1 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4 Use __mmask64 type instead of __mmask8 for __M argument.
5 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9 __mmask16 for __M argument.
10 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
11 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12 to __mmask16 instead of __mmask8.
13 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
14 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
15 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
16 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
17 instead of __mmask16 for __U argument.
18 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
19 __mmask16 instead of __mmask8 for __U argument.
20 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
22 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
24 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
26 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
28 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
30 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
32 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
34 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
36 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
38 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
40 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
41 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
42 return type as well as __M argument type and all casts from __mmask8
44 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
45 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
46 return type as well as __M argument type and all casts from __mmask8
48 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
49 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
50 return type as well as __M argument type and all casts from __mmask8
52 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
53 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
54 return type as well as __M argument type and all casts from __mmask8
56 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
57 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
60 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
62 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
63 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
66 2018-07-11 Paul Koning <ni1d@arrl.net>
68 * doc/md.texi (define_subst): Document how multiple occurrences of
69 the same argument in the replacement pattern are handled.
71 2018-07-11 Paul Koning <ni1d@arrl.net>
73 * doc/extend.texi (Common Variable Attributes): Move "mode" into
75 (Common Type Attributes): Add "mode" attribute.
77 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
79 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
80 stream DECL_ORIGINAL_TYPE.
81 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
82 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
83 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
84 Do not walk original type.
85 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
87 (write_ts_decl_non_common_tree_pointers): Do not stream
89 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
90 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
92 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
94 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
95 thread twice from the same starting edge.
97 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
99 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
100 * gimple.c (gimple_call_nonnull_result_p): ...here...
101 (gimple_call_nonnull_arg): ...and here.
102 * gimple.h (gimple_call_nonnull_result_p): New.
103 (gimple_call_nonnull_arg): New.
105 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
107 * config/arm/arm-cpus.in: Move information from fpu field of each
108 cpu definition to the isa field.
109 * config/arm/parsecpu.awk (fpu): Delete match rule.
110 (gen_comm_data): Don't add bits from the CPU's FPU entry.
112 2018-07-11 Richard Biener <rguenther@suse.de>
115 * dwarf2out.c (init_sections_and_labels): Use
116 output_asm_line_debug_info consistently.
117 (dwarf2out_early_finish): Likewise.
118 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
121 2018-07-11 Richard Biener <rguenther@suse.de>
123 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
124 Rework father_bb setting in a way to avoid propagating constants
125 multiple times on a loop body.
127 2018-07-10 Mark Wielaard <mark@klomp.org>
130 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
132 2018-07-10 Richard Biener <rguenther@suse.de>
134 * hash-map.h (hash_map::iterator::operator*): Return
135 references to key and value.
137 2018-07-10 Jakub Jelinek <jakub@redhat.com>
140 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
141 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
142 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
143 (gimplify_omp_for): For composite loops, move outer
144 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
145 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
146 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
147 TREE_LIST for both the original class iterator and the "last" helper
148 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
149 loop, remember has_decl_expr from outer composite loops for the
150 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
152 2018-07-09 Martin Sebor <msebor@redhat.com>
156 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
157 accessing implicitly initialized array elements.
158 * expr.c (string_constant): Handle string initializers of
159 character arrays within aggregates.
160 * gimple-fold.c (fold_array_ctor_reference): Add argument.
161 Store element offset. As a special case, handle zero size.
162 (fold_nonarray_ctor_reference): Same.
163 (fold_ctor_reference): Add argument. Store subobject offset.
164 * gimple-fold.h (fold_ctor_reference): Add argument.
166 2018-07-09 Paul Koning <ni1d@arrl.net>
168 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
169 (pdp11_insn_cost): New function.
170 (pdp11_md_asm_adjust): New function.
171 (TARGET_INVALID_WITHIN_DOLOOP): Define.
172 (pdp11_rtx_costs): Update to match machine better.
173 (output_addr_const_pdp11): Correct format mismatch warnings.
174 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
175 * config/pdp11/pdp11.md: General change to add base_cost and/or
176 length attributes for use by new pdp11_insn_cost function.
177 (MIN_BRANCH): Correct definition.
179 (doloop_end): Use standard pattern name for looping pattern.
180 (doloop_end_nocc): New.
181 (movsf): Add another constraint alternative.
182 (zero_extendqihi2): Add constraint alternatives for not in place
184 (zero_extendhisi2): Remove.
185 (shift patterns): Add CC handling variants.
189 (define_peephole2): New peephole to recognize mov that sets CC for
192 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
194 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
195 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
198 2018-07-09 Jakub Jelinek <jakub@redhat.com>
201 * real.c (real_nextafter): Return true if result is denormal.
203 2018-07-09 Martin Liska <mliska@suse.cz>
205 * common.opt: Add back wrongly removed attribute.
207 2018-07-09 Richard Biener <rguenther@suse.de>
210 * dwarf2out.c (gen_block_die): For an early generated DIE
211 always output high/low PC attributes.
213 2018-07-09 Tom de Vries <tdevries@suse.de>
215 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
216 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
217 onto VAR_DECL with abstract origin.
219 2018-07-07 Jim Wilson <jimw@sifive.com>
221 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
223 2018-07-07 Tom de Vries <tdevries@suse.de>
225 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
227 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
229 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
231 (combine_bound): Use wide-int overflow calculation instead of
233 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
235 * fold-const.c (int_const_binop_2): Same.
236 (extract_muldiv_1): Same.
237 (fold_div_compare): Same.
238 (fold_abs_const): Same.
240 * poly-int.h (add): Same.
244 * predict.c (predict_iv_comparison): Same.
245 * profile-count.c (slow_safe_scale_64bit): Same.
246 * simplify-rtx.c (simplify_const_binary_operation): Same.
247 * tree-chrec.c (tree_fold_binomial): Same.
248 * tree-data-ref.c (split_constant_offset_1): Same.
249 * tree-if-conv.c (idx_within_array_bound): Same.
250 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
251 * tree-ssa-phiopt.c (minmax_replacement): Same.
252 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
253 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
254 * vr-values.c (vr_values::adjust_range_with_scev): Same.
255 * wide-int.cc (wi::add_large): Same.
256 (wi::mul_internal): Same.
257 (wi::sub_large): Same.
258 (wi::divmod_internal): Same.
259 * wide-int.h: Change overflow type to overflow_type for neg, add,
260 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
261 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
262 mul_internal, divmod_internal.
263 (overflow_type): New enum.
264 (accumulate_overflow): New.
266 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
268 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
269 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
271 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
273 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
274 argument is checked for zero before entering loop, avoid checking again.
276 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
278 * gimplify.h (generic_expr_could_trap_p): Set as global function.
279 * gimplify.h (generic_expr_could_trap_p): Likwise.
280 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
282 2018-07-06 Jakub Jelinek <jakub@redhat.com>
284 PR tree-optimization/86401
285 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
286 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
287 (fold_bit_and_mask): ... here. New helper function for match.pd.
288 * fold-const.h (fold_bit_and_mask): Declare.
289 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
291 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
294 * target.def (translate_mode_attribute): New hook.
295 * targhooks.h (default_translate_mode_attribute): Declare.
296 * targhooks.c (default_translate_mode_attribute): New function.
297 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
298 * doc/tm.texi: Regenerate.
299 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
300 (rs6000_translate_mode_attribute): New function.
302 2018-07-06 Paul Koning <ni1d@arrl.net>
304 * doc/md.texi (define_split): Document DONE and FAIL.
305 (define_peephole2): Ditto.
307 2018-07-05 Jeff Law <law@redhat.com>
309 PR tree-optimization/86010
310 * tree-ssa-dse.c (compute_trims): More aggressively trim at
311 both the head and tail of mem* and str* calls.
313 2018-07-05 Jim Wilson <jimw@sifive.com>
315 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
317 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
319 * config/aarch64/aarch64-simd.md: correct flags text for
320 MIN_EXPR replacement.
322 2018-07-05 James Clarke <jrtc27@jrtc27.com>
324 * configure: Regenerated.
326 2018-07-05 Carl Love <cel@us.ibm.com>
328 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
329 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
330 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
331 VSX_BUILTIN_DOUBLEL_V4SF.
333 2018-07-05 Martin Sebor <msebor@redhat.com>
336 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
337 than its domain to compute its the upper bound of a char array.
339 2018-07-05 Nathan Sidwell <nathan@acm.org>
341 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
342 * doc/cpp.texi: Update comment.
343 * doc/tm.texi: Rebuilt.
344 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
345 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
346 * doc/extend.texi (Backwards Compatibility): Clarify it is system
347 headers affected by extern "C".
348 * system.h: Poison NO_IMPLICIT_EXTERN_C.
349 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
350 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
351 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
352 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
353 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
354 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
355 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
356 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
357 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
358 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
359 config/sparc/sp64-elf.h, config/spu/spu.h,
360 config/stormy16/stormy16.h, config/v850/v850.h,
361 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
362 define NO_IMPLICIT_EXTERN_C.
363 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
365 2018-07-05 Tamar Christina <tamar.christina@arm.com>
368 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
369 instead of GET_MODE_SIZE when comparing Units.
371 2018-07-05 Tamar Christina <tamar.christina@arm.com>
374 * rtlanal.c (set_noop_p): Constrain on mode change,
375 include hard-reg-set.h
377 2018-07-05 Tamar Christina <tamar.christina@arm.com>
379 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
381 2018-07-05 Jakub Jelinek <jakub@redhat.com>
384 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
387 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
389 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
391 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
394 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
396 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
398 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
401 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
403 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
405 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
408 2018-07-04 Martin Liska <mliska@suse.cz>
410 * tree-switch-conversion.c: Define
411 max_ratio_for_speed and max_ratio_for_size constants.
413 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
414 Martin Liska <mliska@suse.cz>
419 * common.opt: Rename align options with 'str_' prefix.
420 * common/config/i386/i386-common.c (set_malign_value): New
422 (ix86_handle_option): Use it to set -falign-* options/
423 * config/aarch64/aarch64-protos.h (struct tune_params): Change
424 type from int to string.
425 * config/aarch64/aarch64.c: Update default values from int
427 * config/alpha/alpha.c (alpha_override_options_after_change):
429 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
430 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
431 max skip conditionally.
432 * config/i386/freebsd.h (SUBALIGN_LOG): New.
433 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
434 max skip conditionally.
435 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
436 max skip conditionally.
