal-fn.c (mask_load_lanes_direct): New macro.
15953 (mask_store_lanes_direct): Likewise.
15954 (expand_mask_load_optab_fn): Handle masked operations.
15955 (expand_mask_load_lanes_optab_fn): New macro.
15956 (expand_mask_store_optab_fn): Handle masked operations.
15957 (expand_mask_store_lanes_optab_fn): New macro.
15958 (direct_mask_load_lanes_optab_supported_p): Likewise.
15959 (direct_mask_store_lanes_optab_supported_p): Likewise.
15960 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15962 (vect_load_lanes_supported): Likewise.
15963 * tree-vect-data-refs.c (strip_conversion): New function.
15964 (can_group_stmts_p): Likewise.
15965 (vect_analyze_data_ref_accesses): Use it instead of checking
15966 for a pair of assignments.
15967 (vect_store_lanes_supported): Take a masked_p parameter.
15968 (vect_load_lanes_supported): Likewise.
15969 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15970 vect_store_lanes_supported and vect_load_lanes_supported.
15971 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15972 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15973 parameter. Don't allow gaps for masked accesses.
15974 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15975 and vect_load_lanes_supported.
15976 (get_load_store_type): Take a masked_p parameter and update
15977 call to get_group_load_store_type.
15978 (vectorizable_store): Update call to get_load_store_type.
15979 Handle IFN_MASK_STORE_LANES.
15980 (vectorizable_load): Update call to get_load_store_type.
15981 Handle IFN_MASK_LOAD_LANES.
15983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15984 Alan Hayward <alan.hayward@arm.com>
15985 David Sherwood <david.sherwood@arm.com>
15987 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15989 * config/aarch64/aarch64-protos.h
15990 (aarch64_sve_struct_memory_operand_p): Declare.
15991 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15992 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15993 (VPRED, vpred): Handle SVE structure modes.
15994 * config/aarch64/constraints.md (Utx): New constraint.
15995 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15996 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15997 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15998 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15999 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16000 structure modes. Split into pieces after RA.
16001 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16002 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16004 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16005 SVE structure modes.
16006 (aarch64_classify_address): Likewise.
16007 (sizetochar): Move earlier in file.
16008 (aarch64_print_operand): Handle SVE register lists.
16009 (aarch64_array_mode): New function.
16010 (aarch64_sve_struct_memory_operand_p): Likewise.
16011 (TARGET_ARRAY_MODE): Redefine.
16013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16014 Alan Hayward <alan.hayward@arm.com>
16015 David Sherwood <david.sherwood@arm.com>
16017 * target.def (array_mode): New target hook.
16018 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16019 * doc/tm.texi: Regenerate.
16020 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16021 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16022 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16023 targetm.array_mode.
16024 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16028 Alan Hayward <alan.hayward@arm.com>
16029 David Sherwood <david.sherwood@arm.com>
16031 * fold-const.c (fold_binary_loc): Check the argument types
16032 rather than the result type when testing for a vector operation.
16034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16036 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16037 * doc/tm.texi: Regenerate.
16039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16040 Alan Hayward <alan.hayward@arm.com>
16041 David Sherwood <david.sherwood@arm.com>
16043 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16044 (sve): Document new AArch64 extension.
16045 * doc/md.texi (w): Extend the description of the AArch64
16046 constraint to include SVE vectors.
16047 (Upl, Upa): Document new AArch64 predicate constraints.
16048 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16050 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16051 (msve-vector-bits=): New option.
16052 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16053 SVE when these are disabled.
16054 (sve): New extension.
16055 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16056 modes. Adjust their number of units based on aarch64_sve_vg.
16057 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16058 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16059 aarch64_addr_query_type.
16060 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16061 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16062 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16063 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16064 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16065 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16066 (aarch64_simd_imm_zero_p): Delete.
16067 (aarch64_check_zero_based_sve_index_immediate): Declare.
16068 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16069 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16070 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16071 (aarch64_sve_float_mul_immediate_p): Likewise.
16072 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16073 rather than an rtx.
16074 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16075 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16076 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16077 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16078 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16079 (aarch64_regmode_natural_size): Likewise.
16080 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16081 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16083 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16084 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16085 for VG and the SVE predicate registers.
16086 (V_ALIASES): Add a "z"-prefixed alias.
16087 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16088 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16089 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16090 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16091 (REG_CLASS_NAMES): Add entries for them.
16092 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16093 and the predicate registers.
16094 (aarch64_sve_vg): Declare.
16095 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16096 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16097 (REGMODE_NATURAL_SIZE): Define.
16098 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16100 * config/aarch64/aarch64.c: Include cfgrtl.h.
16101 (simd_immediate_info): Add a constructor for series vectors,
16102 and an associated step field.
16103 (aarch64_sve_vg): New variable.
16104 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16105 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16106 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16107 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16108 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16109 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16110 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16111 (aarch64_get_mask_mode): New functions.
16112 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16113 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16114 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16115 predicate modes and predicate registers. Explicitly restrict
16116 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16117 to store a vector mode if it is recognized by
16118 aarch64_classify_vector_mode.
16119 (aarch64_regmode_natural_size): New function.
16120 (aarch64_hard_regno_caller_save_mode): Return the original mode
16122 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16123 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16124 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16125 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16127 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16128 does not overlap dest if the function is frame-related. Handle
16130 (aarch64_split_add_offset): New function.
16131 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16132 them aarch64_add_offset.
16133 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16134 and update call to aarch64_sub_sp.
16135 (aarch64_add_cfa_expression): New function.
16136 (aarch64_expand_prologue): Pass extra temporary registers to the
16137 functions above. Handle the case in which we need to emit new
16138 DW_CFA_expressions for registers that were originally saved
16139 relative to the stack pointer, but now have to be expressed
16140 relative to the frame pointer.
16141 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16143 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16144 IP0 and IP1 values for SVE frames.
16145 (aarch64_expand_vec_series): New function.
16146 (aarch64_expand_sve_widened_duplicate): Likewise.
16147 (aarch64_expand_sve_const_vector): Likewise.
16148 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16149 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16150 into the register, rather than emitting a SET directly.
16151 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16152 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16153 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16154 (offset_9bit_signed_scaled_p): New functions.
16155 (aarch64_replicate_bitmask_imm): New function.
16156 (aarch64_bitmask_imm): Use it.
16157 (aarch64_cannot_force_const_mem): Reject expressions involving
16158 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16159 (aarch64_classify_index): Handle SVE indices, by requiring
16160 a plain register index with a scale that matches the element size.
16161 (aarch64_classify_address): Handle SVE addresses. Assert that
16162 the mode of the address is VOIDmode or an integer mode.
16163 Update call to aarch64_classify_symbol.
16164 (aarch64_classify_symbolic_expression): Update call to
16165 aarch64_classify_symbol.
16166 (aarch64_const_vec_all_in_range_p): New function.
16167 (aarch64_print_vector_float_operand): Likewise.
16168 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16169 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16170 and the FP immediates 1.0 and 0.5.
16171 (aarch64_print_address_internal): Handle SVE addresses.
16172 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16173 (aarch64_regno_regclass): Handle predicate registers.
16174 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16176 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16177 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16178 (aarch64_convert_sve_vector_bits): New function.
16179 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16180 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16181 rather than an rtx.
16182 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16183 Handle SVE vector and predicate modes. Accept VL-based constants
16184 that need only one temporary register, and VL offsets that require
16185 no temporary registers.
16186 (aarch64_conditional_register_usage): Mark the predicate registers
16187 as fixed if SVE isn't available.
16188 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16189 Return true for SVE vector and predicate modes.
16190 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16191 rather than an unsigned int. Handle SVE modes.
16192 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16194 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16196 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16197 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16198 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16199 (aarch64_sve_float_mul_immediate_p): New functions.
16200 (aarch64_sve_valid_immediate): New function.
16201 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16202 Explicitly reject structure modes. Check for INDEX constants.
16203 Handle PTRUE and PFALSE constants.
16204 (aarch64_check_zero_based_sve_index_immediate): New function.
16205 (aarch64_simd_imm_zero_p): Delete.
16206 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16207 vector modes. Accept constants in the range of CNT[BHWD].
16208 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16209 ask for an Advanced SIMD mode.
16210 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16211 (aarch64_simd_vector_alignment): Handle SVE predicates.
16212 (aarch64_vectorize_preferred_vector_alignment): New function.
16213 (aarch64_simd_vector_alignment_reachable): Use it instead of
16215 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16216 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16218 (MAX_VECT_LEN): Delete.
16219 (expand_vec_perm_d): Add a vec_flags field.
16220 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16221 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16222 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16224 (aarch64_evpc_rev): Rename to...
16225 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16226 (aarch64_evpc_rev_global): New function.
16227 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16228 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16230 (aarch64_evpc_sve_tbl): New function.
16231 (aarch64_expand_vec_perm_const_1): Update after rename of
16232 aarch64_evpc_rev. Handle SVE permutes too, trying
16233 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16234 than aarch64_evpc_tbl.
16235 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16236 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16237 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16238 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16239 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16240 (aarch64_expand_sve_vcond): New functions.
16241 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16242 of aarch64_vector_mode_p.
16243 (aarch64_dwarf_poly_indeterminate_value): New function.
16244 (aarch64_compute_pressure_classes): Likewise.
16245 (aarch64_can_change_mode_class): Likewise.
16246 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16247 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16248 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16249 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16250 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16251 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16252 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16253 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16255 (Dn, Dl, Dr): Accept const as well as const_vector.
16256 (Dz): Likewise. Compare against CONST0_RTX.
16257 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16258 of "vector" where appropriate.
16259 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16260 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16261 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16262 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16263 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16264 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16265 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16266 (v_int_equiv): Extend to SVE modes.
16267 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16269 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16270 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16271 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16272 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16273 (SVE_COND_FP_CMP): New int iterators.
16274 (perm_hilo): Handle the new unpack unspecs.
16275 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16277 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16278 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16279 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16280 (aarch64_equality_operator, aarch64_constant_vector_operand)
16281 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16282 (aarch64_sve_nonimmediate_operand): Likewise.
16283 (aarch64_sve_general_operand): Likewise.
16284 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16285 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16286 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16287 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16288 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16289 (aarch64_sve_float_arith_immediate): Likewise.
16290 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16291 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16292 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16293 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16294 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16295 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16296 (aarch64_sve_float_arith_operand): Likewise.
16297 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16298 (aarch64_sve_float_mul_operand): Likewise.
16299 (aarch64_sve_vec_perm_operand): Likewise.
16300 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16301 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16302 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16303 as well as const_vector.
16304 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16305 in file. Use CONST0_RTX and CONSTM1_RTX.
16306 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16307 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16308 Use aarch64_simd_imm_zero.
16309 * config/aarch64/aarch64-sve.md: New file.
16310 * config/aarch64/aarch64.md: Include it.
16311 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16312 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16313 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16314 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16315 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16316 (sve): New attribute.
16317 (enabled): Disable instructions with the sve attribute unless
16319 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16320 aarch64_expand_mov_immediate.
16321 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16322 CNT[BHSD] immediates.
16323 (movti): Split CONST_POLY_INT moves into two halves.
16324 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16325 Split additions that need a temporary here if the destination
16326 is the stack pointer.
16327 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16328 (*add<mode>3_poly_1): New instruction.
16329 (set_clobber_cc): New expander.
16331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16333 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16334 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16335 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16336 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16337 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16338 Change innermode from fixed_mode_size to machine_mode.
16339 (simplify_subreg): Update call accordingly. Handle a constant-sized
16340 subreg of a variable-length CONST_VECTOR.
16342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16343 Alan Hayward <alan.hayward@arm.com>
16344 David Sherwood <david.sherwood@arm.com>
16346 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16347 (add_offset_to_base): New function, split out from...
16348 (create_mem_ref): ...here. When handling a scale other than 1,
16349 check first whether the address is valid without the offset.
16350 Add it into the base if so, leaving the index and scale as-is.
16352 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16355 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16356 fold_for_warn before checking if arg2 is INTEGER_CST.
16358 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16360 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16361 (store_multiple_operation): Delete.
16362 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16363 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16364 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16365 guarded by TARGET_STRING.
16366 (rs6000_output_load_multiple): Delete.
16367 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16368 OPTION_MASK_STRING / TARGET_STRING handling.
16369 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16370 (const rs6000_opt_masks) <"string">: Change mask to 0.
16371 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16372 (MASK_STRING): Delete.
16373 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16375 (load_multiple): Delete.
16382 (store_multiple): Delete.
16389 (movmemsi_8reg): Delete.
16390 (corresponding unnamed define_insn): Delete.
16391 (movmemsi_6reg): Delete.
16392 (corresponding unnamed define_insn): Delete.
16393 (movmemsi_4reg): Delete.
16394 (corresponding unnamed define_insn): Delete.