437 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
438 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
439 max skip conditionally.
440 * config/i386/i386.c (struct ptt): Change type from int to
442 (ix86_default_align): Set default values.
443 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
444 max skip conditionally.
445 * config/i386/iamcu.h (SUBALIGN_LOG): New.
446 (ASM_OUTPUT_MAX_SKIP_ALIGN):
447 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
448 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
449 max skip conditionally.
450 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
451 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
452 * config/i386/x86-64.h (SUBALIGN_LOG): New.
453 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
454 max skip conditionally.
455 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
456 * config/ia64/ia64.c (ia64_option_override): Set default values
457 for alignment options.
458 * config/m68k/m68k.c: Handle new str_align_* options.
459 * config/mips/mips.c (mips_set_compression_mode): Change
461 (mips_option_override): Set default values for options.
462 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
464 * config/rs6000/rs6000.c (rs6000_option_override_internal):
466 * config/rx/rx.c (rx_option_override): Likewise.
467 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
468 (LABEL_ALIGN): Use align_labels_log.
469 (LOOP_ALIGN): Use align_loops_align.
470 * config/s390/s390.c (s390_asm_output_function_label): Use new
472 * config/sh/sh.c (sh_override_options_after_change):
473 Change type of constants.
474 * config/spu/spu.c (spu_sched_init): Likewise.
475 * config/sparc/sparc.c (sparc_option_override): Set default
477 * config/visium/visium.c (visium_option_override): Likewise.
478 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
479 emit p2align format with last argument if it's not needed.
480 * doc/invoke.texi: Document extended format of -falign-*.
481 * final.c: Use align_labels alignment.
482 * flags.h (struct target_flag_state): Change type to use
484 (struct align_flags_tuple): New.
485 (struct align_flags): Likewise.
486 (align_loops_log): Redefine macro to use new types.
487 (align_loops_max_skip): Redefine macro to use new types.
488 (align_jumps_log): Redefine macro to use new types.
489 (align_jumps_max_skip): Redefine macro to use new types.
490 (align_labels_log): Redefine macro to use new types.
491 (align_labels_max_skip): Redefine macro to use new types.
492 (align_functions_log): Redefine macro to use new types.
493 (align_loops): Redefine macro to use new types.
494 (align_jumps): Redefine macro to use new types.
495 (align_labels): Redefine macro to use new types.
496 (align_functions): Redefine macro to use new types.
497 (align_functions_max_skip): Redefine macro to use new types.
498 (align_loops_value): New macro.
499 (align_jumps_value): New macro.
500 (align_labels_value): New macro.
501 (align_functions_value): New macro.
502 * function.c (invoke_set_current_function_hook): Propagate
503 alignment values from flags to global variables default in
505 * ipa-icf.c (sem_function::equals_wpa): Use
506 cl_optimization_option_eq instead of memcmp.
507 * lto-streamer.h (cl_optimization_stream_out): Support streaming
509 (cl_optimization_stream_in): Likewise.
510 * optc-save-gen.awk: Support strings in cl_optimization.
511 * opth-gen.awk: Likewise.
512 * opts.c (finish_options): Remove error checking of invalid
514 (MAX_CODE_ALIGN): Remove.
515 (MAX_CODE_ALIGN_VALUE): Likewise.
516 (parse_and_check_align_values): New function.
517 (check_alignment_argument): Likewise.
518 (common_handle_option): Use check_alignment_argument.
519 * opts.h (parse_and_check_align_values): Declare.
520 * toplev.c (init_alignments): Remove.
521 (read_log_maxskip): New.
522 (parse_N_M): Likewise.
523 (parse_alignment_opts): Likewise.
524 (backend_init_target): Remove usage of init_alignments.
525 * toplev.h (parse_alignment_opts): Declare.
526 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
528 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
529 * tree.c (cl_option_hasher::equal): New.
530 * varasm.c: Use new global macros.
532 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
534 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
535 Use a simpler align directive also if MAXSKIP = ALIGN-1.
536 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
537 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
538 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
539 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
540 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
541 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
542 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
543 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
544 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
545 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
546 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
548 2018-07-04 Martin Liska <mliska@suse.cz>
549 Jonathan Wakely <jwakely@redhat.com>
551 * coverage.c: Use correct type.
552 * doc/invoke.texi: Language correction.
554 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
557 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
558 ENDBRANCH for non-tail call which may return via indirect branch.
559 * doc/extend.texi: Document indirect_return attribute.
561 2018-07-03 Martin Sebor <msebor@redhat.com>
563 PR tree-optimization/86274
564 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
566 (format_floating): Correct handling of infinities and NaNs.
568 2018-07-03 Martin Sebor <msebor@redhat.com>
570 * print-tree.c (print_real_cst): New function.
571 (print_node_brief): Call it.
574 2018-07-03 Jeff Law <law@redhat.com>
576 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
577 into a single pattern.
579 * config/h8300/h8300.md (ors code_iterator): New.
580 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
581 a single pattern and single splitter.
582 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
583 (iorqi3_1, xorqi3_1): Likewise.
584 (iorqi3, xorqi3 expanders): Similarly.
586 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
587 (movmd_internal) into a single pattern using the P mode iterator.
588 (movmd splitters): Similarly.
589 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
590 (movsd splitters): Similarly.
592 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
593 ADDB, ADDW and ADDL into a single ADD attribute which selects the
594 right table based on the size of the operand.
595 * config/h8300/h8300.md (length_table): Corresponding changes. All
596 references to "addb", "addw" and "addl" changed to "add".
597 (btst patterns): Merge two variants into a single pattern.
598 (tstqi, tsthi): Likewise.
599 (addhi3_incdec, addsi3_incdec): Likewise.
600 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
601 (mulhi3, mulsi3): Likewise.
602 (udivhi3, udivsi3): Likewise.
603 (divhi3, divsi3): Likewise.
604 (andorqi3, andorhi3, andorsi3): Likewise.
606 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
609 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
610 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
612 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
614 PR tree-optimization/85694
615 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
616 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
618 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
619 (<u>avg<mode>3_ceil): New patterns.
621 2018-07-03 David Malcolm <dmalcolm@redhat.com>
623 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
624 scan-tree-dump directive.
625 * gcc.dg/vect/slp-perm-2.c: Likewise.
626 * gcc.dg/vect/slp-perm-3.c: Likewise.
627 * gcc.dg/vect/slp-perm-5.c: Likewise.
628 * gcc.dg/vect/slp-perm-6.c: Likewise.
629 * gcc.dg/vect/slp-perm-7.c: Likewise.
630 * gcc.dg/vect/slp-perm-8.c: Likewise.
632 2018-07-03 Marek Polacek <polacek@redhat.com>
635 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
637 2018-07-03 Richard Biener <rguenther@suse.de>
640 * tree-ssa-structalias.c (find_func_clobbers): Properly
641 handle indirect calls.
643 2018-07-03 Jeff Law <law@redhat.com>
645 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
646 (shifts): New code iterator.
647 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
648 expander. Fix HImode handling on H8/SX.
649 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
650 (subqi3, subhi3, subsi3 expanders): Likewise.
651 (andqi3, andhi3, andsi3 expanders): Likewise.
652 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
653 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
654 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
655 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
656 (zero_extendqihi2, zero_extendqisi2): Likewise.
657 (extendqihi2, extendqisi2): Likewise.
658 (rotlqi3, rotlhi3, rotlsi3): Likewise.
659 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
660 (rotlqi3_1, rotlhi3_1): Likewise.
661 (logicalhi3_sn, logicalsi3_sn): Likewise.
662 (logicalhi3, logicalsi3): Likewise.
664 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
666 * tree-vect-patterns.c (vect_recog_rotate_pattern)
667 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
668 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
669 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
670 type to append_pattern_def_seq instead of creating a stmt_vec_info
672 (build_mask_conversion): Likewise. Remove vinfo argument.
673 (vect_add_conversion_to_patterm): Likewise, renaming to...
674 (vect_add_conversion_to_pattern): ...this.
675 (vect_recog_mask_conversion_pattern): Update call to
676 build_mask_conversion. Pass the vector type to
677 append_pattern_def_seq here too.
678 (vect_recog_gather_scatter_pattern): Update call to
679 vect_add_conversion_to_pattern.
681 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
683 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
684 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
685 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
686 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
687 STMT_VINFO_PATTERN_DEF_SEQ to null here.
688 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
689 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
690 append_pattern_def_seq instead of new_pattern_def_seq.
691 (vect_recog_divmod_pattern): Do both of the above.
692 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
695 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
697 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
698 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
699 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
700 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
701 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
702 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
703 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
704 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
705 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
706 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
707 parameter with a single stmt_vec_info.
708 (vect_recog_func_ptr): Likewise.
709 (vect_recog_gather_scatter_pattern): Likewise, folding in...
710 (vect_try_gather_scatter_pattern): ...this.
711 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
712 the stmt_vec_info of the statement to be matched. Don't clear
713 STMT_VINFO_RELATED_STMT.
714 (vect_pattern_recog): Update call accordingly.
716 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
718 PR tree-optimization/85694
719 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
720 (uavgM3_ceil): Document new optabs.
721 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
722 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
724 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
725 (savg_ceil_optab): New optabs.
726 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
727 (vect_vect_recog_func_ptrs): Add it.
728 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
729 constant directly from the associated lhs.
731 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
733 * tree-vect-patterns.c (vect_split_statement): New function.
734 (vect_convert_input): Use it to try to split an existing cast.
736 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
738 * poly-int.h (print_hex): New function.
739 * dumpfile.h (dump_dec, dump_hex): Declare.
740 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
741 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
742 min_input_precision, operation_precision and operation_sign.
743 * tree-vect-patterns.c (vect_get_range_info): New function.
744 (vect_same_loop_or_bb_p, vect_single_imm_use)
745 (vect_operation_fits_smaller_type): Delete.
746 (vect_look_through_possible_promotion): Add an optional
747 single_use_p parameter.
748 (vect_recog_over_widening_pattern): Rewrite to use new
749 stmt_vec_info infomration. Handle one operation at a time.
750 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
751 (vect_truncatable_operation_p, vect_set_operation_type)
752 (vect_set_min_input_precision): New functions.
753 (vect_determine_min_output_precision_1): Likewise.
754 (vect_determine_min_output_precision): Likewise.
755 (vect_determine_precisions_from_range): Likewise.