16395 (movmemsi_2reg): Delete.
16396 (corresponding unnamed define_insn): Delete.
16397 (movmemsi_1reg): Delete.
16398 (corresponding unnamed define_insn): Delete.
16399 * config/rs6000/rs6000.opt (mno-string): New.
16400 (mstring): Replace by deprecation warning stub.
16401 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16403 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16405 * regrename.c (regrename_do_replace): If replacing the same
16406 reg multiple times, try to reuse last created gen_raw_REG.
16409 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16410 main to workaround a bug in GDB.
16412 2018-01-12 Tom de Vries <tom@codesourcery.com>
16415 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16417 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16419 PR rtl-optimization/80481
16420 * ira-color.c (get_cap_member): New function.
16421 (allocnos_conflict_by_live_ranges_p): Use it.
16422 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16423 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16425 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16428 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16429 (*saddl_se_1): Ditto.
16430 (*ssubsi_1): Ditto.
16431 (*ssubl_se_1): Ditto.
16433 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16435 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16436 rather than wi::to_widest for DR_INITs.
16437 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16438 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16439 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16441 (vect_analyze_group_access_1): Note that here.
16443 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16445 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16446 polynomial type sizes.
16448 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16450 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16451 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16452 (gimple_add_tmp_var): Likewise.
16454 2018-01-12 Martin Liska <mliska@suse.cz>
16456 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16457 (gimple_alloc_sizes): Likewise.
16458 (dump_gimple_statistics): Use PRIu64 in printf format.
16459 * gimple.h: Change uint64_t to int.
16461 2018-01-12 Martin Liska <mliska@suse.cz>
16463 * tree-core.h: Use uint64_t instead of int.
16464 * tree.c (tree_node_counts): Likewise.
16465 (tree_node_sizes): Likewise.
16466 (dump_tree_statistics): Use PRIu64 in printf format.
16468 2018-01-12 Martin Liska <mliska@suse.cz>
16470 * Makefile.in: As qsort_chk is implemented in vec.c, add
16471 vec.o to linkage of gencfn-macros.
16472 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16473 passing the info to record_node_allocation_statistics.
16474 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16476 * ggc-common.c (struct ggc_usage): Add operator== and use
16477 it in operator< and compare function.
16478 * mem-stats.h (struct mem_usage): Likewise.
16479 * vec.c (struct vec_usage): Remove operator< and compare
16480 function. Can be simply inherited.
16482 2018-01-12 Martin Jambor <mjambor@suse.cz>
16485 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16486 * tree-ssa-math-opts.c: Include domwalk.h.
16487 (convert_mult_to_fma_1): New function.
16488 (fma_transformation_info): New type.
16489 (fma_deferring_state): Likewise.
16490 (cancel_fma_deferring): New function.
16491 (result_of_phi): Likewise.
16492 (last_fma_candidate_feeds_initial_phi): Likewise.
16493 (convert_mult_to_fma): Added deferring logic, split actual
16494 transformation to convert_mult_to_fma_1.
16495 (math_opts_dom_walker): New type.
16496 (math_opts_dom_walker::after_dom_children): New method, body moved
16497 here from pass_optimize_widening_mul::execute, added deferring logic
16499 (pass_optimize_widening_mul::execute): Moved most of code to
16500 math_opts_dom_walker::after_dom_children.
16501 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16502 * config/i386/i386.c (ix86_option_override_internal): Added
16503 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16505 2018-01-12 Richard Biener <rguenther@suse.de>
16508 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16509 inline instance vars.
16511 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16514 * config/rx/rx.c (rx_is_restricted_memory_address):
16515 Handle SUBREG case.
16517 2018-01-12 Richard Biener <rguenther@suse.de>
16519 PR tree-optimization/80846
16520 * target.def (split_reduction): New target hook.
16521 * targhooks.c (default_split_reduction): New function.
16522 * targhooks.h (default_split_reduction): Declare.
16523 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16524 target requests first reduce vectors by combining low and high
16526 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16527 (get_vectype_for_scalar_type_and_size): Export.
16528 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16529 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16530 * doc/tm.texi: Regenerate.
16531 * config/i386/i386.c (ix86_split_reduction): Implement
16532 TARGET_VECTORIZE_SPLIT_REDUCTION.
16534 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16537 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16538 in PIC mode except for TARGET_VXWORKS_RTP.
16539 * config/sparc/sparc.c: Include cfgrtl.h.
16540 (TARGET_INIT_PIC_REG): Define.
16541 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16542 (sparc_pic_register_p): New predicate.
16543 (sparc_legitimate_address_p): Use it.
16544 (sparc_legitimize_pic_address): Likewise.
16545 (sparc_delegitimize_address): Likewise.
16546 (sparc_mode_dependent_address_p): Likewise.
16547 (gen_load_pcrel_sym): Remove 4th parameter.
16548 (load_got_register): Adjust call to above. Remove obsolete stuff.
16549 (sparc_expand_prologue): Do not call load_got_register here.
16550 (sparc_flat_expand_prologue): Likewise.
16551 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16552 (sparc_use_pseudo_pic_reg): New function.
16553 (sparc_init_pic_reg): Likewise.
16554 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16555 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16557 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16559 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16560 Add item for branch_cost.
16562 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16564 PR rtl-optimization/83565
16565 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16566 not extend the result to a larger mode for rotate operations.
16567 (num_sign_bit_copies1): Likewise.
16569 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16572 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16574 Use values-Xc.o for -pedantic.
16575 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16577 2018-01-12 Martin Liska <mliska@suse.cz>
16580 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16582 (possible_polymorphic_call_targets): Use it.
16583 (ipa_devirt): Likewise.
16585 2018-01-12 Martin Liska <mliska@suse.cz>
16587 * profile-count.h (enum profile_quality): Use 0 as invalid
16588 enum value of profile_quality.
16590 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16592 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16593 -mext-string options.
16595 2018-01-12 Richard Biener <rguenther@suse.de>
16597 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16598 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16599 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16601 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16603 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16605 * configure.ac (--with-long-double-format): Add support for the
16606 configuration option to change the default long double format on
16608 * config.gcc (powerpc*-linux*-*): Likewise.
16609 * configure: Regenerate.
16610 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16611 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16612 used without modification.
16614 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16616 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16617 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16618 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16619 MISC_BUILTIN_SPEC_BARRIER.
16620 (rs6000_init_builtins): Likewise.
16621 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16623 (speculation_barrier): New define_insn.
16624 * doc/extend.texi: Document __builtin_speculation_barrier.
16626 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16629 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16630 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16631 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16633 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16634 integral modes instead of "ss" and "sd".
16635 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16636 vectors with 32-bit and 64-bit elements.
16637 (vecdupssescalarmodesuffix): New mode attribute.
16638 (vec_dup<mode>): Use it.
16640 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16643 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16644 frame if argument is passed on stack.
16646 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16649 * ree.c (combine_reaching_defs): Optimize also
16650 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16651 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16653 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16655 PR middle-end/83189
16656 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16658 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16660 PR middle-end/83718
16661 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16662 after they are computed.
16664 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16666 PR tree-optimization/83695
16667 * gimple-loop-linterchange.cc
16668 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16669 reset cached scev information after interchange.
16670 (pass_linterchange::execute): Remove call to scev_reset_htab.
16672 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16674 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16675 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16676 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16677 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16678 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16679 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16680 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16681 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16682 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16683 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16684 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16685 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16686 (V_lane_reg): Likewise.
16687 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16689 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16690 (vfmal_lane_low<mode>_intrinsic,
16691 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16692 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16693 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16694 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16695 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16696 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16698 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16700 * config/arm/arm-cpus.in (fp16fml): New feature.
16701 (ALL_SIMD): Add fp16fml.
16702 (armv8.2-a): Add fp16fml as an option.
16703 (armv8.3-a): Likewise.
16704 (armv8.4-a): Add fp16fml as part of fp16.
16705 * config/arm/arm.h (TARGET_FP16FML): Define.
16706 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16708 * config/arm/arm-modes.def (V2HF): Define.
16709 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16710 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16711 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16712 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16713 vfmsl_low, vfmsl_high): New set of builtins.
16714 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16715 (vfml_op): New code attribute.
16716 (VFMLHALVES): New int iterator.
16717 (VFML, VFMLSEL): New mode attributes.
16718 (V_reg): Define mapping for V2HF.
16719 (V_hi, V_lo): New mode attributes.
16720 (VF_constraint): Likewise.
16721 (vfml_half, vfml_half_selector): New int attributes.
16722 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16724 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16725 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16727 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16728 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16729 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16730 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16732 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16733 Document new effective target and option set.
16735 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16737 * config/arm/arm-cpus.in (armv8_4): New feature.
16738 (ARMv8_4a): New fgroup.
16739 (armv8.4-a): New arch.
16740 * config/arm/arm-tables.opt: Regenerate.
16741 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16742 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16743 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16744 Add matching rules for -march=armv8.4-a and extensions.
16745 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16747 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16750 * config/rx/rx.md (BW): New mode attribute.
16751 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16753 2018-01-11 Richard Biener <rguenther@suse.de>
16755 PR tree-optimization/83435
16756 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16757 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16758 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16760 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16761 Alan Hayward <alan.hayward@arm.com>
16762 David Sherwood <david.sherwood@arm.com>
16764 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16766 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16767 (aarch64_print_address_internal): Use it to check for a zero offset.
16769 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16770 Alan Hayward <alan.hayward@arm.com>
16771 David Sherwood <david.sherwood@arm.com>
16773 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16774 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16775 Return a poly_int64 rather than a HOST_WIDE_INT.
16776 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16777 rather than a HOST_WIDE_INT.
16778 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16779 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16780 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16781 final_offset from HOST_WIDE_INT to poly_int64.
16782 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16783 to_constant when getting the number of units in an Advanced SIMD
16785 (aarch64_builtin_vectorized_function): Check for a constant number
16787 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16789 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16790 attribute instead of GET_MODE_NUNITS.
16791 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16792 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16793 GET_MODE_SIZE for fixed-size registers.
16794 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16795 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16796 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16797 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16798 (aarch64_print_operand, aarch64_print_address_internal)
16799 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16800 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16801 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16802 Handle polynomial GET_MODE_SIZE.
16803 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16804 wider than SImode without modification.
16805 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16806 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16807 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16808 passing and returning SVE modes.
16809 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16810 rather than GEN_INT.
16811 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16812 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16813 (aarch64_allocate_and_probe_stack_space): Likewise.
16814 (aarch64_layout_frame): Cope with polynomial offsets.
16815 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16816 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16817 polynomial offsets.
16818 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16819 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16820 poly_int64 rather than a HOST_WIDE_INT.
16821 (aarch64_get_separate_components, aarch64_process_components)
16822 (aarch64_expand_prologue, aarch64_expand_epilogue)
16823 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16824 (aarch64_anchor_offset): New function, split out from...
16825 (aarch64_legitimize_address): ...here.
16826 (aarch64_builtin_vectorization_cost): Handle polynomial
16827 TYPE_VECTOR_SUBPARTS.
16828 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16830 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16831 number of elements from the PARALLEL rather than the mode.
16832 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16833 rather than GET_MODE_BITSIZE.
16834 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16835 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16836 (aarch64_expand_vec_perm_const_1): Handle polynomial
16837 d->perm.length () and d->perm elements.
16838 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16839 Apply to_constant to d->perm elements.
16840 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16841 polynomial CONST_VECTOR_NUNITS.
16842 (aarch64_move_pointer): Take amount as a poly_int64 rather
16844 (aarch64_progress_pointer): Avoid temporary variable.
16845 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16846 the mode attribute instead of GET_MODE.
16848 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16849 Alan Hayward <alan.hayward@arm.com>
16850 David Sherwood <david.sherwood@arm.com>
16852 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16853 x exists before using it.
16854 (aarch64_add_constant_internal): Rename to...
16855 (aarch64_add_offset_1): ...this. Replace regnum with separate
16856 src and dest rtxes. Handle the case in which they're different,
16857 including when the offset is zero. Replace scratchreg with an rtx.
16858 Use 2 additions if there is no spare register into which we can
16859 move a 16-bit constant.
16860 (aarch64_add_constant): Delete.
16861 (aarch64_add_offset): Replace reg with separate src and dest
16862 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16863 Use aarch64_add_offset_1.
16864 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16865 an rtx rather than an int. Take the delta as a poly_int64
16866 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16867 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16868 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16869 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16870 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16871 and aarch64_add_sp.
16872 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16873 aarch64_add_constant.
16875 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16877 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16878 Use scalar_float_mode.
16880 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16882 * config/aarch64/aarch64-simd.md
16883 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16884 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16885 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16886 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16887 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16888 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16889 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16890 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16891 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16892 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16894 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16897 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16898 targ_options->x_arm_arch_string is non NULL.