756 (vect_determine_precisions_from_users): Likewise.
757 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
758 (vect_vect_recog_func_ptrs): Put over_widening first.
760 (vect_pattern_recog): Call vect_determine_precisions.
762 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
764 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
765 statements that have been replaced by further pattern statements.
766 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
768 2018-07-03 Richard Biener <rguenther@suse.de>
770 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
771 always set *dt. Dump vectype in vectype overload.
772 * dumpfile.h (dump_gimple_expr): New function.
773 (dump_gimple_expr_loc): Likewise.
774 * dumpfile.c (dump_gimple_expr): New function.
775 (dump_gimple_expr_loc): Likewise.
777 2018-07-02 Jeff Law <law@redhat.com>
779 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
780 the H8/300, H8/300H and H8/S variants into a single pattern.
781 (movhi_h8300, movqi_h8300hs): Similarly.
782 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
783 (QHI mode iterator): New.
785 * config/h8300/h8300.md: Remove trailing whitespace.
787 2018-07-02 Jim Wilson <jimw@sifive.com>
789 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
790 instead of emit_insn for interrupt returns.
791 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
792 (riscv_sret, riscv_uret): Likewise.
794 2018-07-02 David Malcolm <dmalcolm@redhat.com>
796 * pretty-print.c (selftest::test_pp_format): Move save and restore
797 of quotes to class auto_fix_quotes, and add an instance.
798 * selftest.c: Include "intl.h".
799 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
800 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
801 * selftest.h (selftest::auto_fix_quotes): New class.
803 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
805 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
806 (aarch64_sve_prepare_conditional_op): Remove.
807 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
808 Allow aarch64_simd_reg_or_zero as select operand; remove
809 the aarch64_sve_prepare_conditional_op call.
810 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
811 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
812 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
813 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
814 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
815 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
816 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
817 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
818 and a splitters to match all of the *_any patterns.
819 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
821 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
822 (SVE_COND_FP_BINARY_REV): Remove.
823 (sve_int_op_rev, sve_fp_op_rev): New.
824 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
825 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
826 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
827 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
828 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
829 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
830 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
831 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
832 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
834 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
835 Remove match_dup 1 from the inner unspec.
836 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
838 * config/aarch64/aarch64.md (movprfx): New attr.
839 (length): Default movprfx to 8.
840 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
841 (*madd<SVE_I>, *msub<SVE_I): Likewise.
842 (*<su>mul<SVE_I>3_highpart): Likewise.
843 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
844 (*v<ASHIFT><SVE_I>3): Likewise.
845 (*<su><MAXMIN><SVE_I>3): Likewise.
846 (*<su><MAXMIN><SVE_F>3): Likewise.
847 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
848 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
849 (*div<SVE_F>4): Likewise.
851 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
853 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
856 2018-07-02 Richard Biener <rguenther@suse.de>
858 PR tree-optimization/86363
859 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
860 memset argument refers to a non-variable address.
862 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
864 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
865 {PLUS,MINUS}_EXPR code to...
866 (adjust_symbolic_bound): ...here,
867 (combine_bound): ...here,
868 (set_value_range_with_overflow): ...and here.
870 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
872 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
874 (extract_range_from_abs_expr): ...here.
876 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
878 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
879 -fno-omit-frame-pointer when not optimizing.
881 2018-07-02 Martin Liska <mliska@suse.cz>
884 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
885 (propagate_nothrow): Likewise.
887 2018-07-02 Martin Liska <mliska@suse.cz>
890 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
892 2018-07-02 David Malcolm <dmalcolm@redhat.com>
894 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
895 function in r262149, changing "loc" param from source_location to
896 const dump_location_t &.
897 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
898 declaration, as above.
900 2018-07-01 Paul Koning <ni1d@arrl.net>
902 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
903 -munit-asm, -mgnu-asm, -mdec-asm.
904 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
905 (pdp11_output_labelref): New.
906 (pdp11_output_def): New.
907 (pdp11_output_addr_vec_elt): New.
908 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
909 %# and %@ format codes.
910 (pdp11_option_override): New.
911 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
912 (pdp11_output_ident): New.
913 (pdp11_asm_named_section): New.
914 (pdp11_asm_init_sections): New.
915 (pdp11_file_start): New.
916 (pdp11_file_end): New.
917 (output_ascii): Use .ascii/.asciz for -mdec-asm.
918 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
919 %o, like %c but octal.
920 (pdp11_option_override): New.
921 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
923 (DATA_SECTION_ASM_OP): Ditto.
924 (READONLY_DATA_SECTION_ASM_OP): New.
925 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
926 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
927 (ASM_OUTPUT_LABELREF): Ditto.
928 (ASM_OUTPUT_DEF): Ditto.
929 (ASM_OUTPUT_EXTERNAL): New.
930 (ASM_OUTPUT_SOURCE_FILENAME): New.
931 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
932 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
933 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
934 %# and %@ format codes.
935 * config/pdp11/pdp11.opt (mgnu-asm): New.
936 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
937 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
938 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
940 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
942 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
943 dereferencing path[] beyond its length.
945 (debug_all_paths): New.
946 (rewire_first_differing_edge): New.
947 (adjust_paths_after_duplication): New.
948 (duplicate_thread_path): Call adjust_paths_after_duplication.
950 (thread_through_all_blocks): Add new argument to
951 duplicate_thread_path.
953 2018-06-30 Jim Wilson <jimw@sifive.com>
955 * config/riscv/predicates.md (p2m1_shift_operand): New.
956 (high_mask_shift_operand): New.
957 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
958 pattern using p2m1_shift_operand.
959 (lshsi3_zero_extend_3+2): New combiner pattern using
960 high_mask_shift_operand.
962 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
964 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
966 (vect_recog_rotate_pattern): ...here.
967 (vect_convert_input): Try to insert casts of invariants in the
969 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
970 preheader to be empty.
972 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
974 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
975 vector type. If given, install it in the new statement's
977 (vect_element_precision): New function.
978 (vect_unpromoted_value): New struct.
979 (vect_unpromoted_value::vect_unpromoted_value): New function.
980 (vect_unpromoted_value::set_op): Likewise.
981 (vect_look_through_possible_promotion): Likewise.
982 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
983 (vect_widened_op_tree, vect_convert_input): Likewise.
984 (vect_convert_inputs, vect_convert_output): Likewise.
985 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
986 to handle the optional cast of the multiplication result and
987 vect_widened_op_tree to detect the widened multiplication itself.
988 Do not require the input and output of promotion casts to have
989 the same sign, but base the signedness of the operation on the
990 input rather than the result. If the pattern includes two
991 promotions, check that those promotions have the same sign.
992 Do not restrict the MULT_EXPR handling to a double-width result;
993 handle quadruple-width results and wider. Use vect_convert_inputs
994 to convert the inputs to the common type.
995 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
996 to handle the optional cast of the ABS result. Also allow a sign
997 change or a sign extension between the ABS and MINUS.
998 Use vect_widened_op_tree to detect the widened subtraction and use
999 vect_convert_inputs to convert the inputs to the common type.
1000 (vect_handle_widen_op_by_const): Delete.
1001 (vect_recog_widen_op_pattern): New function.
1002 (vect_recog_widen_mult_pattern): Use it.
1003 (vect_recog_widen_shift_pattern): Likewise.
1004 (vect_recog_widen_sum_pattern): Use
1005 vect_look_through_possible_promotion to handle the promoted
1008 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1010 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1011 the containing gimple_seq *.
1012 * gimple-iterator.h (gsi_for_stmt): Declare it.
1013 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1014 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1015 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1016 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1017 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1019 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1021 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1022 statement being replaced is part of an existing pattern
1023 definition sequence, inserting the new pattern statements before
1025 (vect_pattern_recog_1): Don't return a bool. If the statement
1026 is already part of a pattern, instead apply pattern matching
1027 to the pattern definition statements. Don't clear the
1028 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1029 (vect_pattern_recog): Don't break after the first match;
1030 continue processing the pattern definition statements instead.
1031 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1033 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1035 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1036 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1037 (vect_recog_widen_sum_pattern): Use it.
1039 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1041 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1042 phi is not a pattern statement and has not been replaced by
1043 a pattern statement.
1044 * tree-vect-patterns.c (type_conversion_p): Don't check
1045 STMT_VINFO_IN_PATTERN_P.
1046 (vect_recog_vector_vector_shift_pattern): Likewise.
1047 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1048 the pattern statement rather than the original statement; check
1049 directly for a WIDEN_MULT_EXPR here.
1050 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1051 vect_is_simple_use to return the pattern statement rather
1052 than the original statement; use is_pattern_stmt_p to check
1053 for such a pattern statement.
1054 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1055 to return the pattern statement rather than the original statement;
1056 don't do the same transformation here.
1057 (vect_is_simple_use): If the defining statement has been replaced
1058 by a pattern statement, return the pattern statement instead.
1059 Remove the corresponding (local) transformation from the vectype
1062 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1064 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1065 end and default to null.
1066 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1067 (vectorizable_reduction): Update calls accordingly, dropping the
1068 gimple ** argument if the passed-back statement isn't needed.
1069 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1070 (vect_recog_rotate_pattern): Likewise.
1071 (vect_recog_mask_conversion_pattern): Likewise.
1072 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1073 (vect_mask_constant_operand_p): Likewise.
1074 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1075 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1076 (get_group_load_store_type, get_load_store_type): Likewise.
1077 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1078 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1079 (vectorizable_conversion, vectorizable_assignment): Likewise.
1080 (vectorizable_shift, vectorizable_operation): Likewise.
1081 (vectorizable_store, vect_is_simple_cond): Likewise.
1082 (vectorizable_condition, vectorizable_comparison): Likewise.
1083 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1084 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1085 and move it to the end. Cope with null def_stmt_outs.
1087 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1089 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1091 2018-06-29 Jeff Law <law@redhat.com>
1093 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1094 displacements for TARGET_V850E2V3 and newer.
1095 (TARGET_LRA_P): Remove. Defaults to LRA now.
1096 * config/v850/v850.md (sign23byte_load): Remove.
1097 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1098 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1100 2018-06-29 Martin Liska <mliska@suse.cz>
1103 * coverage.c (coverage_init): Mangle full path name.
1104 * doc/invoke.texi: Document the change.
1105 * gcov-io.c (mangle_path): New.
1106 * gcov-io.h (mangle_path): Likewise.
1107 * gcov.c (mangle_name): Use mangle_path for path mangling.
1109 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1111 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1112 if starting source register is not even.
1114 2018-06-29 Martin Liska <mliska@suse.cz>
1116 PR tree-optimization/86263
1117 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1118 Make edge redirection.
1120 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1122 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1123 (dump_scope_depth): New variable.
1124 (get_dump_scope_depth): New function.
1125 (dump_begin_scope): New function.
1126 (dump_end_scope): New function.
1127 * dumpfile.h (get_dump_scope_depth): New declaration.
1128 (dump_begin_scope): New declaration.
1129 (dump_end_scope): New declaration.
1130 (class auto_dump_scope): New class.
1131 (AUTO_DUMP_SCOPE): New macro.
1132 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1135 2018-06-29 Richard Biener <rguenther@suse.de>
1137 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1138 compute_all_dependences succeeds.
1139 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1140 exceed --param loop-max-datarefs-for-datadeps.
1142 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1144 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1146 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1149 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1150 alternative 0 in preferred_for_speed attribute.
1152 2018-06-28 Paul Koning <ni1d@arrl.net>
1154 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1155 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1156 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1157 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1158 for shift insn patterns.
1160 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1162 * cgraph.c (cgraph_node::get_body): Replace assignments to
1163 "dump_file" with calls to set_dump_file.
1164 * dumpfile.c (alt_dump_file): Make static, and group with...
1165 (alt_flags): ...this definition.
1166 (dumps_are_enabled): New variable.
1167 (refresh_dumps_are_enabled): New function.
1168 (set_dump_file): New function.
1169 (set_alt_dump_file): New function.
1170 (gcc::dump_manager::dump_start): Replace assignments to
1171 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1173 (gcc::dump_manager::dump_finish): Likewise.
1174 * dumpfile.h (alt_dump_file): Delete decl.
1175 (dumps_are_enabled): New variable decl.
1176 (set_dump_file): New function decl.
1177 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1179 * tree-nested.c (lower_nested_functions): Replace assignments to
1180 "dump_file" with calls to set_dump_file.
1182 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1184 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1185 goto_locus of each outgoing edge of each basic block.
1187 2018-06-28 Richard Biener <rguenther@suse.de>
1189 * dwarf2out.c (decl_scope_table): Remove.
1190 (push_decl_scope): Likewise.
1191 (pop_decl_scope): Likewise.
1192 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1193 (gen_struct_or_union_type_die): Likewise.
1194 (gen_tagged_type_die): Likewise.
1195 (dwarf2out_init): Do not initialize decl_scope_table.
1196 (dwarf2out_c_finalize): Do not free it.
1198 2018-06-28 Richard Biener <rguenther@suse.de>
1200 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1201 deciding whether to not re-use a DIE.
1203 2018-06-28 Richard Biener <rguenther@suse.de>
1205 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1206 DW_AT_abstract_origin attribute.
1208 2018-06-28 Martin Liska <mliska@suse.cz>
1210 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1211 Use newly introduced constants.
1212 * tree-switch-conversion.h (struct jump_table_cluster):
1213 Define max_ratio_for_size and max_ratio_for_speed.
1215 2018-06-28 Martin Liska <mliska@suse.cz>
1217 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1218 Add new checking assert to catch invalid state.
1219 (jump_table_cluster::can_be_handled): Handle single case
1221 (jump_table_cluster::is_beneficial): Bail out for such case.
1222 (bit_test_cluster::find_bit_tests):
1223 Add new checking assert to catch invalid state.
1224 (bit_test_cluster::can_be_handled): Handle single case
1226 (bit_test_cluster::is_beneficial): Bail out for such case.
1227 (switch_decision_tree::analyze_switch_statement):
1230 2018-06-28 Martin Liska <mliska@suse.cz>
1232 * common.opt: Introduce -completion option.
1233 * gcc.c (driver_handle_option): Handle it.
1234 (driver::main): Print completions if completion
1236 * opt-suggestions.c (option_proposer::get_completions):
1238 (option_proposer::suggest_completion): Likewise.
1239 (option_proposer::find_param_completions): Likewise.
1240 (verify_autocompletions): Likewise.
1241 (test_completion_valid_options): Likewise.
1242 (test_completion_valid_params): Likewise.
1243 (in_completion_p): Likewise.
1244 (empty_completion_p): Likewise.
1245 (test_completion_partial_match): Likewise.
1246 (test_completion_garbage): Likewise.
1247 (opt_proposer_c_tests): Likewise.
1248 * opt-suggestions.h: Declare new functions.
1249 * opts.c (common_handle_option): Handle OPT__completion_.
1250 * selftest-run-tests.c (selftest::run_tests): Add
1251 opt_proposer_c_tests.
1252 * selftest.c (assert_str_startswith): New.
1253 * selftest.h (assert_str_startswith): Likewise.
1254 (opt_proposer_c_tests): New.
1255 (ASSERT_STR_STARTSWITH): Likewise.
1257 2018-06-28 Martin Liska <mliska@suse.cz>
1259 * Makefile.in: Add opt-suggestions.o.
1260 * gcc-main.c: Include opt-suggestions.h.
1261 * gcc.c (driver::driver): Likewise.
1262 (driver::~driver): Remove m_option_suggestions.
1263 (driver::build_option_suggestions): Moved to option_proposer.
1264 (driver::suggest_option): Likewise.
1265 (driver::handle_unrecognized_options): Use option_proposer.
1266 * gcc.h (class driver): Add new memver m_option_proposer.
1267 * opt-suggestions.c: New file.
1268 * opt-suggestions.h: New file.
1270 2018-06-28 Martin Liska <mliska@suse.cz>
1272 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1273 (auto_string_vec::~auto_string_vec): Likewise.
1275 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1277 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1278 prevent_decl_creation_for_types fields up and add reset_location field.
1279 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1280 statement if id->reset_location is true.
1281 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1282 id->reset_location is true.
1283 (copy_phis_for_bb): Force input_location on the arguments if
1284 id->reset_location is true.
1285 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1286 is set on the function to be inlined.
1288 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1290 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1292 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1294 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1295 registers for Pmode.
1296 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1297 hard registers for the clobbered pseudo.
1299 2018-06-27 Paul Koning <ni1d@arrl.net>
1301 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1302 mutually exclusive options.
1303 * config/pdp11/constraints.md (h): New constraint.
1304 (O): Update definition to match shift code generation.
1305 (D): New constraint.
1306 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1308 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1310 (output_jump): Change arguments.
1311 (pdp11_fixed_cc_regs): New function.
1312 (pdp11_cc_mode): Ditto.
1313 (pdp11_expand_shift): Ditto.
1314 (pdp11_assemble_shift): Ditto.
1315 (pdp11_small_shift): Ditto.
1316 (pdp11_branch_cost): Remove.
1317 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1319 (pdp11_register_move_cost): Update for CC registers.
1320 (pdp11_rtx_costs): Add case for LSHIFTRT.
1321 (pdp11_output_jump): Add CCNZ mode conditional branches.
1322 (notice_update_cc_on_set): Remove.
1323 (pdp11_cc_mode): New function.
1324 (simple_memory_operand): Correct pre/post decrement case.
1325 (no_side_effect_operand): New function.
1326 (pdp11_regno_reg_class): Add CC_REGS class.
1327 (pdp11_fixed_cc_regs): New function.
1328 (pdp11_small_shift): New function.
1329 (pdp11_expand_shift): New function to expand shift insns.
1330 (pdp11_assemble_shift): New function to output shifts.
1331 (pdp11_branch_cost): Remove.
1332 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1333 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1334 (WCHAR_TYPE): Ditto.
1335 (PTRDIFF_TYPE): Ditto.
1336 (ADJUST_INSN_LENGTH): New macro.
1337 (FIXED_REGISTERS): Add CC registers.
1338 (CALL_USED_REGISTERS): Ditto.
1340 (REG_CLASS_NAMES): Ditto.
1341 (REG_CLASS_CONTENTS): Ditto.
1342 (SELECT_CC_MODE): Use new function.
1343 (TARGET_FLAGS_REGNUM): New macro.
1344 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1345 (cc0_reg_rtx): Remove.
1346 (CC_STATUS_MDEP): Remove.
1347 (CC_STATUS_MDEFP_INIT): Remove.
1348 (CC_IN_FPU): Remove.
1349 (NOTICE_UPDATE_CC): Remove.
1350 (REGISTER_NAMES): Add CC registers.
1351 (BRANCH_COST): Change to constant 1.
1352 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1354 * config/pdp11/pdp11.opt (mbcopy): Remove.
1355 (mbcopy-builtin): Remove.
1356 (mbranch-cheap): Remove.
1357 (mbranch-expensive): Remove.
1358 * config/pdp11/predicates.md (expand_shift_operand): Update to
1359 match shift code generation.
1360 (ccnz_operator): New predicate.
1361 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1362 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1363 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1365 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1366 description of O constraint.
1368 2018-06-27 Jeff Law <law@redhat.com>
1369 Austin Law <austinklaw@gmail.com>
1371 * config/v850/v850.md (addsi3_set_flags): New pattern.
1372 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1373 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1374 (zero_extendhisi2_v850_set_flags): Likewise.
1375 (zero_extendqisi2_v850_set_flags): Likewise.
1376 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1377 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1378 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1380 * config/v850/v850-protos.h (notice_update_cc): Remove.
1381 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1382 (v850_print_operand): Handle 'D' and "d".
1383 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1384 Add handling of arithmetic/logical operations compared against zero.
1385 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1386 Do not look at v850_compare_op, instead get mode from last argument.
1387 (v850_gen_compare): Remove
1388 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1389 after reload for prologue insns.
1390 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1392 (construct_save_jarl): Likewise.
1393 (TARGET_FLAGS_REGNUM): Define.
1394 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1395 (NOTICE_UPDATE_CC): Remove.
1396 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1397 than cc0. Conditionalize on reload_completed.
1398 (cmpsi_insn, setfcc_insn): Likewise.
1399 (tst1 splitter): Turn into define_and_split which sets the flags
1401 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1402 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1403 (cstoresf4, cstoredf4): Clobber the flags.
1404 (cmpsi, cmpsf, cmpdf): Remove expanders.
1405 (setf_insn): Remove pattern.