16900 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16902 * config/aarch64/aarch64.h
16903 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16905 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16908 * expmed.c (emit_store_flag_force): Swap if const op0
16909 and change VOIDmode to mode of op0.
16911 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16913 PR rtl-optimization/83761
16914 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16915 than bytes to mode_for_size.
16917 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16919 PR middle-end/83189
16920 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16921 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16924 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16926 PR middle-end/83575
16927 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16928 when in layout mode.
16929 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16930 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16933 2018-01-10 Michael Collison <michael.collison@arm.com>
16935 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16936 * config/aarch64/aarch64-option-extension.def: Add
16937 AARCH64_OPT_EXTENSION of 'fp16fml'.
16938 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16939 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16940 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16941 * config/aarch64/constraints.md (Ui7): New constraint.
16942 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16943 (VFMLA_SEL_W): Ditto.
16946 (VFMLA16_LOW): New int iterator.
16947 (VFMLA16_HIGH): Ditto.
16948 (UNSPEC_FMLAL): New unspec.
16949 (UNSPEC_FMLSL): Ditto.
16950 (UNSPEC_FMLAL2): Ditto.
16951 (UNSPEC_FMLSL2): Ditto.
16952 (f16mac): New code attribute.
16953 * config/aarch64/aarch64-simd-builtins.def
16954 (aarch64_fmlal_lowv2sf): Ditto.
16955 (aarch64_fmlsl_lowv2sf): Ditto.
16956 (aarch64_fmlalq_lowv4sf): Ditto.
16957 (aarch64_fmlslq_lowv4sf): Ditto.
16958 (aarch64_fmlal_highv2sf): Ditto.
16959 (aarch64_fmlsl_highv2sf): Ditto.
16960 (aarch64_fmlalq_highv4sf): Ditto.
16961 (aarch64_fmlslq_highv4sf): Ditto.
16962 (aarch64_fmlal_lane_lowv2sf): Ditto.
16963 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16964 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16965 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16966 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16967 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16968 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16969 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16970 (aarch64_fmlal_lane_highv2sf): Ditto.
16971 (aarch64_fmlsl_lane_highv2sf): Ditto.
16972 (aarch64_fmlal_laneq_highv2sf): Ditto.
16973 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16974 (aarch64_fmlalq_lane_highv4sf): Ditto.
16975 (aarch64_fmlsl_lane_highv4sf): Ditto.
16976 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16977 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16978 * config/aarch64/aarch64-simd.md:
16979 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16980 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16981 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16982 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16983 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16984 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16985 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16986 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16987 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16988 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16989 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16990 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16991 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16992 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16993 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16994 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16995 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16996 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16997 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16998 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16999 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17000 (vfmlsl_low_u32): Ditto.
17001 (vfmlalq_low_u32): Ditto.
17002 (vfmlslq_low_u32): Ditto.
17003 (vfmlal_high_u32): Ditto.
17004 (vfmlsl_high_u32): Ditto.
17005 (vfmlalq_high_u32): Ditto.
17006 (vfmlslq_high_u32): Ditto.
17007 (vfmlal_lane_low_u32): Ditto.
17008 (vfmlsl_lane_low_u32): Ditto.
17009 (vfmlal_laneq_low_u32): Ditto.
17010 (vfmlsl_laneq_low_u32): Ditto.
17011 (vfmlalq_lane_low_u32): Ditto.
17012 (vfmlslq_lane_low_u32): Ditto.
17013 (vfmlalq_laneq_low_u32): Ditto.
17014 (vfmlslq_laneq_low_u32): Ditto.
17015 (vfmlal_lane_high_u32): Ditto.
17016 (vfmlsl_lane_high_u32): Ditto.
17017 (vfmlal_laneq_high_u32): Ditto.
17018 (vfmlsl_laneq_high_u32): Ditto.
17019 (vfmlalq_lane_high_u32): Ditto.
17020 (vfmlslq_lane_high_u32): Ditto.
17021 (vfmlalq_laneq_high_u32): Ditto.
17022 (vfmlslq_laneq_high_u32): Ditto.
17023 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17024 (AARCH64_FL_FOR_ARCH8_4): New.
17025 (AARCH64_ISA_F16FML): New ISA flag.
17026 (TARGET_F16FML): New feature flag for fp16fml.
17027 (doc/invoke.texi): Document new fp16fml option.
17029 2018-01-10 Michael Collison <michael.collison@arm.com>
17031 * config/aarch64/aarch64-builtins.c:
17032 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17033 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17034 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17035 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17036 (AARCH64_ISA_SHA3): New ISA flag.
17037 (TARGET_SHA3): New feature flag for sha3.
17038 * config/aarch64/iterators.md (sha512_op): New int attribute.
17039 (CRYPTO_SHA512): New int iterator.
17040 (UNSPEC_SHA512H): New unspec.
17041 (UNSPEC_SHA512H2): Ditto.
17042 (UNSPEC_SHA512SU0): Ditto.
17043 (UNSPEC_SHA512SU1): Ditto.
17044 * config/aarch64/aarch64-simd-builtins.def
17045 (aarch64_crypto_sha512hqv2di): New builtin.
17046 (aarch64_crypto_sha512h2qv2di): Ditto.
17047 (aarch64_crypto_sha512su0qv2di): Ditto.
17048 (aarch64_crypto_sha512su1qv2di): Ditto.
17049 (aarch64_eor3qv8hi): Ditto.
17050 (aarch64_rax1qv2di): Ditto.
17051 (aarch64_xarqv2di): Ditto.
17052 (aarch64_bcaxqv8hi): Ditto.
17053 * config/aarch64/aarch64-simd.md:
17054 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17055 (aarch64_crypto_sha512su0qv2di): Ditto.
17056 (aarch64_crypto_sha512su1qv2di): Ditto.
17057 (aarch64_eor3qv8hi): Ditto.
17058 (aarch64_rax1qv2di): Ditto.
17059 (aarch64_xarqv2di): Ditto.
17060 (aarch64_bcaxqv8hi): Ditto.
17061 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17062 (vsha512h2q_u64): Ditto.
17063 (vsha512su0q_u64): Ditto.
17064 (vsha512su1q_u64): Ditto.
17065 (veor3q_u16): Ditto.
17066 (vrax1q_u64): Ditto.
17067 (vxarq_u64): Ditto.
17068 (vbcaxq_u16): Ditto.
17069 * config/arm/types.md (crypto_sha512): New type attribute.
17070 (crypto_sha3): Ditto.
17071 (doc/invoke.texi): Document new sha3 option.
17073 2018-01-10 Michael Collison <michael.collison@arm.com>
17075 * config/aarch64/aarch64-builtins.c:
17076 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17077 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17078 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17079 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17080 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17081 (AARCH64_ISA_SM4): New ISA flag.
17082 (TARGET_SM4): New feature flag for sm4.
17083 * config/aarch64/aarch64-simd-builtins.def
17084 (aarch64_sm3ss1qv4si): Ditto.
17085 (aarch64_sm3tt1aq4si): Ditto.
17086 (aarch64_sm3tt1bq4si): Ditto.
17087 (aarch64_sm3tt2aq4si): Ditto.
17088 (aarch64_sm3tt2bq4si): Ditto.
17089 (aarch64_sm3partw1qv4si): Ditto.
17090 (aarch64_sm3partw2qv4si): Ditto.
17091 (aarch64_sm4eqv4si): Ditto.
17092 (aarch64_sm4ekeyqv4si): Ditto.
17093 * config/aarch64/aarch64-simd.md:
17094 (aarch64_sm3ss1qv4si): Ditto.
17095 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17096 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17097 (aarch64_sm4eqv4si): Ditto.
17098 (aarch64_sm4ekeyqv4si): Ditto.
17099 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17100 (sm3part_op): Ditto.
17101 (CRYPTO_SM3TT): Ditto.
17102 (CRYPTO_SM3PART): Ditto.
17103 (UNSPEC_SM3SS1): New unspec.
17104 (UNSPEC_SM3TT1A): Ditto.
17105 (UNSPEC_SM3TT1B): Ditto.
17106 (UNSPEC_SM3TT2A): Ditto.
17107 (UNSPEC_SM3TT2B): Ditto.
17108 (UNSPEC_SM3PARTW1): Ditto.
17109 (UNSPEC_SM3PARTW2): Ditto.
17110 (UNSPEC_SM4E): Ditto.
17111 (UNSPEC_SM4EKEY): Ditto.
17112 * config/aarch64/constraints.md (Ui2): New constraint.
17113 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17114 * config/arm/types.md (crypto_sm3): New type attribute.
17115 (crypto_sm4): Ditto.
17116 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17117 (vsm3tt1aq_u32): Ditto.
17118 (vsm3tt1bq_u32): Ditto.
17119 (vsm3tt2aq_u32): Ditto.
17120 (vsm3tt2bq_u32): Ditto.
17121 (vsm3partw1q_u32): Ditto.
17122 (vsm3partw2q_u32): Ditto.
17123 (vsm4eq_u32): Ditto.
17124 (vsm4ekeyq_u32): Ditto.
17125 (doc/invoke.texi): Document new sm4 option.
17127 2018-01-10 Michael Collison <michael.collison@arm.com>
17129 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17130 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17131 (AARCH64_FL_FOR_ARCH8_4): New.
17132 (AARCH64_FL_V8_4): New flag.
17133 (doc/invoke.texi): Document new armv8.4-a option.
17135 2018-01-10 Michael Collison <michael.collison@arm.com>
17137 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17138 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17139 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17140 * config/aarch64/aarch64-option-extension.def: Add
17141 AARCH64_OPT_EXTENSION of 'sha2'.
17142 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17143 (crypto): Disable sha2 and aes if crypto disabled.
17144 (crypto): Enable aes and sha2 if enabled.
17145 (simd): Disable sha2 and aes if simd disabled.
17146 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17148 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17149 (TARGET_SHA2): New feature flag for sha2.
17150 (TARGET_AES): New feature flag for aes.
17151 * config/aarch64/aarch64-simd.md:
17152 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17153 conditional on TARGET_AES.
17154 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17155 (aarch64_crypto_sha1hsi): Make pattern conditional
17157 (aarch64_crypto_sha1hv4si): Ditto.
17158 (aarch64_be_crypto_sha1hv4si): Ditto.
17159 (aarch64_crypto_sha1su1v4si): Ditto.
17160 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17161 (aarch64_crypto_sha1su0v4si): Ditto.
17162 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17163 (aarch64_crypto_sha256su0v4si): Ditto.
17164 (aarch64_crypto_sha256su1v4si): Ditto.
17165 (doc/invoke.texi): Document new aes and sha2 options.
17167 2018-01-10 Martin Sebor <msebor@redhat.com>
17169 PR tree-optimization/83781
17170 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17173 2018-01-11 Martin Sebor <msebor@gmail.com>
17174 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17176 PR tree-optimization/83501
17177 PR tree-optimization/81703
17179 * tree-ssa-strlen.c (get_string_cst): Rename...
17180 (get_string_len): ...to this. Handle global constants.
17181 (handle_char_store): Adjust.
17183 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17184 Jim Wilson <jimw@sifive.com>
17186 * config/riscv/riscv-protos.h (riscv_output_return): New.
17187 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17188 (riscv_attribute_table, riscv_output_return),
17189 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17190 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17191 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17193 (riscv_expand_prologue): Add early return for naked function.
17194 (riscv_expand_epilogue): Likewise.
17195 (riscv_function_ok_for_sibcall): Return false for naked function.
17196 (riscv_set_current_function): New.
17197 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17198 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17199 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17200 * doc/extend.texi (RISC-V Function Attributes): New.
17202 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17204 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17205 check for 128-bit long double before checking TCmode.
17206 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17207 128-bit long doubles before checking TFmode or TCmode.
17208 (FLOAT128_IBM_P): Likewise.
17210 2018-01-10 Martin Sebor <msebor@redhat.com>
17212 PR tree-optimization/83671
17213 * builtins.c (c_strlen): Unconditionally return zero for the empty
17215 Use -Warray-bounds for warnings.
17216 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17217 for non-constant array indices with COMPONENT_REF, arrays of
17218 arrays, and pointers to arrays.
17219 (gimple_fold_builtin_strlen): Determine and set length range for
17220 non-constant character arrays.
17222 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17224 PR middle-end/81897
17225 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17228 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17230 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17232 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17235 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17236 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17237 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17238 indexed_or_indirect_operand predicate.
17239 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17240 (*vsx_le_perm_load_v8hi): Likewise.
17241 (*vsx_le_perm_load_v16qi): Likewise.
17242 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17243 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17244 (*vsx_le_perm_store_v8hi): Likewise.
17245 (*vsx_le_perm_store_v16qi): Likewise.
17246 (eight unnamed splitters): Likewise.
17248 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17250 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17251 * config/rs6000/emmintrin.h: Likewise.
17252 * config/rs6000/mmintrin.h: Likewise.
17253 * config/rs6000/xmmintrin.h: Likewise.
17255 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17258 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17260 * tree.c (tree_nop_conversion): Return true for location wrapper
17262 (maybe_wrap_with_location): New function.
17263 (selftest::check_strip_nops): New function.
17264 (selftest::test_location_wrappers): New function.
17265 (selftest::tree_c_tests): Call it.
17266 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17267 (maybe_wrap_with_location): New decl.
17268 (EXPR_LOCATION_WRAPPER_P): New macro.
17269 (location_wrapper_p): New inline function.
17270 (tree_strip_any_location_wrapper): New inline function.
17272 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17275 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17276 stack_realign_offset for the largest alignment of stack slot
17278 (ix86_find_max_used_stack_alignment): New function.
17279 (ix86_finalize_stack_frame_flags): Use it. Set
17280 max_used_stack_alignment if we don't realign stack.
17281 * config/i386/i386.h (machine_function): Add
17282 max_used_stack_alignment.
17284 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17286 * config/arm/arm.opt (-mbranch-cost): New option.
17287 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17290 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17293 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17294 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17296 2018-01-10 Richard Biener <rguenther@suse.de>
17299 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17300 early out so it also covers the case where we have a non-NULL
17303 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17305 PR tree-optimization/83753
17306 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17307 for non-strided grouped accesses if the number of elements is 1.
17309 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17312 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17313 * i386.h (TARGET_USE_GATHER): Define.
17314 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17316 2018-01-10 Martin Liska <mliska@suse.cz>
17319 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17320 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17322 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17323 CLEANUP_NO_PARTITIONING is not set.
17325 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17327 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17328 for vectors, as a partial revert of r254296.
17329 * rtl.h (const_vec_p): Delete.
17330 (const_vec_duplicate_p): Don't test for vector CONSTs.
17331 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17332 * expmed.c (make_tree): Likewise.
17335 * common.md (E, F): Use CONSTANT_P instead of checking for
17337 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17338 checking for CONST_VECTOR.
17340 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17342 PR middle-end/83575
17343 * predict.c (force_edge_cold): Handle in more sane way edges
17344 with no prediction.
17346 2018-01-09 Carl Love <cel@us.ibm.com>
17348 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17350 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17351 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17352 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17353 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17354 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17355 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17356 * config/rs6000/rs6000-protos.h: Add extern defition for
17357 rs6000_generate_float2_double_code.
17358 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17360 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17361 (float2_v2df): Add define_expand.
17363 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17366 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17367 op_mode in the force_to_mode call.
17369 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17371 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17372 instead of checking each element individually.
17373 (aarch64_evpc_uzp): Likewise.
17374 (aarch64_evpc_zip): Likewise.
17375 (aarch64_evpc_ext): Likewise.
17376 (aarch64_evpc_rev): Likewise.
17377 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17378 instead of checking each element individually. Return true without
17380 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17381 whether all selected elements come from the same input, instead of
17382 checking each element individually. Remove calls to gen_rtx_REG,
17383 start_sequence and end_sequence and instead assert that no rtl is
17386 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17388 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17389 order of HIGH and CONST checks.
17391 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17393 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17394 if the destination isn't an SSA_NAME.
17396 2018-01-09 Richard Biener <rguenther@suse.de>
17398 PR tree-optimization/83668
17399 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17401 (canonicalize_loop_form): ... here, renamed from ...
17402 (canonicalize_loop_closed_ssa_form): ... this and amended to
17403 swap successor edges for loop exit blocks to make us use
17404 the RPO order we need for initial schedule generation.
17406 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17408 PR tree-optimization/64811
17409 * match.pd: When optimizing comparisons with Inf, avoid
17410 introducing or losing exceptions from comparisons with NaN.
17412 2018-01-09 Martin Liska <mliska@suse.cz>
17415 * asan.c (shadow_mem_size): Add gcc_assert.
17417 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17419 Don't save registers in main().
17422 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17423 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17424 * config/avr/avr.c (avr_set_current_function): Don't error if
17425 naked, OS_task or OS_main are specified at the same time.
17426 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17428 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17430 * common/config/avr/avr-common.c (avr_option_optimization_table):
17431 Switch on -mmain-is-OS_task for optimizing compilations.
17433 2018-01-09 Richard Biener <rguenther@suse.de>
17435 PR tree-optimization/83572
17436 * graphite.c: Include cfganal.h.
17437 (graphite_transform_loops): Connect infinite loops to exit
17438 and remove fake edges at the end.
17440 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17442 * ipa-inline.c (edge_badness): Revert accidental checkin.
17444 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17447 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17448 symbols; not inline clones.
17450 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17453 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17454 hard registers. Formatting fixes.
17456 PR preprocessor/83722
17457 * gcc.c (try_generate_repro): Pass
17458 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17459 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17462 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17463 Kito Cheng <kito.cheng@gmail.com>
17465 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17466 (riscv_leaf_function_p): Delete.
17467 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17469 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17471 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17473 (do_ifelse): New function.
17474 (do_isel): New function.
17475 (do_sub3): New function.
17476 (do_add3): New function.
17477 (do_load_mask_compare): New function.
17478 (do_overlap_load_compare): New function.
17479 (expand_compare_loop): New function.
17480 (expand_block_compare): Call expand_compare_loop() when appropriate.
17481 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17482 option description.
17483 (-mblock-compare-inline-loop-limit): New option.
17485 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17488 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17489 Reverse order of second and third operands in first alternative.
17490 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17491 of first and second elements in UNSPEC_VPERMR vector.
17492 (altivec_expand_vec_perm_le): Likewise.
17494 2018-01-08 Jeff Law <law@redhat.com>
17496 PR rtl-optimizatin/81308
17497 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17498 (process_switch): If group_case_labels makes a change, then set
17500 (pass_convert_switch::execute): If a switch is converted, then
17501 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17503 PR rtl-optimization/81308
17504 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17507 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17509 PR target/83663 - Revert r255946
17510 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17511 generation for cases where splatting a value is not useful.
17512 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17513 across a vec_duplicate and a paradoxical subreg forming a vector
17514 mode to a vec_concat.
17516 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17518 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17519 -march=armv8.3-a variants.
17520 * config/arm/t-multilib: Likewise.
17521 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17523 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17525 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17527 (cceq_ior_compare_complement): Give it a name so I can use it, and
17528 change boolean_or_operator predicate to boolean_operator so it can
17529 be used to generate a crand.
17530 (eqne): New code iterator.
17531 (bd/bd_neg): New code_attrs.
17532 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17533 a single define_insn.
17534 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17535 decrement (bdnzt/bdnzf/bdzt/bdzf).
17536 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17537 with the new names of the branch decrement patterns, and added the
17538 names of the branch decrement conditional patterns.
17540 2018-01-08 Richard Biener <rguenther@suse.de>
17542 PR tree-optimization/83563
17543 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17546 2018-01-08 Richard Biener <rguenther@suse.de>
17548 PR middle-end/83713
17549 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17551 2018-01-08 Richard Biener <rguenther@suse.de>
17553 PR tree-optimization/83685
17554 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17555 references to abnormals.
17557 2018-01-08 Richard Biener <rguenther@suse.de>
17560 * dwarf2out.c (output_indirect_strings): Handle empty
17561 skeleton_debug_str_hash.
17562 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17564 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17566 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17567 (emit_store_direct): Likewise.
17568 (arc_trampoline_adjust_address): Likewise.
17569 (arc_asm_trampoline_template): New function.
17570 (arc_initialize_trampoline): Use asm_trampoline_template.
17571 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17572 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17573 * config/arc/arc.md (flush_icache): Delete pattern.
17575 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17577 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17578 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17581 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17584 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17585 by not USED_FOR_TARGET.
17586 (make_pass_resolve_sw_modes): Likewise.
17588 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17590 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17593 2018-01-08 Richard Biener <rguenther@suse.de>
17595 PR middle-end/83580
17596 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17598 2018-01-08 Richard Biener <rguenther@suse.de>
17600 PR middle-end/83517
17601 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17603 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17605 PR middle-end/81897
17606 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17607 basic blocks with a small number of successors.
17608 (convert_control_dep_chain_into_preds): Improve handling of
17610 (dump_predicates): Split apart into...
17611 (dump_pred_chain): ...here...
17612 (dump_pred_info): ...and here.
17613 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17614 (can_chain_union_be_invalidated_p): Improve check for invalidation
17616 (uninit_uses_cannot_happen): Avoid unnecessary if
17617 convert_control_dep_chain_into_preds yielded nothing.
17619 2018-01-06 Martin Sebor <msebor@redhat.com>
17621 PR tree-optimization/83640
17622 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17623 subtracting negative offset from size.
17624 (builtin_access::overlap): Adjust offset bounds of the access to fall
17625 within the size of the object if possible.
17627 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17629 PR rtl-optimization/83699
17630 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17631 extract_bit_field_as_subreg to cases in which the extracted
17632 value is also a vector.
17634 * lra-constraints.c (process_alt_operands): Test for the equivalence
17635 substitutions when detecting a possible reload cycle.
17637 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17640 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17641 by default if flag_selective_schedling{,2}. Formatting fixes.
17643 PR rtl-optimization/83682
17644 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17645 if it has non-VECTOR_MODE element mode.
17646 (vec_duplicate_p): Likewise.
17648 PR middle-end/83694
17649 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17650 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17652 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17655 * config/i386/i386-builtin.def
17656 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17657 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17658 Require also OPTION_MASK_ISA_AVX512F in addition to
17659 OPTION_MASK_ISA_GFNI.
17660 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17661 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17662 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17663 to OPTION_MASK_ISA_GFNI.
17664 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17665 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17666 OPTION_MASK_ISA_AVX512BW.
17667 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17668 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17669 addition to OPTION_MASK_ISA_GFNI.
17670 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17671 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17672 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17673 to OPTION_MASK_ISA_GFNI.
17674 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17675 a requirement for all ISAs rather than any of them with a few
17677 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17679 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17680 bitmasks to be enabled with 3 exceptions, instead of requiring any
17681 enabled ISA with lots of exceptions.
17682 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17683 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17684 Change avx512bw in isa attribute to avx512f.
17685 * config/i386/sgxintrin.h: Add license boilerplate.
17686 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17687 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17688 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17689 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17691 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17692 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17693 temporarily sse2 rather than sse if not enabled already.
17696 * config/i386/sse.md (VI248_VLBW): Rename to ...
17697 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17698 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17699 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17700 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17701 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17702 mode iterator instead of VI248_VLBW.
17704 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17706 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17707 (record_modified): Skip clobbers; add debug output.
17708 (param_change_prob): Use sreal frequencies.
17710 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17712 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17713 punt for user-aligned variables.
17715 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17717 * tree-chrec.c (chrec_contains_symbols): Return true for
17720 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17723 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17724 of (x|y) == x for BICS pattern.
17726 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17728 PR tree-optimization/83605
17729 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17730 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17733 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17735 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17736 * config/epiphany/rtems.h: New file.
17738 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17739 Uros Bizjak <ubizjak@gmail.com>
17742 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17743 QIreg_operand instead of register_operand predicate.
17744 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17745 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17746 comments instead of -fmitigate[-_]rop.
17748 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17751 * cgraphunit.c (symbol_table::compile): Switch to text_section
17752 before calling assembly_start debug hook.
17753 * run-rtl-passes.c (run_rtl_passes): Likewise.
17756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17758 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17759 range_int_cst_p rather than !symbolic_range_p before calling
17760 extract_range_from_multiplicative_op_1.
17762 2018-01-04 Jeff Law <law@redhat.com>
17764 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17765 redundant test in assertion.
17767 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17769 * doc/rtl.texi: Document machine_mode wrapper classes.
17771 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17773 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17774 using tree_to_uhwi.
17776 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17778 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17779 the VEC_PERM_EXPR fold to fail.
17781 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17784 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17785 to switched_sections.
17787 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17790 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17791 test for d.testing.
17793 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17796 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17797 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17799 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17802 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17803 is BLKmode and bitpos not zero or mode change is needed.
17805 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17808 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17811 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17814 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17815 instead of MULT rtx. Update all corresponding splitters.
17816 (*saddl_se): Ditto.
17817 (*ssub<modesuffix>): Ditto.
17818 (*ssubl_se): Ditto.
17819 (*cmp_sadd_di): Update split patterns.
17820 (*cmp_sadd_si): Ditto.
17821 (*cmp_sadd_sidi): Ditto.
17822 (*cmp_ssub_di): Ditto.
17823 (*cmp_ssub_si): Ditto.
17824 (*cmp_ssub_sidi): Ditto.
17825 * config/alpha/predicates.md (const23_operand): New predicate.