1406 (addsi3): Turn into define_and_split which clobbers the flags after
1407 reload and a suitable pattern (addsi3_clobber_flags) for use after
1409 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1410 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1411 (ashrsi3, ashrsi3_v850e2): Likewise.
1412 (bins): Clobber the flags.
1413 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1414 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1415 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1416 (call_value_internal_short, call_value_internal_long): Likewise.
1417 (callt_save_interrupt, callt_return_interrupt): Likewise.
1418 (save_interrupt, return_interrupt): Likewise.
1419 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1420 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1421 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1422 (All FP comparisons): Only allow after reload has completed.
1424 (divh, divhu): Tweak output template.
1425 (branch_z_normal, branch_z_invert): Remove
1426 (branch_nz_normal, branch_nz_invert): Likewise.
1427 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1429 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1430 * config/v850/v850.c (notice_update_cc): Remove.
1431 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1432 (CC_NO_CARRY): Likewise.
1433 (NOTICE_UPDATE_CC): Define to nothing.
1434 * config/v850/v850.md: Remove block comment on cc0 handling
1435 Remove "cc" attribute from all patterns. Remove cc_status handling
1436 from all patterns. Minor formatting fixes.
1438 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1440 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1441 (cortex-a76.cortex-a55): Likewise.
1442 * config/aarch64/aarch64-tune.md: Regenerate.
1443 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1444 cortex-a76.cortex-a55.
1446 2018-06-27 Jeff Law <law@redhat.com>
1448 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1449 (MULTILIB_DIRNAMES): Similarly.
1451 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1453 * gimple.h (gimple_return_retbnd): Delete.
1454 (gimple_return_set_retbnd): Likewise.
1455 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1456 gimple_return_set_retbnd.
1457 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1458 gimple_return_retbnd and adjust.
1459 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1460 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1461 Explicitly return NULL in a couple more cases. Move assertion
1462 on debug statements and remove unreachable code.
1463 (reset_debug_binding): Do not test id->retbnd.
1464 (expand_call_inline): Do not set it.
1466 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1468 * configure.ac: Add --disable-gcov option.
1469 * configure: Regenerate.
1470 * Makefile.in: Honour @enable_gcov@.
1471 * doc/install.texi: Document --disable-gcov.
1473 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1475 * config/arm/arm-cpus.in (cortex-a76): New entry.
1476 (cortex-a76.cortex-a55): Likewise.
1477 * config/arm/arm-tables.opt: Regenerate.
1478 * config/arm/arm-tune.md: Likewise.
1479 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1480 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1481 cortex-a76.cortex-a55.
1483 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1486 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1488 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1490 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1492 (EPILOGUE_USES): Likewise.
1494 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1496 * tree-inline.c (remap_location): New function extracted from...
1497 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1498 (copy_phis_for_bb): ...here. Call remap_location.
1499 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1501 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1503 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1504 unaligned vsx for 16B memset.
1506 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1509 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1510 ieee128_float_type_node to long_double_type_node unless
1511 TARGET_LONG_DOUBLE_128 is set.
1513 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1515 * cfgloop.c (get_loop_location): Convert return type from
1516 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1517 by implicit construction from rtx_insn *, and using
1518 dump_user_location_t::from_function_decl for the fallback case.
1519 * cfgloop.h (get_loop_location): Convert return type from
1520 location_t to dump_user_location_t.
1521 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1522 dump_printf_loc to pass in a dump_location_t rather than a
1523 location_t, via the gimple stmt.
1524 * coverage.c (get_coverage_counts): Update calls to
1525 dump_printf_loc to pass in dump_location_t rather than a
1527 * doc/optinfo.texi (Dump types): Convert example of
1528 dump_printf_loc from taking "locus" to taking "insn". Update
1529 description of the "_loc" calls to cover dump_location_t.
1530 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1532 (dump_user_location_t::dump_user_location_t): New constructors,
1533 from gimple *stmt and rtx_insn *.
1534 (dump_user_location_t::from_function_decl): New function.
1535 (dump_loc): Make static.
1536 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1537 const dump_location_t &.
1538 (dump_generic_expr_loc): Delete.
1539 (dump_printf_loc): Convert param "loc" from location_t to
1540 const dump_location_t &.
1541 (selftest::test_impl_location): New function.
1542 (selftest::dumpfile_c_tests): New function.
1543 * dumpfile.h: Include "profile-count.h".
1544 (class dump_user_location_t): New class.
1545 (struct dump_impl_location_t): New struct.
1546 (class dump_location_t): New class.
1547 (dump_printf_loc): Convert 2nd param from source_location to
1548 const dump_location_t &.
1549 (dump_generic_expr_loc): Delete.
1550 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1551 const dump_location_t &.
1552 * gimple-fold.c (fold_gimple_assign): Update call to
1553 dump_printf_loc to pass in a dump_location_t rather than a
1554 location_t, via the gimple stmt.
1555 (gimple_fold_call): Likewise.
1556 * gimple-loop-interchange.cc
1557 (loop_cand::analyze_iloop_reduction_var): Update for change to
1558 check_reduction_path.
1559 (tree_loop_interchange::interchange): Update for change to
1561 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1562 change in return-type of find_loop_location.
1563 (graphite_regenerate_ast_isl): Likewise.
1564 * graphite-optimize-isl.c (optimize_isl): Likewise.
1565 * graphite.c (graphite_transform_loops): Likewise.
1566 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1567 pass in a dump_location_t rather than a location_t, via the
1569 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1570 * ipa.c (walk_polymorphic_call_targets): Likewise.
1571 * loop-unroll.c (report_unroll): Convert "locus" param from
1572 location_t to dump_location_t.
1573 (decide_unrolling): Update for change to get_loop_location's
1575 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1576 location_t to dump_user_location_t.
1577 (grid_find_single_omp_among_assignments_1): Updates calls to
1578 dump_printf_loc to pass in a dump_location_t rather than a
1579 location_t, via the gimple stmt.
1580 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1581 location_t to dump_location_t. Updates calls to dump_printf_loc
1582 to pass in a dump_location_t rather than a location_t, via the
1584 (grid_inner_loop_gridifiable_p): Likewise.
1585 (grid_dist_follows_simple_pattern): Likewise.
1586 (grid_gfor_follows_tiling_pattern): Likewise.
1587 (grid_target_follows_gridifiable_pattern): Likewise.
1588 (grid_attempt_target_gridification): Convert initialization
1589 of local "grid" from memset to zero-initialization; FIXME: does
1590 this require C++11? Update call to dump_printf_loc to pass in a
1591 optinfo_location rather than a location_t, via the gimple stmt.
1592 * profile.c (read_profile_edge_counts): Updates call to
1593 dump_printf_loc to pass in a dump_location_t rather than a
1595 (compute_branch_probabilities): Likewise.
1596 * selftest-run-tests.c (selftest::run_tests): Call
1598 * selftest.h (dumpfile_c_tests): New decl.
1599 * tree-loop-distribution.c (pass_loop_distribution::execute):
1600 Update for change in return type of find_loop_location.
1601 * tree-parloops.c (parallelize_loops): Likewise.
1602 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1603 "locus" from location_t to dump_user_location_t.
1604 (canonicalize_loop_induction_variables): Likewise.
1605 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1606 for change in return type of find_loop_location.
1607 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1608 to dump_printf_loc to pass in a dump_location_t rather than a
1609 location_t, via the stmt.
1610 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1612 * tree-vect-loop-manip.c (find_loop_location): Convert return
1613 type from source_location to dump_user_location_t.
1614 (vect_do_peeling): Update for above change.
1615 (vect_loop_versioning): Update for change in type of
1617 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1618 from location_t to dump_user_location_t.
1619 (vect_estimate_min_profitable_iters): Update for change in type
1621 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1622 location_t to dump_location_t.
1623 (vect_slp_bb): Update for change in type of vect_location.
1624 * tree-vectorizer.c (vect_location): Convert from source_location
1625 to dump_user_location_t.
1626 (try_vectorize_loop_1): Update for change in vect_location's type.
1627 (vectorize_loops): Likewise.
1628 (increase_alignment): Likewise.
1629 * tree-vectorizer.h (vect_location): Convert from source_location
1630 to dump_user_location_t.
1631 (find_loop_location): Convert return type from source_location to
1632 dump_user_location_t.
1633 (check_reduction_path): Convert 1st param from location_t to
1634 dump_user_location_t.
1635 * value-prof.c (check_counter): Update call to dump_printf_loc to
1636 pass in a dump_user_location_t rather than a location_t; update
1637 call to error_at for change in type of "locus".
1638 (check_ic_target): Update call to dump_printf_loc to
1639 pass in a dump_user_location_t rather than a location_t, via the
1642 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1644 * config/s390/s390.h (enum processor_flags): Do not use
1645 default tune parameter when -march was specified.
1647 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1650 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1651 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1654 2018-06-26 Richard Biener <rguenther@suse.de>
1656 PR tree-optimization/86287
1658 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1659 (vect_analyze_loop): Initialize n_stmts.
1661 2018-06-26 Richard Biener <rguenther@suse.de>
1664 * fold-const.c (fold_convertible_p): Pointer extension
1667 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1670 * dwarf2out.c (loc_list_has_views): Adjust comments.
1671 (dw_loc_list): Split single cross-partition range with
1674 2018-06-25 Jeff Law <law@redhat.com>
1676 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1677 on -mbig-switch by default.
1679 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1681 (const_float_0_operand): Remove unused predicate.
1682 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1683 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1684 (recipsf2): New expander. Original pattern now called
1686 (recipdf2, recipdf2_insn): Similarly.
1687 (rsqrtsf2, rsqrtsf2_insn): Similarly
1688 (rsqrtdf2, rsqrtdf2_insn): Similarly
1690 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1692 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1693 Simplify logic for FreeBSD (twice).
1695 2018-06-25 Martin Sebor <msebor@redhat.com>
1697 PR tree-optimization/86204
1698 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1699 a strnlen result if it's less than the length of the string.
1701 2018-06-25 Martin Sebor <msebor@redhat.com>
1703 PR tree-optimization/85700
1704 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1705 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1706 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1708 2018-06-25 Martin Sebor <msebor@redhat.com>
1710 * doc/extend.texi (Zero-length arrays): Update and clarify.
1712 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1714 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1715 added IEEE/IBM long double multilib support on PowerPC little
1716 endian Linux systems.
1717 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1718 (MULTILIB_DEFAULTS): Likewise.