17826 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17827 Look for ASHIFT, not MULT inner operand.
17828 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17830 2018-01-04 Martin Liska <mliska@suse.cz>
17832 PR gcov-profile/83669
17833 * gcov.c (output_intermediate_file): Add version to intermediate
17835 * doc/gcov.texi: Document new field 'version' in intermediate
17836 file format. Fix location of '-k' option of gcov command.
17838 2018-01-04 Martin Liska <mliska@suse.cz>
17841 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17843 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17845 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17847 2018-01-03 Martin Sebor <msebor@redhat.com>
17849 PR tree-optimization/83655
17850 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17851 checking calls with invalid arguments.
17853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17855 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17856 (vectorizable_mask_load_store): Delete.
17857 (vectorizable_call): Return false for masked loads and stores.
17858 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17859 instead of gimple_assign_rhs1.
17860 (vectorizable_load): Handle IFN_MASK_LOAD.
17861 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17865 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17867 (vectorizable_mask_load_store): ...here.
17868 (vectorizable_load): ...and here.
17870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17872 * tree-vect-stmts.c (vect_build_all_ones_mask)
17873 (vect_build_zero_merge_argument): New functions, split out from...
17874 (vectorizable_load): ...here.
17876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17878 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17880 (vectorizable_mask_load_store): ...here.
17881 (vectorizable_store): ...and here.
17883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17885 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17887 (vectorizable_mask_load_store): ...here.
17889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17891 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17892 (vect_model_store_cost): Take a vec_load_store_type instead of a
17894 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17895 (vect_model_store_cost): Take a vec_load_store_type instead of a
17897 (vectorizable_mask_load_store): Update accordingly.
17898 (vectorizable_store): Likewise.
17899 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17903 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17904 IFN_MASK_LOAD calls here rather than...
17905 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17908 Alan Hayward <alan.hayward@arm.com>
17909 David Sherwood <david.sherwood@arm.com>
17911 * expmed.c (extract_bit_field_1): For vector extracts,
17912 fall back to extract_bit_field_as_subreg if vec_extract
17915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17916 Alan Hayward <alan.hayward@arm.com>
17917 David Sherwood <david.sherwood@arm.com>
17919 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17920 they are variable or constant sized.
17921 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17922 slots for constant-sized data.
17924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17925 Alan Hayward <alan.hayward@arm.com>
17926 David Sherwood <david.sherwood@arm.com>
17928 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17929 handling COND_EXPRs with boolean comparisons, try to find a better
17930 basis for the mask type than the boolean itself.
17932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17934 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17935 is calculated and how it can be overridden.
17936 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17937 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17939 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17943 Alan Hayward <alan.hayward@arm.com>
17944 David Sherwood <david.sherwood@arm.com>
17946 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17947 Remove the mode argument.
17948 (aarch64_simd_valid_immediate): Remove the mode and inverse
17950 * config/aarch64/iterators.md (bitsize): New iterator.
17951 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17952 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17953 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17954 aarch64_simd_valid_immediate.
17955 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17956 (aarch64_reg_or_bic_imm): Likewise.
17957 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17958 with an insn_type enum and msl with a modifier_type enum.
17959 Replace element_width with a scalar_mode. Change the shift
17960 to unsigned int. Add constructors for scalar_float_mode and
17961 scalar_int_mode elements.
17962 (aarch64_vect_float_const_representable_p): Delete.
17963 (aarch64_can_const_movi_rtx_p)
17964 (aarch64_simd_scalar_immediate_valid_for_move)
17965 (aarch64_simd_make_constant): Update call to
17966 aarch64_simd_valid_immediate.
17967 (aarch64_advsimd_valid_immediate_hs): New function.
17968 (aarch64_advsimd_valid_immediate): Likewise.
17969 (aarch64_simd_valid_immediate): Remove mode and inverse
17970 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17971 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17972 and aarch64_float_const_representable_p on the result.
17973 (aarch64_output_simd_mov_immediate): Remove mode argument.
17974 Update call to aarch64_simd_valid_immediate and use of
17975 simd_immediate_info.
17976 (aarch64_output_scalar_simd_mov_immediate): Update call
17979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17980 Alan Hayward <alan.hayward@arm.com>
17981 David Sherwood <david.sherwood@arm.com>
17983 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17984 (mode_nunits): Likewise CONST_MODE_NUNITS.
17985 * machmode.def (ADJUST_NUNITS): Document.
17986 * genmodes.c (mode_data::need_nunits_adj): New field.
17987 (blank_mode): Update accordingly.
17988 (adj_nunits): New variable.
17989 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17991 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17992 listed in adj_nunits.
17993 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17994 listed in adj_nunits. Don't emit case statements for such modes.
17995 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17996 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17997 nothing if adj_nunits is nonnull.
17998 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17999 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18000 (emit_mode_fbit): Update use of print_maybe_const_decl.
18001 (emit_move_size): Likewise. Treat the array as non-const
18003 (emit_mode_adjustments): Handle adj_nunits.
18005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18007 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18008 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18009 (VECTOR_MODES): Use it.
18010 (make_vector_modes): Take the prefix as an argument.
18012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18013 Alan Hayward <alan.hayward@arm.com>
18014 David Sherwood <david.sherwood@arm.com>
18016 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18017 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18018 for MODE_VECTOR_BOOL.
18019 * machmode.def (VECTOR_BOOL_MODE): Document.
18020 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18021 (make_vector_bool_mode): New function.
18022 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18024 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18025 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18027 * stor-layout.c (int_mode_for_mode): Likewise.
18028 * tree.c (build_vector_type_for_mode): Likewise.
18029 * varasm.c (output_constant_pool_2): Likewise.
18030 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18031 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18032 for MODE_VECTOR_BOOL.
18033 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18034 of mode class checks.
18035 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18036 instead of a list of mode class checks.
18037 (expand_vector_scalar_condition): Likewise.
18038 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18041 Alan Hayward <alan.hayward@arm.com>
18042 David Sherwood <david.sherwood@arm.com>
18044 * machmode.h (mode_size): Change from unsigned short to
18046 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18047 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18048 or if measurement_type is not polynomial.
18049 (fixed_size_mode::includes_p): Check for constant-sized modes.
18050 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18051 return a poly_uint16 rather than an unsigned short.
18052 (emit_mode_size): Change the type of mode_size from unsigned short
18053 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18054 (emit_mode_adjustments): Cope with polynomial vector sizes.
18055 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18057 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18059 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18060 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18061 * caller-save.c (setup_save_areas): Likewise.
18062 (replace_reg_with_saved_mem): Likewise.
18063 * calls.c (emit_library_call_value_1): Likewise.
18064 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18065 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18066 (gen_lowpart_for_combine): Likewise.
18067 * convert.c (convert_to_integer_1): Likewise.
18068 * cse.c (equiv_constant, cse_insn): Likewise.
18069 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18070 (cselib_subst_to_values): Likewise.
18071 * dce.c (word_dce_process_block): Likewise.
18072 * df-problems.c (df_word_lr_mark_ref): Likewise.
18073 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18074 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18075 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18076 (rtl_for_decl_location): Likewise.
18077 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18078 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18079 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18080 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18081 (expand_expr_real_1): Likewise.
18082 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18083 (pad_below): Likewise.
18084 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18085 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18086 * ira.c (get_subreg_tracking_sizes): Likewise.
18087 * ira-build.c (ira_create_allocno_objects): Likewise.
18088 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18089 (ira_sort_regnos_for_alter_reg): Likewise.
18090 * ira-costs.c (record_operand_costs): Likewise.
18091 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18092 (resolve_simple_move): Likewise.
18093 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18094 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18095 (lra_constraints): Likewise.
18096 (CONST_POOL_OK_P): Reject variable-sized modes.
18097 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18098 (add_pseudo_to_slot, lra_spill): Likewise.
18099 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18100 * optabs-query.c (get_best_extraction_insn): Likewise.
18101 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18102 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18103 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18104 * recog.c (offsettable_address_addr_space_p): Likewise.
18105 * regcprop.c (maybe_mode_change): Likewise.
18106 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18107 * regrename.c (build_def_use): Likewise.
18108 * regstat.c (dump_reg_info): Likewise.
18109 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18110 (find_reloads, find_reloads_subreg_address): Likewise.
18111 * reload1.c (eliminate_regs_1): Likewise.
18112 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18113 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18114 (simplify_binary_operation_1, simplify_subreg): Likewise.
18115 * targhooks.c (default_function_arg_padding): Likewise.
18116 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18117 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18118 (verify_gimple_assign_ternary): Likewise.
18119 * tree-inline.c (estimate_move_cost): Likewise.
18120 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18121 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18122 (get_address_cost_ainc): Likewise.
18123 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18124 (vect_supportable_dr_alignment): Likewise.
18125 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18126 (vectorizable_reduction): Likewise.
18127 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18128 (vectorizable_operation, vectorizable_load): Likewise.
18129 * tree.c (build_same_sized_truth_vector_type): Likewise.
18130 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18131 * var-tracking.c (emit_note_insn_var_location): Likewise.
18132 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18133 (ADDR_VEC_ALIGN): Likewise.
18135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18136 Alan Hayward <alan.hayward@arm.com>
18137 David Sherwood <david.sherwood@arm.com>
18139 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18141 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18142 or if measurement_type is polynomial.
18143 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18144 * combine.c (make_extraction): Likewise.
18145 * dse.c (find_shift_sequence): Likewise.
18146 * dwarf2out.c (mem_loc_descriptor): Likewise.
18147 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18148 (extract_bit_field, extract_low_bits): Likewise.
18149 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18150 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18151 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18152 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18153 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18154 * reload.c (find_reloads): Likewise.
18155 * reload1.c (alter_reg): Likewise.
18156 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18157 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18158 * tree-if-conv.c (predicate_mem_writes): Likewise.
18159 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18160 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18161 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18162 * valtrack.c (dead_debug_insert_temp): Likewise.
18163 * varasm.c (mergeable_constant_section): Likewise.
18164 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18167 Alan Hayward <alan.hayward@arm.com>
18168 David Sherwood <david.sherwood@arm.com>
18170 * expr.c (expand_assignment): Cope with polynomial mode sizes
18171 when assigning to a CONCAT.
18173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18174 Alan Hayward <alan.hayward@arm.com>
18175 David Sherwood <david.sherwood@arm.com>
18177 * machmode.h (mode_precision): Change from unsigned short to
18179 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18181 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18182 or if measurement_type is not polynomial.
18183 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18184 in which the mode is already known to be a scalar_int_mode.
18185 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18186 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18188 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18189 for GET_MODE_PRECISION.
18190 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18191 for GET_MODE_PRECISION.
18192 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18194 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18195 (expand_field_assignment, make_extraction): Likewise.
18196 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18197 (get_last_value): Likewise.
18198 * convert.c (convert_to_integer_1): Likewise.
18199 * cse.c (cse_insn): Likewise.
18200 * expr.c (expand_expr_real_1): Likewise.
18201 * lra-constraints.c (simplify_operand_subreg): Likewise.
18202 * optabs-query.c (can_atomic_load_p): Likewise.
18203 * optabs.c (expand_atomic_load): Likewise.
18204 (expand_atomic_store): Likewise.
18205 * ree.c (combine_reaching_defs): Likewise.
18206 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18207 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18208 * tree.h (type_has_mode_precision_p): Likewise.
18209 * ubsan.c (instrument_si_overflow): Likewise.
18211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18212 Alan Hayward <alan.hayward@arm.com>
18213 David Sherwood <david.sherwood@arm.com>
18215 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18216 polynomial numbers of units.
18217 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18218 (valid_vector_subparts_p): New function.
18219 (build_vector_type): Remove temporary shim and take the number
18220 of units as a poly_uint64 rather than an int.
18221 (build_opaque_vector_type): Take the number of units as a
18222 poly_uint64 rather than an int.
18223 * tree.c (build_vector_from_ctor): Handle polynomial
18224 TYPE_VECTOR_SUBPARTS.
18225 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18226 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18227 (build_vector_from_val): If the number of units is variable,
18228 use build_vec_duplicate_cst for constant operands and
18229 VEC_DUPLICATE_EXPR otherwise.
18230 (make_vector_type): Remove temporary is_constant ().
18231 (build_vector_type, build_opaque_vector_type): Take the number of
18232 units as a poly_uint64 rather than an int.
18233 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18235 * cfgexpand.c (expand_debug_expr): Likewise.
18236 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18237 (store_constructor, expand_expr_real_1): Likewise.
18238 (const_scalar_mask_from_tree): Likewise.
18239 * fold-const-call.c (fold_const_reduction): Likewise.
18240 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18241 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18242 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18243 (fold_relational_const): Likewise.
18244 (native_interpret_vector): Likewise. Change the size from an
18245 int to an unsigned int.