1719 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1721 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1722 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1723 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1725 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1728 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1729 (REORDER_45): Likewise.
1731 2018-06-25 Jeff Law <law@redhat.com>
1733 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1734 dividend to 32 bits. Adjust length.
1735 (udivmodhi4): Cleanup output template. Fix length.
1737 2018-06-25 Carl Love <cel@us.ibm.com>
1739 * config/rs6000/vsx.md: Change word selector to prefered location.
1741 2018-06-25 Richard Biener <rguenther@suse.de>
1743 PR tree-optimization/86304
1744 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1745 epilogue-if-converted loops as well.
1747 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1749 * lto-section-out.c (lto_begin_section): Do not print section
1750 name for noaddr and unnumbered dumps.
1752 2018-06-25 Richard Biener <rguenther@suse.de>
1754 * tree-vectorizer.h (struct vec_info_shared): New structure
1755 with parts split out from struct vec_info and loop_nest from
1756 struct _loop_vec_info.
1757 (struct vec_info): Adjust accordingly.
1758 (struct _loop_vec_info): Likewise.
1759 (LOOP_VINFO_LOOP_NEST): Adjust.
1760 (LOOP_VINFO_DATAREFS): Likewise.
1761 (LOOP_VINFO_DDRS): Likewise.
1762 (struct _bb_vec_info): Likewise.
1763 (BB_VINFO_DATAREFS): Likewise.
1764 (BB_VINFO_DDRS): Likewise.
1765 (struct _stmt_vec_info): Add dr_aux member.
1766 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1767 (DR_MISALIGNMENT_UNINITIALIZED): New.
1768 (set_dr_misalignment): Adjust.
1769 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1770 (vect_analyze_loop): Adjust prototype.
1771 (vect_analyze_loop_form): Likewise.
1772 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1773 Compute dependences lazily.
1774 (vect_record_base_alignments): Use shared datarefs/ddrs.
1775 (vect_verify_datarefs_alignment): Likewise.
1776 (vect_analyze_data_refs_alignment): Likewise.
1777 (vect_analyze_data_ref_accesses): Likewise.
1778 (vect_analyze_data_refs): Likewise.
1779 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1780 constructor parameter for shared part.
1781 (vect_analyze_loop_form): Pass in shared part and adjust.
1782 (vect_analyze_loop_2): Pass in storage for the number of
1783 stmts. Move loop nest finding to the caller. Compute
1785 (vect_analyze_loop): Pass in shared part.
1786 (vect_transform_loop): Verify shared datarefs are unchanged.
1787 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1788 constructor parameter for shared part.
1789 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1790 (vect_slp_bb): Verify shared datarefs are unchanged before
1792 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1794 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1795 DR_MISALIGNMENT_UNINITIALIZED.
1796 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1797 parameter for shared part.
1798 (vec_info::~vec_info): Adjust.
1799 (vec_info_shared::vec_info_shared): New.
1800 (vec_info_shared::~vec_info_shared): Likewise.
1801 (vec_info_shared::save_datarefs): Likewise.
1802 (vec_info_shared::check_datarefs): Likewise.
1803 (try_vectorize_loop_1): Construct shared part live for analyses
1804 of a single loop for multiple vector sizes.
1805 * tree-parloops.c (gather_scalar_reductions): Adjust.
1807 2018-06-25 Richard Biener <rguenther@suse.de>
1809 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1810 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1811 (vect_analyze_data_refs): Remove similar code from here and
1812 simplify accordingly.
1814 2018-06-25 Richard Biener <rguenther@suse.de>
1816 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1817 for reverse storage order accesses rather than asserting
1818 they cannot happen here.
1820 2018-06-25 Tom de Vries <tdevries@suse.de>
1823 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1824 Use data16 instead of .byte for insn prefix.
1826 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1829 * parser.c (make_char_string_pack): Pass this literal chars
1830 through cpp_interpret_string.
1831 (cp_parser_userdef_numeric_literal): Check the result of
1832 make_char_string_pack.
1834 2018-06-24 Maya Rashish <coypu@sdf.org>
1836 * ginclude/stddef.h: Simplify conditions around avoiding
1837 re-definition of __size_t.
1839 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1841 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1842 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1844 2018-06-22 Maya Rashish <coypu@sdf.org>
1846 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1847 made to OpenBSD and NetBSD through the years.
1849 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1851 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1852 behavior of vec_pack (vector double, vector double) to match
1853 behavior of vec_float2 (vector double, vector double).
1855 2018-06-22 Olivier Hainque <hainque@adacore.com>
1857 * gimplify.c (gimplify_function_tree): Prevent creation
1858 of a trampoline for the address of the current function
1859 passed to entry/exit instrumentation hooks.
1861 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1864 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1867 2018-06-22 Martin Liska <mliska@suse.cz>
1869 PR tree-optimization/86263
1870 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1871 Bail out if is_enabled is false.
1872 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1874 (jump_table_cluster::is_enabled): New function.
1876 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1878 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1879 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1880 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1881 (lto_input_ts_binfo_tree_pointers): Likewise.
1882 * tree-streamer-out.c (streamer_write_tree_bitfields,
1883 write_ts_binfo_tree_pointers): Likewise.
1884 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1886 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1888 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1890 2018-06-22 Martin Liska <mliska@suse.cz>
1892 * symbol-summary.h (get): Make it pure and inline move
1893 functionality from ::get function.
1894 (get): Remove and inline into ::get and ::get_create.
1895 (get_create): Move code from ::get function.
1897 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1900 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1901 -x assembler-with-cpp.
1903 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1905 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1906 _FILE_OFFSET_BITS=64 for C++.
1908 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1910 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1911 conversion insn that shows up when pr85657-3.c is compiled using
1912 IEEE 128-bit long double.
1913 (neg<mode>2_internal): Use the correct mode to check whether the
1914 mode is IBM extended.
1915 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1916 multiply and divide external functions from being created more
1919 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1921 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1923 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1924 the edge can be forwarded.
1925 (cfg_layout_merge_blocks): Likewise.
1927 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1929 * except.c (finish_eh_generation): Commit edge insertions only after
1930 the EH edges have been redirected from post-landing to landing pads.
1932 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1934 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1935 create_tmp_var_for to create the FRAME decl.
1936 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1938 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1940 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1941 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1942 debug statement when resetting its value.
1943 (expand_call_inline): Copy the locus of the call onto the assignment
1944 of the return value, if any. Use local variable in more cases.
1946 2018-06-21 Martin Liska <mliska@suse.cz>
1948 * ipa-pure-const.c (propagate_nothrow): Use
1949 funct_state_summaries->get.
1950 (dump_malloc_lattice): Likewise.
1951 (propagate_malloc): Likewise.
1953 2018-06-21 Richard Biener <rguenther@suse.de>
1955 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1956 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1957 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1959 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1960 BLOCK_ABSTRACT_ORIGIN unconditionally.
1962 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1964 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1966 * ipa-reference.c (ipa_reference_c_finalize): Delete
1967 ipa_ref_opt_sum_summaries and set it to NULL.
1969 2018-06-21 Tom de Vries <tdevries@suse.de>
1971 PR tree-optimization/85859
1972 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1973 test with comment from bb_no_side_effects_p.
1975 2018-06-21 Richard Biener <rguenther@suse.de>
1977 PR tree-optimization/86232
1978 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1979 max for constant niter.
1981 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1983 * config/aarch64/aarch64-simd.md
1984 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1986 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1988 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1989 Make opernads of the unspec commutative.
1991 2018-06-21 Richard Biener <rguenther@suse.de>
1993 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1994 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1995 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1996 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1998 (vect_record_base_alignments): Do not record base alignment
1999 for gathers or scatters.
2000 (vect_compute_data_ref_alignment): Drop return value that is always
2001 true. Bail out early for gathers or scatters.
2002 (vect_enhance_data_refs_alignment): Bail out early for gathers
2004 (vect_find_same_alignment_drs): Likewise.
2005 (vect_analyze_data_refs_alignment): Remove dead code.
2006 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2007 (vect_analyze_data_refs): For possible gathers or scatters do
2008 not create an alternate DR, just check their possible validity
2009 and mark them. Adjust DECL_NONALIASED handling to not rely
2011 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2012 update inits of gathers or scatters.
2013 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2014 Also copy gather/scatter flag to pattern vinfo.
2016 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2018 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2019 behavior of vec_packsu (vector unsigned long long, vector unsigned
2020 long long) to match behavior of vec_packs with same signature.
2022 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2023 Thomas Schwinge <thomas@codesourcery.com>
2024 Cesar Philippidis <cesar@codesourcery.com>
2026 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2027 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2028 (gimplify_adjust_omp_clauses): Likewise.
2029 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2030 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2031 (gimplify_omp_target_update): Update handling of acc update and
2033 * omp-low.c (install_var_field): Remove unused parameter
2034 base_pointers_restrict.
2035 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2036 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2038 (omp_target_base_pointers_restrict_p): Delete.
2039 (scan_omp_target): Update call to scan_sharing_clauses.
2040 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2042 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2043 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2044 (convert_local_omp_clauses): Likewise.
2045 * tree-pretty-print.c (dump_omp_clause): Likewise.
2046 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2048 (omp_clause_code_name): Likewise.
2050 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2053 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2056 PR tree-optimization/86231
2057 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2058 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2060 2018-06-20 Tom de Vries <tdevries@suse.de>
2062 PR tree-optimization/86097
2063 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2064 iv type if signedness of iv type is not the same as that of *nit.
2066 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2068 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2069 EDGE_EH edges, verify they are all EDGE_EH.
2071 2018-06-20 Maya Rashish <coypu@sdf.org>
2073 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2075 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2077 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2078 * config/aarch64/aarch64.c (xgene1_tunings): Add
2079 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2080 (aarch64_mode_valid_for_sched_fusion_p):
2081 Allow 16-byte modes.
2082 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2083 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2085 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2087 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2088 * config/aarch64/iterators.md (VQ2): New mode iterator.
2090 2018-06-20 Martin Liska <mliska@suse.cz>
2092 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2093 Change default ratio from 10 to 8.
2095 2018-06-20 Martin Liska <mliska@suse.cz>
2097 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2099 (bit_test_cluster::find_bit_tests): Likewise.
2100 (switch_decision_tree::analyze_switch_statement): Find clusters.
2101 * tree-switch-conversion.h (struct jump_table_cluster): Document
2104 2018-06-20 Martin Liska <mliska@suse.cz>
2106 * tree-switch-conversion.c (switch_conversion::collect):
2107 Record m_uniq property.
2108 (switch_conversion::expand): Bail out for special conditions.
2109 (group_cluster::~group_cluster): New.
2110 (group_cluster::group_cluster): Likewise.
2111 (group_cluster::dump): Likewise.
2112 (jump_table_cluster::emit): New.
2113 (switch_decision_tree::fix_phi_operands_for_edges): New.
2114 (struct case_node): Remove struct.
2115 (jump_table_cluster::can_be_handled): New.
2116 (case_values_threshold): Moved to header.
2117 (reset_out_edges_aux): Likewise.
2118 (jump_table_cluster::is_beneficial): New.
2119 (bit_test_cluster::can_be_handled): Likewise.
2120 (add_case_node): Remove.
2121 (bit_test_cluster::is_beneficial): New.
2122 (case_bit_test::cmp): New.
2123 (bit_test_cluster::emit): New.
2124 (expand_switch_as_decision_tree_p): Remove.
2125 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2126 (fix_phi_operands_for_edge): Likewise.
2127 (switch_decision_tree::analyze_switch_statement): New.
2128 (compute_cases_per_edge): Move ...
2129 (switch_decision_tree::compute_cases_per_edge): ... here.
2130 (try_switch_expansion): Likewise.
2131 (switch_decision_tree::try_switch_expansion): Likewise.
2132 (record_phi_operand_mapping): Likewise.
2133 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2134 (emit_case_decision_tree): Likewise.
2135 (switch_decision_tree::emit): Likewise.
2136 (balance_case_nodes): Likewise.
2137 (switch_decision_tree::balance_case_nodes): Likewise.
2138 (dump_case_nodes): Likewise.
2139 (switch_decision_tree::dump_case_nodes): Likewise.
2140 (emit_jump): Likewise.
2141 (switch_decision_tree::emit_jump): Likewise.
2142 (emit_cmp_and_jump_insns): Likewise.
2143 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2144 (emit_case_nodes): Likewise.
2145 (switch_decision_tree::emit_case_nodes): Likewise.
2146 (conditional_probability): Remove.
2147 * tree-switch-conversion.h (enum cluster_type): New.
2149 (struct cluster): Likewise.
2150 (cluster::cluster): Likewise.
2151 (struct simple_cluster): Likewise.
2152 (simple_cluster::simple_cluster): Likewise.
2153 (struct group_cluster): Likewise.
2154 (struct jump_table_cluster): Likewise.
2155 (struct bit_test_cluster): Likewise.
2156 (struct min_cluster_item): Likewise.
2157 (struct case_tree_node): Likewise.
2158 (case_tree_node::case_tree_node): Likewise.
2159 (jump_table_cluster::case_values_threshold): Likewise.
2160 (struct case_bit_test): Likewise.
2161 (struct switch_decision_tree): Likewise.
2162 (struct switch_conversion): Likewise.
2163 (switch_decision_tree::reset_out_edges_aux): Likewise.
2165 2018-06-20 Martin Liska <mliska@suse.cz>
2167 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2168 (hoist_edge_and_branch_if_true): Likewise.
2169 (expand_switch_using_bit_tests_p): Likewise.
2170 (struct case_bit_test): Likewise.
2171 (case_bit_test_cmp): Likewise.
2172 (emit_case_bit_tests): Likewise.
2173 (switch_conversion::switch_conversion): New class.
2174 (struct switch_conv_info): Remove old struct.
2175 (collect_switch_conv_info): More to ...
2176 (switch_conversion::collect): ... this.
2177 (check_range): Likewise.
2178 (switch_conversion::check_range): Likewise.
2179 (check_all_empty_except_final): Likewise.
2180 (switch_conversion::check_all_empty_except_final): Likewise.
2181 (check_final_bb): Likewise.
2182 (switch_conversion::check_final_bb): Likewise.
2183 (create_temp_arrays): Likewise.
2184 (switch_conversion::create_temp_arrays): Likewise.
2185 (free_temp_arrays): Likewise.
2186 (gather_default_values): Likewise.
2187 (switch_conversion::gather_default_values): Likewise.
2188 (build_constructors): Likewise.
2189 (switch_conversion::build_constructors): Likewise.
2190 (constructor_contains_same_values_p): Likewise.
2191 (switch_conversion::contains_same_values_p): Likewise.
2192 (array_value_type): Likewise.
2193 (switch_conversion::array_value_type): Likewise.
2194 (build_one_array): Likewise.
2195 (switch_conversion::build_one_array): Likewise.
2196 (build_arrays): Likewise.
2197 (switch_conversion::build_arrays): Likewise.
2198 (gen_def_assigns): Likewise.
2199 (switch_conversion::gen_def_assigns): Likewise.
2200 (prune_bbs): Likewise.
2201 (switch_conversion::prune_bbs): Likewise.
2202 (fix_phi_nodes): Likewise.
2203 (switch_conversion::fix_phi_nodes): Likewise.
2204 (gen_inbound_check): Likewise.
2205 (switch_conversion::gen_inbound_check): Likewise.
2206 (process_switch): Use the newly created class.
2207 (switch_conversion::expand): New.
2208 (switch_conversion::~switch_conversion): New.
2209 * tree-switch-conversion.h: New file.
2211 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2213 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2214 tree-vect-patterns.c.
2215 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2216 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2217 (vect_recog_sad_pattern): Likewise.
2218 (vect_recog_widen_sum_pattern): Likewise.
2219 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2220 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2221 (vect_recog_rotate_pattern): Likewise.
2222 (vect_recog_mult_pattern): Likewise.
2223 (vect_recog_vector_vector_shift_pattern): Likewise.
2224 (vect_recog_divmod_pattern): Likewise.
2225 (vect_recog_mixed_size_cond_pattern): Likewise.
2226 (vect_recog_bool_pattern): Likewise.
2227 (vect_recog_mask_conversion_pattern): Likewise.
2228 (vect_try_gather_scatter_pattern): Likewise.
2229 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2230 (vect_recog_over_widening_pattern): Likewise.
2231 (vect_recog_gather_scatter_pattern): Likewise.
2232 (vect_recog_func_ptr): Move from tree-vectorizer.h
2233 (vect_vect_recog_func_ptrs): Move further down the file.
2234 (vect_recog_func): Likewise. Remove the third argument.
2235 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2236 (vect_pattern_recog_1): Expect the pattern function to do any
2237 necessary target tests. Also expect it to provide a vector type.
2238 Remove the type_in handling.
2240 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2242 * tree-vect-patterns.c (vect_pattern_detected): New function.
2243 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2244 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2245 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2246 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2247 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2248 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2249 (vect_recog_mask_conversion_pattern)
2250 (vect_try_gather_scatter_pattern): Likewise.
2252 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2254 * tree-vect-patterns.c (vect_get_internal_def): New function.
2255 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2256 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2257 (search_type_for_mask_1): Use it.
2259 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2261 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2262 redundant WIDEN_SUM_EXPR handling.
2263 (vect_recog_sad_pattern): Likewise.
2265 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2267 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2268 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2269 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2270 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2271 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2273 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2275 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2276 use the stmt_vec_info of the original bb statement for the
2277 new zero assignment, even if the call is part of a pattern.
2279 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2281 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2282 that the sequence is attached to the original statement rather
2283 than the pattern statement.
2284 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2285 PATTERN_DEF_SEQ from the original statement rather than
2286 the main pattern statement.
2287 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2288 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2289 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2291 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2293 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2294 definition statements before the early exit for statements that aren't
2296 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2298 (vect_transform_loop): ...here. Process pattern definition
2299 statements without first checking whether the main pattern
2300 statement is live or relevant.
2302 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2304 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2305 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2307 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2309 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2310 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2311 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2312 (expand_block_compare): Change select_block_compare_mode call.
2313 (expand_strncmp_align_check): Use new functions, fix comment.
2314 (emit_final_str_compare_gpr): New function.
2315 (expand_strn_compare): Refactor and clean up code.
2316 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2318 2018-06-19 Tony Reix <tony.reix@atos.com>
2319 Damien Bergamini <damien.bergamini@atos.com>
2320 David Edelsohn <dje.gcc@gmail.com>
2322 * collect2.c (static_obj): New variable.
2323 (static_libs): New variable.
2324 (is_in_list): Uncomment declaration.
2325 (main): Track AIX libraries linked statically.
2326 (is_in_list): Uncomment definition.
2327 (scan_prog_file): Don't add AIX shared libraries initializer
2328 to constructor list if linking statically.
2330 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2332 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2334 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2336 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2338 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2341 2018-06-19 Martin Liska <mliska@suse.cz>
2343 * config/i386/i386.c (ix86_can_inline_p): Do not use
2344 ipa_fn_summaries::get_create.
2345 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2347 (devirtualization_time_bonus): Likewise.
2348 (ipcp_propagate_stage): Likewise.
2349 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2350 (edge_set_predicate): Likewise.
2351 (evaluate_conditions_for_known_args): Likewise.
2352 (evaluate_properties_for_edge): Likewise.
2353 (ipa_call_summary::reset): Tranform to ...
2354 (ipa_call_summary::~ipa_call_summary): ... this.
2355 (ipa_fn_summary::reset): Transform to ...
2356 (ipa_fn_summary::~ipa_fn_summary): ... this.
2357 (ipa_fn_summary_t::remove): Rename to ...
2358 (ipa_fn_summary_t::remove_callees): ... this.
2359 (ipa_fn_summary_t::duplicate): Use placement new
2360 instead of memory copy.
2361 (ipa_call_summary_t::duplicate): Likewise.
2362 (ipa_call_summary_t::remove): Remove.
2363 (dump_ipa_call_summary): Change get_create to get.
2364 (ipa_dump_fn_summary): Dump only when summary exists.
2365 (analyze_function_body): Use symbol_summary::get instead
2367 (compute_fn_summary): Likewise.
2368 (estimate_edge_devirt_benefit): Likewise.
2369 (estimate_edge_size_and_time): Likewise.
2370 (inline_update_callee_summaries): Likewise.
2371 (remap_edge_change_prob): Likewise.
2372 (remap_edge_summaries): Likewise.
2373 (ipa_merge_fn_summary_after_inlining): Likewise.
2374 (write_ipa_call_summary): Likewise.