18246 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18247 TYPE_VECTOR_SUBPARTS.
18248 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18249 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18250 duplicating a non-constant operand into a variable-length vector.
18251 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18252 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18253 * ipa-icf.c (sem_variable::equals): Likewise.
18254 * match.pd: Likewise.
18255 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18256 * print-tree.c (print_node): Likewise.
18257 * stor-layout.c (layout_type): Likewise.
18258 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18259 * tree-cfg.c (verify_gimple_comparison): Likewise.
18260 (verify_gimple_assign_binary): Likewise.
18261 (verify_gimple_assign_ternary): Likewise.
18262 (verify_gimple_assign_single): Likewise.
18263 * tree-pretty-print.c (dump_generic_node): Likewise.
18264 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18265 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18266 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18267 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18268 (vect_shift_permute_load_chain): Likewise.
18269 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18270 (expand_vector_condition, optimize_vector_constructor): Likewise.
18271 (lower_vec_perm, get_compute_type): Likewise.
18272 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18273 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18274 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18275 (vect_recog_mask_conversion_pattern): Likewise.
18276 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18277 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18278 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18279 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18280 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18281 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18282 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18283 (supportable_widening_operation): Likewise.
18284 (supportable_narrowing_operation): Likewise.
18285 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18287 * varasm.c (output_constant): Likewise.
18289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18290 Alan Hayward <alan.hayward@arm.com>
18291 David Sherwood <david.sherwood@arm.com>
18293 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18294 so that both the length == 3 and length != 3 cases set up their
18295 own permute vectors. Add comments explaining why we know the
18296 number of elements is constant.
18297 (vect_permute_load_chain): Likewise.
18299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18300 Alan Hayward <alan.hayward@arm.com>
18301 David Sherwood <david.sherwood@arm.com>
18303 * machmode.h (mode_nunits): Change from unsigned char to
18305 (ONLY_FIXED_SIZE_MODES): New macro.
18306 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18307 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18308 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18310 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18311 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18312 or if measurement_type is not polynomial.
18313 * genmodes.c (ZERO_COEFFS): New macro.
18314 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18316 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18317 Use ZERO_COEFFS when emitting initializers.
18318 * data-streamer.h (bp_pack_poly_value): New function.
18319 (bp_unpack_poly_value): Likewise.
18320 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18321 for GET_MODE_NUNITS.
18322 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18323 for GET_MODE_NUNITS.
18324 * tree.c (make_vector_type): Remove temporary shim and make
18325 the real function take the number of units as a poly_uint64
18326 rather than an int.
18327 (build_vector_type_for_mode): Handle polynomial nunits.
18328 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18329 * emit-rtl.c (const_vec_series_p_1): Likewise.
18330 (gen_rtx_CONST_VECTOR): Likewise.
18331 * fold-const.c (test_vec_duplicate_folding): Likewise.
18332 * genrecog.c (validate_pattern): Likewise.
18333 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18334 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18335 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18336 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18337 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18338 * rtlanal.c (subreg_get_info): Likewise.
18339 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18340 (vect_grouped_load_supported): Likewise.
18341 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18342 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18343 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18344 (simplify_const_unary_operation, simplify_binary_operation_1)
18345 (simplify_const_binary_operation, simplify_ternary_operation)
18346 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18347 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18348 instead of CONST_VECTOR_NUNITS.
18349 * varasm.c (output_constant_pool_2): Likewise.
18350 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18351 explicit-encoded elements in the XVEC for variable-length vectors.
18353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18355 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18358 Alan Hayward <alan.hayward@arm.com>
18359 David Sherwood <david.sherwood@arm.com>
18361 * coretypes.h (fixed_size_mode): Declare.
18362 (fixed_size_mode_pod): New typedef.
18363 * builtins.h (target_builtins::x_apply_args_mode)
18364 (target_builtins::x_apply_result_mode): Change type to
18365 fixed_size_mode_pod.
18366 * builtins.c (apply_args_size, apply_result_size, result_vector)
18367 (expand_builtin_apply_args_1, expand_builtin_apply)
18368 (expand_builtin_return): Update accordingly.
18370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18372 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18373 * cselib.c (cselib_hash_rtx): Likewise.
18374 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18375 CONST_VECTOR encoding.
18377 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18378 Jeff Law <law@redhat.com>
18381 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18382 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18383 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18384 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18387 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18388 explicitly probe *sp in a noreturn function if there were any callee
18389 register saves or frame pointer is needed.
18391 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18394 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18395 BLKmode for ternary, binary or unary expressions.
18398 * var-tracking.c (delete_vta_debug_insn): New inline function.
18399 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18400 insns from get_insns () to NULL instead of each bb separately.
18401 Use delete_vta_debug_insn. No longer static.
18402 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18403 delete_vta_debug_insns callers.
18404 * rtl.h (delete_vta_debug_insns): Declare.
18405 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18406 instead of variable_tracking_main.
18408 2018-01-03 Martin Sebor <msebor@redhat.com>
18410 PR tree-optimization/83603
18411 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18412 arguments past the endof the argument list in functions declared
18413 without a prototype.
18414 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18415 Avoid checking when arguments are null.
18417 2018-01-03 Martin Sebor <msebor@redhat.com>
18420 * doc/extend.texi (attribute const): Fix a typo.
18421 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18422 issuing -Wsuggest-attribute for void functions.
18424 2018-01-03 Martin Sebor <msebor@redhat.com>
18426 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18427 offset_int::from instead of wide_int::to_shwi.
18428 (maybe_diag_overlap): Remove assertion.
18429 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18430 * gimple-ssa-sprintf.c (format_directive): Same.
18431 (parse_directive): Same.
18432 (sprintf_dom_walker::compute_format_length): Same.
18433 (try_substitute_return_value): Same.
18435 2018-01-03 Jeff Law <law@redhat.com>
18437 PR middle-end/83654
18438 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18439 non-constant residual for zero at runtime and avoid probing in
18440 that case. Reorganize code for trailing problem to mirror handling
18443 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18445 PR tree-optimization/83501
18446 * tree-ssa-strlen.c (get_string_cst): New.
18447 (handle_char_store): Call get_string_cst.
18449 2018-01-03 Martin Liska <mliska@suse.cz>
18451 PR tree-optimization/83593
18452 * tree-ssa-strlen.c: Include tree-cfg.h.
18453 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18454 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18455 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18457 (strlen_dom_walker::before_dom_children): Call
18458 gimple_purge_dead_eh_edges. Dump tranformation with details
18460 (strlen_dom_walker::before_dom_children): Update call by adding
18461 new argument cleanup_eh.
18462 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18464 2018-01-03 Martin Liska <mliska@suse.cz>
18467 * cif-code.def (VARIADIC_THUNK): New enum value.
18468 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18471 2018-01-03 Jan Beulich <jbeulich@suse.com>
18473 * sse.md (mov<mode>_internal): Tighten condition for when to use
18474 vmovdqu<ssescalarsize> for TI and OI modes.
18476 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18478 Update copyright years.
18480 2018-01-03 Martin Liska <mliska@suse.cz>
18483 * ipa-visibility.c (function_and_variable_visibility): Skip
18484 functions with noipa attribure.
18486 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18488 * gcc.c (process_command): Update copyright notice dates.
18489 * gcov-dump.c (print_version): Ditto.
18490 * gcov.c (print_version): Ditto.
18491 * gcov-tool.c (print_version): Ditto.
18492 * gengtype.c (create_file): Ditto.
18493 * doc/cpp.texi: Bump @copying's copyright year.
18494 * doc/cppinternals.texi: Ditto.
18495 * doc/gcc.texi: Ditto.
18496 * doc/gccint.texi: Ditto.
18497 * doc/gcov.texi: Ditto.
18498 * doc/install.texi: Ditto.
18499 * doc/invoke.texi: Ditto.
18501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18503 * vector-builder.h (vector_builder::m_full_nelts): Change from
18504 unsigned int to poly_uint64.
18505 (vector_builder::full_nelts): Update prototype accordingly.
18506 (vector_builder::new_vector): Likewise.
18507 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18508 (vector_builder::operator ==): Likewise.
18509 (vector_builder::finalize): Likewise.
18510 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18511 Take the number of elements as a poly_uint64 rather than an
18513 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18514 from unsigned int to poly_uint64.
18515 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18516 (vec_perm_indices::new_vector): Likewise.
18517 (vec_perm_indices::length): Likewise.
18518 (vec_perm_indices::nelts_per_input): Likewise.
18519 (vec_perm_indices::input_nelts): Likewise.
18520 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18521 number of elements per input as a poly_uint64 rather than an
18522 unsigned int. Use the original encoding for variable-length
18523 vectors, rather than clamping each individual element.
18524 For the second and subsequent elements in each pattern,
18525 clamp the step and base before clamping their sum.
18526 (vec_perm_indices::series_p): Handle polynomial element counts.
18527 (vec_perm_indices::all_in_range_p): Likewise.
18528 (vec_perm_indices_to_tree): Likewise.
18529 (vec_perm_indices_to_rtx): Likewise.
18530 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18531 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18532 (tree_vector_builder::new_binary_operation): Handle polynomial
18533 element counts. Return false if we need to know the number
18534 of elements at compile time.
18535 * fold-const.c (fold_vec_perm): Punt if the number of elements
18536 isn't known at compile time.
18538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18540 * vec-perm-indices.h (vec_perm_builder): Change element type
18541 from HOST_WIDE_INT to poly_int64.
18542 (vec_perm_indices::element_type): Update accordingly.
18543 (vec_perm_indices::clamp): Handle polynomial element_types.
18544 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18545 (vec_perm_indices::all_in_range_p): Likewise.
18546 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18548 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18549 polynomial vec_perm_indices element types.
18550 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18551 * fold-const.c (fold_vec_perm): Likewise.
18552 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18553 * tree-vect-generic.c (lower_vec_perm): Likewise.
18554 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18555 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18556 element type to HOST_WIDE_INT.
18558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18559 Alan Hayward <alan.hayward@arm.com>
18560 David Sherwood <david.sherwood@arm.com>
18562 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18563 rather than an int. Use plus_constant.
18564 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18565 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18568 Alan Hayward <alan.hayward@arm.com>
18569 David Sherwood <david.sherwood@arm.com>
18571 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18572 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
18579 mode sizes. Require a constant size for BLKmode parameters
18580 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18581 forces a parameter to be padded at the lsb end in order to
18582 fill a complete number of words, require the parameter size
18583 to be ordered wrt UNITS_PER_WORD.
18585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18586 Alan Hayward <alan.hayward@arm.com>
18587 David Sherwood <david.sherwood@arm.com>
18589 * reload1.c (spill_stack_slot_width): Change element type
18590 from unsigned int to poly_uint64_pod.
18591 (alter_reg): Treat mode sizes as polynomial.
18593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18594 Alan Hayward <alan.hayward@arm.com>
18595 David Sherwood <david.sherwood@arm.com>
18597 * reload.c (complex_word_subreg_p): New function.
18598 (reload_inner_reg_of_subreg, push_reload): Use it.
18600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18601 Alan Hayward <alan.hayward@arm.com>
18602 David Sherwood <david.sherwood@arm.com>
18604 * lra-constraints.c (process_alt_operands): Reject matched
18605 operands whose sizes aren't ordered.
18606 (match_reload): Refer to this check here.
18608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18609 Alan Hayward <alan.hayward@arm.com>
18610 David Sherwood <david.sherwood@arm.com>
18612 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18613 that the mode size is in the set {1, 2, 4, 8, 16}.
18615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18616 Alan Hayward <alan.hayward@arm.com>
18617 David Sherwood <david.sherwood@arm.com>
18619 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18620 Use plus_constant instead of gen_rtx_PLUS.
18622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18623 Alan Hayward <alan.hayward@arm.com>
18624 David Sherwood <david.sherwood@arm.com>
18626 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18627 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18628 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18629 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18630 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18631 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18632 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18633 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18634 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18635 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18637 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18638 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18639 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18640 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18641 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18642 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18643 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18644 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18645 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18646 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18648 * expr.c (emit_move_resolve_push): Treat the input and result
18649 of PUSH_ROUNDING as a poly_int64.
18650 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18651 (emit_push_insn): Likewise.
18652 * lra-eliminations.c (mark_not_eliminable): Likewise.
18653 * recog.c (push_operand): Likewise.
18654 * reload1.c (elimination_effects): Likewise.
18655 * rtlanal.c (nonzero_bits1): Likewise.
18656 * calls.c (store_one_arg): Likewise. Require the padding to be
18657 known at compile time.
18659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18660 Alan Hayward <alan.hayward@arm.com>
18661 David Sherwood <david.sherwood@arm.com>
18663 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18664 Use plus_constant instead of gen_rtx_PLUS.
18666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18667 Alan Hayward <alan.hayward@arm.com>
18668 David Sherwood <david.sherwood@arm.com>
18670 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18671 rather than an int.
18673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18674 Alan Hayward <alan.hayward@arm.com>
18675 David Sherwood <david.sherwood@arm.com>
18677 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18678 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18679 via stack temporaries. Treat the mode size as polynomial too.
18681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18682 Alan Hayward <alan.hayward@arm.com>
18683 David Sherwood <david.sherwood@arm.com>
18685 * expr.c (expand_expr_real_2): When handling conversions involving
18686 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18687 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18688 as a poly_uint64 too.
18690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18691 Alan Hayward <alan.hayward@arm.com>
18692 David Sherwood <david.sherwood@arm.com>
18694 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18697 Alan Hayward <alan.hayward@arm.com>
18698 David Sherwood <david.sherwood@arm.com>
18700 * combine.c (can_change_dest_mode): Handle polynomial
18701 REGMODE_NATURAL_SIZE.
18702 * expmed.c (store_bit_field_1): Likewise.
18703 * expr.c (store_constructor): Likewise.
18704 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18705 and polynomial REGMODE_NATURAL_SIZE.
18706 (gen_lowpart_common): Likewise.
18707 * reginfo.c (record_subregs_of_mode): Likewise.
18708 * rtlanal.c (read_modify_subreg_p): Likewise.
18710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18711 Alan Hayward <alan.hayward@arm.com>
18712 David Sherwood <david.sherwood@arm.com>
18714 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18715 numbers of elements.
18717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18718 Alan Hayward <alan.hayward@arm.com>
18719 David Sherwood <david.sherwood@arm.com>
18721 * match.pd: Cope with polynomial numbers of vector elements.
18723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18724 Alan Hayward <alan.hayward@arm.com>
18725 David Sherwood <david.sherwood@arm.com>
18727 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18728 in a POINTER_PLUS_EXPR.
18730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18731 Alan Hayward <alan.hayward@arm.com>
18732 David Sherwood <david.sherwood@arm.com>
18734 * omp-simd-clone.c (simd_clone_subparts): New function.
18735 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18736 (ipa_simd_modify_function_body): Likewise.
18738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18739 Alan Hayward <alan.hayward@arm.com>
18740 David Sherwood <david.sherwood@arm.com>
18742 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18743 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18744 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18745 (expand_vector_condition, vector_element): Likewise.
18746 (subparts_gt): New function.
18747 (get_compute_type): Use subparts_gt.
18748 (count_type_subparts): Delete.
18749 (expand_vector_operations_1): Use subparts_gt instead of
18750 count_type_subparts.
18752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18753 Alan Hayward <alan.hayward@arm.com>
18754 David Sherwood <david.sherwood@arm.com>
18756 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18757 (vect_compile_time_alias): ...this new function. Do the calculation
18758 on poly_ints rather than trees.
18759 (vect_prune_runtime_alias_test_list): Update call accordingly.
18761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18762 Alan Hayward <alan.hayward@arm.com>
18763 David Sherwood <david.sherwood@arm.com>
18765 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18767 (vect_schedule_slp_instance): Likewise.
18769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18770 Alan Hayward <alan.hayward@arm.com>
18771 David Sherwood <david.sherwood@arm.com>
18773 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18774 constant and extern definitions for variable-length vectors.
18775 (vect_get_constant_vectors): Note that the number of units
18776 is known to be constant.
18778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18779 Alan Hayward <alan.hayward@arm.com>
18780 David Sherwood <david.sherwood@arm.com>
18782 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18783 of units as polynomial. Choose between WIDE and NARROW based
18786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18787 Alan Hayward <alan.hayward@arm.com>
18788 David Sherwood <david.sherwood@arm.com>
18790 * tree-vect-stmts.c (simd_clone_subparts): New function.
18791 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18794 Alan Hayward <alan.hayward@arm.com>
18795 David Sherwood <david.sherwood@arm.com>
18797 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18798 vectors as polynomial. Use build_index_vector for
18799 IFN_GOMP_SIMD_LANE.
18801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18802 Alan Hayward <alan.hayward@arm.com>
18803 David Sherwood <david.sherwood@arm.com>
18805 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18806 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18807 for variable-length vectors.
18808 (vectorizable_mask_load_store): Treat the number of units as
18809 polynomial, asserting that it is constant if the condition has
18810 already been enforced.
18811 (vectorizable_store, vectorizable_load): Likewise.
18813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18814 Alan Hayward <alan.hayward@arm.com>
18815 David Sherwood <david.sherwood@arm.com>
18817 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18818 of units as polynomial. Punt if we can't tell at compile time
18819 which vector contains the final result.
18821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18822 Alan Hayward <alan.hayward@arm.com>
18823 David Sherwood <david.sherwood@arm.com>
18825 * tree-vect-loop.c (vectorizable_induction): Treat the number
18826 of units as polynomial. Punt on SLP inductions. Use an integer
18827 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18828 cast of such a series for variable-length floating-point
18831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18832 Alan Hayward <alan.hayward@arm.com>
18833 David Sherwood <david.sherwood@arm.com>
18835 * tree.h (build_index_vector): Declare.
18836 * tree.c (build_index_vector): New function.
18837 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18838 of units as polynomial, forcibly converting it to a constant if
18839 vectorizable_reduction has already enforced the condition.
18840 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18841 to create a {1,2,3,...} vector.
18842 (vectorizable_reduction): Treat the number of units as polynomial.
18843 Choose vectype_in based on the largest scalar element size rather
18844 than the smallest number of units. Enforce the restrictions
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18852 number of units as polynomial.
18854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18855 Alan Hayward <alan.hayward@arm.com>
18856 David Sherwood <david.sherwood@arm.com>
18858 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18859 * target.def (autovectorize_vector_sizes): Return the vector sizes
18860 by pointer, using vector_sizes rather than a bitmask.
18861 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18862 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18863 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18865 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18866 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18867 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18868 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18869 * omp-general.c (omp_max_vf): Likewise.
18870 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18871 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18872 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18873 * tree-vect-slp.c (vect_slp_bb): Likewise.
18874 * doc/tm.texi: Regenerate.
18875 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18877 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18878 the vector size as a poly_uint64 rather than an unsigned int.
18879 (current_vector_size): Change from an unsigned int to a poly_uint64.
18880 (get_vectype_for_scalar_type): Update accordingly.
18881 * tree.h (build_truth_vector_type): Take the size and number of
18882 units as a poly_uint64 rather than an unsigned int.
18883 (build_vector_type): Add a temporary overload that takes
18884 the number of units as a poly_uint64 rather than an unsigned int.
18885 * tree.c (make_vector_type): Likewise.
18886 (build_truth_vector_type): Take the number of units as a poly_uint64
18887 rather than an unsigned int.
18889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18890 Alan Hayward <alan.hayward@arm.com>
18891 David Sherwood <david.sherwood@arm.com>
18893 * target.def (get_mask_mode): Take the number of units and length
18894 as poly_uint64s rather than unsigned ints.
18895 * targhooks.h (default_get_mask_mode): Update accordingly.
18896 * targhooks.c (default_get_mask_mode): Likewise.
18897 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18898 * doc/tm.texi: Regenerate.
18900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18901 Alan Hayward <alan.hayward@arm.com>
18902 David Sherwood <david.sherwood@arm.com>
18904 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18905 * omp-general.c (omp_max_vf): Likewise.
18906 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18907 (expand_omp_simd): Handle polynomial safelen.
18908 * omp-low.c (omplow_simd_context): Add a default constructor.
18909 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18910 (lower_rec_simd_input_clauses): Update accordingly.
18911 (lower_rec_input_clauses): Likewise.
18913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18914 Alan Hayward <alan.hayward@arm.com>
18915 David Sherwood <david.sherwood@arm.com>
18917 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18918 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18919 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18920 (vect_analyze_slp_cost): Likewise.
18921 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18922 (vect_model_load_cost): Likewise.
18924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18925 Alan Hayward <alan.hayward@arm.com>
18926 David Sherwood <david.sherwood@arm.com>
18928 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18929 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18930 from an unsigned int * to a poly_uint64_pod *.
18931 (calculate_unrolling_factor): New function.
18932 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18935 Alan Hayward <alan.hayward@arm.com>
18936 David Sherwood <david.sherwood@arm.com>
18938 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18939 from an unsigned int to a poly_uint64.
18940 (_loop_vec_info::slp_unrolling_factor): Likewise.
18941 (_loop_vec_info::vectorization_factor): Change from an int
18943 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18944 (vect_get_num_vectors): New function.
18945 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18946 (vect_get_num_copies): Use vect_get_num_vectors.
18947 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18948 to an unsigned int *.
18949 (vect_analyze_data_refs): Change min_vf from an int * to a
18951 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18952 than an unsigned HOST_WIDE_INT.
18953 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18954 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18955 to an unsigned int *.
18956 (vect_analyze_data_ref_dependences): Likewise.
18957 (vect_compute_data_ref_alignment): Handle polynomial vf.
18958 (vect_enhance_data_refs_alignment): Likewise.
18959 (vect_prune_runtime_alias_test_list): Likewise.
18960 (vect_shift_permute_load_chain): Likewise.
18961 (vect_supportable_dr_alignment): Likewise.
18962 (dependence_distance_ge_vf): Take the vectorization factor as a
18963 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18964 (vect_analyze_data_refs): Change min_vf from an int * to a
18966 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18967 vfm1 as a poly_uint64 rather than an int. Make the same change
18968 for the returned bound_scalar.
18969 (vect_gen_vector_loop_niters): Handle polynomial vf.
18970 (vect_do_peeling): Likewise. Update call to
18971 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18972 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18974 * tree-vect-loop.c (vect_determine_vectorization_factor)
18975 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18976 (vect_get_known_peeling_cost): Likewise.
18977 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18978 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18979 (vect_transform_loop): Likewise. Use the lowest possible VF when
18980 updating the upper bounds of the loop.
18981 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18982 rather than an int.
18983 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18984 polynomial unroll factors.
18985 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18986 (vect_make_slp_decision): Likewise.
18987 (vect_supported_load_permutation_p): Likewise, and polynomial
18989 (vect_analyze_slp_cost): Handle polynomial vf.
18990 (vect_slp_analyze_node_operations): Likewise.
18991 (vect_slp_analyze_bb_1): Likewise.
18992 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18993 than an unsigned HOST_WIDE_INT.
18994 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18995 (vectorizable_load): Handle polynomial vf.
18996 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18998 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19001 Alan Hayward <alan.hayward@arm.com>
19002 David Sherwood <david.sherwood@arm.com>
19004 * match.pd: Handle bit operations involving three constants
19005 and try to fold one pair.
19007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19009 * tree-vect-loop-manip.c: Include gimple-fold.h.
19010 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19011 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19012 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19013 Add a path that uses a step of VF instead of 1, but disable it
19015 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19016 and niters_no_overflow parameters. Update calls to
19017 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19018 Create a new SSA name if the latter choses to use a ste other
19019 than zero, and return it via niters_vector_mult_vf_var.
19020 * tree-vect-loop.c (vect_transform_loop): Update calls to
19021 vect_do_peeling, vect_gen_vector_loop_niters and
19022 slpeel_make_loop_iterate_ntimes.
19023 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19024 (vect_gen_vector_loop_niters): Update declarations after above changes.
19026 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19028 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19029 128-bit round to integer instructions.
19030 (ceil<mode>2): Likewise.
19031 (btrunc<mode>2): Likewise.
19032 (round<mode>2): Likewise.
19034 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19036 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19037 unaligned VSX load/store on P8/P9.
19038 (expand_block_clear): Allow the use of unaligned VSX
19039 load/store on P8/P9.
19041 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19043 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19045 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19046 swap associated with both a load and a store.
19048 2018-01-02 Andrew Waterman <andrew@sifive.com>
19050 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19051 * config/riscv/riscv.md (clear_cache): Use it.
19053 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19055 * web.c: Remove out-of-date comment.
19057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19059 * expr.c (fixup_args_size_notes): Check that any existing
19060 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19061 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19062 (emit_single_push_insn): ...here.
19064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19066 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19067 (const_vector_encoded_nelts): New function.
19068 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19069 (const_vector_int_elt, const_vector_elt): Declare.
19070 * emit-rtl.c (const_vector_int_elt_1): New function.
19071 (const_vector_elt): Likewise.
19072 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19073 of CONST_VECTOR_ELT.
19075 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19077 * expr.c: Include rtx-vector-builder.h.
19078 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19079 directly on the tree encoding.
19080 (const_vector_from_tree): Likewise.
19081 * optabs.c: Include rtx-vector-builder.h.
19082 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19083 sequence of "u" values.
19084 * vec-perm-indices.c: Include rtx-vector-builder.h.
19085 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19086 directly on the vec_perm_indices encoding.
19088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19090 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19091 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19092 * rtx-vector-builder.h: New file.
19093 * rtx-vector-builder.c: Likewise.
19094 * rtl.h (rtx_def::u2): Add a const_vector field.
19095 (CONST_VECTOR_NPATTERNS): New macro.
19096 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19097 (CONST_VECTOR_DUPLICATE_P): Likewise.
19098 (CONST_VECTOR_STEPPED_P): Likewise.
19099 (CONST_VECTOR_ENCODED_ELT): Likewise.
19100 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19101 (unwrap_const_vec_duplicate): Likewise.
19102 (const_vec_series_p): Check for a non-duplicated vector encoding.
19103 Say that the function only returns true for integer vectors.
19104 * emit-rtl.c: Include rtx-vector-builder.h.
19105 (gen_const_vec_duplicate_1): Delete.
19106 (gen_const_vector): Call gen_const_vec_duplicate instead of
19107 gen_const_vec_duplicate_1.
19108 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19109 (gen_const_vec_duplicate): Use rtx_vector_builder.
19110 (gen_const_vec_series): Likewise.
19111 (gen_rtx_CONST_VECTOR): Likewise.
19112 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19113 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19114 Build a new vector rather than modifying a CONST_VECTOR in-place.
19115 (handle_special_swappables): Update call accordingly.
19116 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19117 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19118 Build a new vector rather than modifying a CONST_VECTOR in-place.
19119 (handle_special_swappables): Update call accordingly.
19121 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19123 * simplify-rtx.c (simplify_const_binary_operation): Use
19124 CONST_VECTOR_ELT instead of XVECEXP.
19126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19128 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19129 the selector elements to be different from the data elements
19130 if the selector is a VECTOR_CST.
19131 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19132 ssizetype for the selector.
19134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19136 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19137 before testing each element individually.
19138 * tree-vect-generic.c (lower_vec_perm): Likewise.
19140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19142 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19143 * selftest-run-tests.c (selftest::run_tests): Call it.
19144 * vector-builder.h (vector_builder::operator ==): New function.
19145 (vector_builder::operator !=): Likewise.
19146 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19147 (vec_perm_indices::all_from_input_p): New function.
19148 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19149 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19150 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19151 instead of reading the VECTOR_CST directly. Detect whether both
19152 vector inputs are the same before constructing the vec_perm_indices,
19153 and update the number of inputs argument accordingly. Use the
19154 utility functions added above. Only construct sel2 if we need to.
19156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19158 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19159 the broadcast of the low byte.
19160 (expand_mult_highpart): Use an explicit encoding for the permutes.
19161 * optabs-query.c (can_mult_highpart_p): Likewise.
19162 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19163 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19164 (vectorizable_bswap): Likewise.
19165 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19166 explicit encoding for the power-of-2 permutes.
19167 (vect_permute_store_chain): Likewise.
19168 (vect_grouped_load_supported): Likewise.
19169 (vect_permute_load_chain): Likewise.
19171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19173 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19174 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19175 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19176 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19177 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19178 (vect_gen_perm_mask_any): Likewise.
19180 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19182 * int-vector-builder.h: New file.
19183 * vec-perm-indices.h: Include int-vector-builder.h.
19184 (vec_perm_indices): Redefine as an int_vector_builder.
19185 (auto_vec_perm_indices): Delete.
19186 (vec_perm_builder): Redefine as a stand-alone class.
19187 (vec_perm_indices::vec_perm_indices): New function.
19188 (vec_perm_indices::clamp): Likewise.
19189 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19190 (vec_perm_indices::new_vector): New function.
19191 (vec_perm_indices::new_expanded_vector): Update for new
19192 vec_perm_indices class.
19193 (vec_perm_indices::rotate_inputs): New function.
19194 (vec_perm_indices::all_in_range_p): Operate directly on the
19195 encoded form, without computing elided elements.
19196 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19197 encoding. Update for new vec_perm_indices class.
19198 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19199 the given vec_perm_builder.
19200 (expand_vec_perm_var): Update vec_perm_builder constructor.
19201 (expand_mult_highpart): Use vec_perm_builder instead of
19202 auto_vec_perm_indices.
19203 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19204 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19205 or double series encoding as appropriate.
19206 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19207 vec_perm_indices instead of auto_vec_perm_indices.
19208 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19209 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19210 (vect_permute_store_chain): Likewise.
19211 (vect_grouped_load_supported): Likewise.
19212 (vect_permute_load_chain): Likewise.
19213 (vect_shift_permute_load_chain): Likewise.
19214 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19215 (vect_transform_slp_perm_load): Likewise.
19216 (vect_schedule_slp_instance): Likewise.
19217 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19218 (vectorizable_mask_load_store): Likewise.
19219 (vectorizable_bswap): Likewise.
19220 (vectorizable_store): Likewise.
19221 (vectorizable_load): Likewise.
19222 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19223 vec_perm_indices instead of auto_vec_perm_indices. Use
19224 tree_to_vec_perm_builder to read the vector from a tree.
19225 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19226 vec_perm_builder instead of a vec_perm_indices.
19227 (have_whole_vector_shift): Use vec_perm_builder and
19228 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19229 truncation to calc_vec_perm_mask_for_shift.
19230 (vect_create_epilog_for_reduction): Likewise.
19231 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19232 from auto_vec_perm_indices to vec_perm_indices.
19233 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19234 instead of changing individual elements.
19235 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19236 the vector in d.perm.
19237 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19238 from auto_vec_perm_indices to vec_perm_indices.
19239 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19240 instead of changing individual elements.
19241 (arm_vectorize_vec_perm_const): Use new_vector to install
19242 the vector in d.perm.
19243 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19244 Update vec_perm_builder constructor.
19245 (rs6000_expand_interleave): Likewise.
19246 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19247 (rs6000_expand_interleave): Likewise.
19249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19251 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19252 to qimode could truncate the indices.
19253 * optabs.c (expand_vec_perm_var): Likewise.
19255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19257 * Makefile.in (OBJS): Add vec-perm-indices.o.
19258 * vec-perm-indices.h: New file.
19259 * vec-perm-indices.c: Likewise.
19260 * target.h (vec_perm_indices): Replace with a forward class
19262 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19263 * optabs.h: Include vec-perm-indices.h.
19264 (expand_vec_perm): Delete.
19265 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19266 (expand_vec_perm_const): Declare.
19267 * target.def (vec_perm_const_ok): Replace with...
19268 (vec_perm_const): ...this new hook.
19269 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19270 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19271 * doc/tm.texi: Regenerate.
19272 * optabs.def (vec_perm_const): Delete.
19273 * doc/md.texi (vec_perm_const): Likewise.
19274 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19275 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19276 expand_vec_perm for constant permutation vectors. Assert that
19277 the mode of variable permutation vectors is the integer equivalent
19278 of the mode that is being permuted.
19279 * optabs-query.h (selector_fits_mode_p): Declare.
19280 * optabs-query.c: Include vec-perm-indices.h.
19281 (selector_fits_mode_p): New function.
19282 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19283 is defined, instead of checking whether the vec_perm_const_optab
19284 exists. Use targetm.vectorize.vec_perm_const instead of
19285 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19286 fit in the vector mode before using a variable permute.
19287 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19288 vec_perm_indices instead of an rtx.
19289 (expand_vec_perm): Replace with...
19290 (expand_vec_perm_const): ...this new function. Take the selector
19291 as a vec_perm_indices rather than an rtx. Also take the mode of
19292 the selector. Update call to shift_amt_for_vec_perm_mask.
19293 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19294 Use vec_perm_indices::new_expanded_vector to expand the original
19295 selector into bytes. Check whether the indices fit in the vector
19296 mode before using a variable permute.
19297 (expand_vec_perm_var): Make global.
19298 (expand_mult_highpart): Use expand_vec_perm_const.
19299 * fold-const.c: Includes vec-perm-indices.h.
19300 * tree-ssa-forwprop.c: Likewise.
19301 * tree-vect-data-refs.c: Likewise.
19302 * tree-vect-generic.c: Likewise.
19303 * tree-vect-loop.c: Likewise.
19304 * tree-vect-slp.c: Likewise.
19305 * tree-vect-stmts.c: Likewise.
19306 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19308 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19309 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19310 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19311 (aarch64_vectorize_vec_perm_const): ...this new function.
19312 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19313 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19314 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19315 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19316 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19317 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19318 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19320 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19321 check for NEON modes.
19322 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19323 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19324 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19325 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19327 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19328 the old VEC_PERM_CONST conditions.
19329 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19330 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19331 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19332 (ia64_vectorize_vec_perm_const_ok): Merge into...
19333 (ia64_vectorize_vec_perm_const): ...this new function.
19334 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19335 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19336 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19337 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19338 * config/mips/mips.c (mips_expand_vec_perm_const)
19339 (mips_vectorize_vec_perm_const_ok): Merge into...
19340 (mips_vectorize_vec_perm_const): ...this new function.
19341 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19342 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19343 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19344 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19345 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19346 (rs6000_expand_vec_perm_const): Delete.
19347 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19349 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19350 (altivec_expand_vec_perm_const_le): Take each operand individually.
19351 Operate on constant selectors rather than rtxes.
19352 (altivec_expand_vec_perm_const): Likewise. Update call to
19353 altivec_expand_vec_perm_const_le.
19354 (rs6000_expand_vec_perm_const): Delete.
19355 (rs6000_vectorize_vec_perm_const_ok): Delete.
19356 (rs6000_vectorize_vec_perm_const): New function.
19357 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19358 an element count and rtx array.
19359 (rs6000_expand_extract_even): Update call accordingly.
19360 (rs6000_expand_interleave): Likewise.
19361 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19362 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19363 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19364 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19365 (rs6000_expand_vec_perm_const): Delete.
19366 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19367 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19368 (altivec_expand_vec_perm_const_le): Take each operand individually.
19369 Operate on constant selectors rather than rtxes.
19370 (altivec_expand_vec_perm_const): Likewise. Update call to
19371 altivec_expand_vec_perm_const_le.
19372 (rs6000_expand_vec_perm_const): Delete.
19373 (rs6000_vectorize_vec_perm_const_ok): Delete.
19374 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19375 reference to the SPE evmerge intructions.
19376 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19377 an element count and rtx array.
19378 (rs6000_expand_extract_even): Update call accordingly.
19379 (rs6000_expand_interleave): Likewise.
19380 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19381 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19383 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19387 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19388 vector mode and that that mode matches the mode of the data
19390 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19391 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19392 directly using expand_vec_perm_1 when forcing selectors into
19394 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19396 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19398 * optabs-query.h (can_vec_perm_p): Delete.
19399 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19400 * optabs-query.c (can_vec_perm_p): Split into...
19401 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19402 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19403 particular selector is valid.
19404 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19405 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19406 (vect_grouped_load_supported): Likewise.
19407 (vect_shift_permute_load_chain): Likewise.
19408 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19409 (vect_transform_slp_perm_load): Likewise.
19410 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19411 (vectorizable_bswap): Likewise.
19412 (vect_gen_perm_mask_checked): Likewise.
19413 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19414 implementations of variable permutation vectors into account
19415 when deciding which selector to use.
19416 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19417 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19418 with a false third argument.
19419 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19420 to test whether the constant selector is valid and can_vec_perm_var_p
19421 to test whether a variable selector is valid.
19423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19425 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19426 * optabs-query.c (can_vec_perm_p): Likewise.
19427 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19428 instead of vec_perm_indices.
19429 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19430 (vect_gen_perm_mask_checked): Likewise,
19431 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19432 (vect_gen_perm_mask_checked): Likewise,
19434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19436 * optabs-query.h (qimode_for_vec_perm): Declare.
19437 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19438 (qimode_for_vec_perm): ...this new function.
19439 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19441 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19443 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19444 does not have a conditional at the top.
19446 2018-01-02 Richard Biener <rguenther@suse.de>
19448 * ipa-inline.c (big_speedup_p): Fix expression.
19450 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19453 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19456 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19460 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19461 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19462 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19463 cond_taken_branch_cost 3->4.
19465 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19467 PR tree-optimization/83581
19468 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19469 TODO_cleanup_cfg if any changes have been made.
19471 PR middle-end/83608
19472 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19473 convert_modes if target mode has the right side, but different mode
19476 PR middle-end/83609
19477 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19478 last argument when extracting from CONCAT. If either from_real or
19479 from_imag is NULL, use expansion through memory. If result is not
19480 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19481 the parts directly to inner mode, if even that fails, use expansion
19484 PR middle-end/83623
19485 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19486 check for bswap in mode rather than HImode and use that in expand_unop
19489 Copyright (C) 2018 Free Software Foundation, Inc.
19491 Copying and distribution of this file, with or without modification,
19492 are permitted in any medium without royalty provided the copyright
19493 notice and this notice are preserved.