2375 (ipa_fn_summary_write): Likewise.
2376 (ipa_free_fn_summary): Likewise.
2377 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2378 (struct ipa_call_summary): Likewise.
2379 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2381 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2382 (estimate_size_after_inlining): Likewise.
2383 (estimate_growth): Likewise.
2384 (growth_likely_positive): Likewise.
2385 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2386 (inline_call): Likewise.
2387 * ipa-inline.c (caller_growth_limits): Likewise.
2388 (can_inline_edge_p): Likewise.
2389 (can_inline_edge_by_limits_p): Likewise.
2390 (compute_uninlined_call_time): Likewise.
2391 (compute_inlined_call_time): Likewise.
2392 (want_inline_small_function_p): Likewise.
2393 (edge_badness): Likewise.
2394 (update_caller_keys): Likewise.
2395 (update_callee_keys): Likewise.
2396 (inline_small_functions): Likewise.
2397 (inline_to_all_callers_1): Likewise.
2398 (dump_overall_stats): Likewise.
2399 (early_inline_small_functions): Likewise.
2400 (early_inliner): Likewise.
2401 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2402 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2403 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2404 * ipa-split.c (execute_split_functions): Likewise.
2405 * symbol-summary.h: Likewise.
2406 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2408 2018-06-19 Richard Biener <rguenther@suse.de>
2410 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2411 (vectorize_loops): ... here. Fix dbgcnt handling.
2412 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2414 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2417 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2418 ieee128 argument takes up only one (vector) register, not two (floating
2421 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2423 * gimplify.c (gimplify_init_constructor): Really never clear for an
2424 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2426 2018-06-19 Richard Biener <rguenther@suse.de>
2428 PR tree-optimization/86179
2429 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2430 after failed recognition.
2432 2018-06-18 Martin Sebor <msebor@redhat.com>
2435 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2436 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2437 Handle integer subtraction.
2438 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2439 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2441 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2443 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2444 param from rtx to rtx_insn *.
2445 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2447 (frv_ifcvt_modify_insn): Likwise.
2448 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2449 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2450 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2451 as_a <rtx_insn *> cast to local "unprotected_region" once
2452 it's been established that it's not NULL or pc_rtx.
2453 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2454 param "sethi" from rtx to rtx_insn *.
2455 (nds32_group_float_insns): Likewise for param "insn".
2456 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2458 (vax_output_int_subtract): Likewise.
2459 * config/vax/vax.c (vax_output_int_add): Likewise for param
2461 (vax_output_int_subtract): Likewise.
2462 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2463 (emit_pattern_after): Likewise for param "after".
2464 (emit_insn_after): Likewise.
2465 (emit_jump_insn_after): Likewise.
2466 (emit_call_insn_after): Likewise.
2467 (emit_debug_insn_after): Likewise.
2468 (emit_pattern_before): Likewise for param "before".
2469 (emit_insn_before): Likewise.
2470 (emit_jump_insn_before): Likewise.
2471 * final.c (get_insn_template): Likewise for param "insn", removing
2473 * output.h (get_insn_template): Likewise for 2nd param.
2474 * rtl.h (emit_insn_before): Likewise.
2475 (emit_jump_insn_before): Likewise.
2476 (emit_debug_insn_before_noloc): Likewise.
2477 (emit_insn_after): Likewise.
2478 (emit_jump_insn_after): Likewise.
2479 (emit_call_insn_after): Likewise.
2480 (emit_debug_insn_after): Likewise.
2481 (set_insn_deleted): Likewise for param.
2483 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2486 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2487 floating point modes, so that IFmode is numerically greater than
2488 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2489 to declare the ordering. This prevents IFmode from being
2490 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2491 machine. Include rs6000-modes.h to share the fractional values
2492 between genmodes* and the rest of the compiler.
2496 * config/rs6000/rs6000-modes.h: New file.
2497 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2498 meaning of rs6000_long_double_size so that 126..128 selects an
2499 appropriate 128-bit floating point type.
2500 (rs6000_option_override_internal): Likewise.
2501 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2502 (TARGET_LONG_DOUBLE_128): Change the meaning of
2503 rs6000_long_double_size so that 126..128 selects an appropriate
2504 128-bit floating point type.
2505 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2506 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2507 source and destination to match the standard usage.
2508 (truncifkf2): Likewise.
2509 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2510 ISA 2.07 to use an explicit clobber, instead of passing in a
2512 (copysign<mode>3_soft): Likewise.
2514 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2516 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2517 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2518 (vect_slp_analyze_instance_dependence): Likewise.
2519 (vect_enhance_data_refs_alignment): Likewise.
2520 (vect_analyze_data_refs_alignment): Likewise.
2521 (vect_slp_analyze_and_verify_instance_alignment
2522 (vect_analyze_data_ref_accesses): Likewise.
2523 (vect_prune_runtime_alias_test_list): Likewise.
2524 (vect_analyze_data_refs): Likewise.
2525 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2526 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2527 (vect_analyze_scalar_cycles_1): Likewise.
2528 (vect_get_loop_niters): Likewise.
2529 (vect_analyze_loop_form_1): Likewise.
2530 (vect_update_vf_for_slp): Likewise.
2531 (vect_analyze_loop_operations): Likewise.
2532 (vect_analyze_loop): Likewise.
2533 (vectorizable_induction): Likewise.
2534 (vect_transform_loop): Likewise.
2535 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2536 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2537 (vect_make_slp_decision): Likewise.
2538 (vect_detect_hybrid_slp): Likewise.
2539 (vect_slp_analyze_operations): Likewise.
2540 (vect_slp_bb): Likewise.
2541 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2542 (vectorizable_bswap): Likewise.
2543 (vectorizable_call): Likewise.
2544 (vectorizable_simd_clone_call): Likewise.
2545 (vectorizable_conversion): Likewise.
2546 (vectorizable_assignment): Likewise.
2547 (vectorizable_shift): Likewise.
2548 (vectorizable_operation): Likewise.
2549 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2551 2018-06-18 Martin Sebor <msebor@redhat.com>
2553 PR tree-optimization/81384
2554 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2555 * builtins.c (expand_builtin_strnlen): New function.
2556 (expand_builtin): Call it.
2557 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2558 * builtins.def (BUILT_IN_STRNLEN): New.
2559 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2560 Warn for bounds in excess of maximum object size.
2561 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2562 single-value ranges. Handle strnlen.
2563 (handle_builtin_strlen): Handle strnlen.
2564 (strlen_check_and_optimize_stmt): Same.
2565 * doc/extend.texi (Other Builtins): Document strnlen.
2567 2018-06-18 Maya Rashish <coypu@sdf.org>
2569 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2570 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2571 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2573 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2575 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2577 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2579 * tree.c (escaped_string::escape): Replace cast to char * by
2580 const_cast<char *> (unescaped).
2582 2018-06-18 Nick Clifton <nickc@redhat.com>
2585 * tree.c (escaped_string): New class. Converts an unescaped
2586 string into its escaped equivalent.
2587 (warn_deprecated_use): Use the new class to convert the
2588 deprecation message, if present.
2589 (test_escaped_strings): New self test.
2590 (test_c_tests): Add test_escaped_strings.
2591 * doc/extend.texi (deprecated): Add a note that the
2592 deprecation message is affected by the -fmessage-length
2593 option, and that control characters will be escaped.
2594 (#pragma GCC error): Document this pragma.
2595 (#pragma GCC warning): Likewise.
2596 * doc/invoke.texi (-fmessage-length): Document this option's
2597 effect on the #warning and #error preprocessor directives and
2598 the deprecated attribute.
2600 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2602 * tree.c (decl_value_expr_lookup): Revert latest change.
2603 (decl_value_expr_insert): Likewise.
2605 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2607 * gimplify.c (nonlocal_vlas): Delete.
2608 (nonlocal_vla_vars): Likewise.
2609 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2611 (gimplify_body): Do not create and destroy nonlocal_vlas.
2612 * tree-nested.c: Include diagnostic.h.
2613 (use_pointer_in_frame): Tweak.
2614 (lookup_field_for_decl): Add assertion and declare the transformation.
2615 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2616 internal error when the reference is in a wrong context. Do not
2617 create a debug decl by default.
2618 (note_nonlocal_block_vlas): Delete.
2619 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2620 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2621 create a debug decl by default.
2622 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2623 call is in a wrong context.
2624 (fixup_vla_decls): New function.
2625 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2626 debug variables were created.
2627 * tree.c (decl_value_expr_lookup): Add checking assertion.
2628 (decl_value_expr_insert): Likewise.
2630 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2633 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2634 * tree-scalar-evolution.c (interpret_expr): Likewise.
2635 (expression_expensive_p): Likewise.
2636 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2637 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2638 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2639 (ssa_defined_by_minus_one_stmt_p): New.
2641 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2644 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2645 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2646 * dojump.c (do_jump): Likewise.
2647 * expr.c (expand_expr_real_2): Check operand type's sign.
2648 * fold-const.c (const_unop): Handle ABSU_EXPR.
2649 (fold_abs_const): Likewise.
2650 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2651 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2652 (strip_sign_op_1): Likesise.
2653 * match.pd: Add new pattern to generate ABSU_EXPR.
2654 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2655 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2656 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2657 * tree-inline.c (estimate_operator_cost): Likewise.
2658 * tree-pretty-print.c (dump_generic_node): Likewise.
2659 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2660 * tree.def (ABSU_EXPR): New.
2662 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2665 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2666 documented as preserved for backward compatibility only.
2667 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2669 PR rtl-optimization/86108
2670 * bb-reorder.c (create_forwarder_block): Renamed to ...
2671 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2672 jump from new landing pad to the second part.
2673 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2676 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2679 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2680 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2681 Only call store_expr for halves if the mode is the same.
2684 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2685 Fix up comment formatting.
2687 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2689 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2690 (typed_splay_tree::closure,
2691 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2692 (typed_splay_tree::typed_splay_tree,
2693 typed_splay_tree::operator =): Declared private.
2694 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2695 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2696 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2697 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2698 typed_splay_tree::splay_tree_splay,
2699 typed_splay_tree::splay_tree_foreach_helper,
2700 typed_splay_tree::splay_tree_insert,
2701 typed_splay_tree::splay_tree_remove,
2702 typed_splay_tree::splay_tree_lookup,
2703 typed_splay_tree::splay_tree_predecessor,
2704 typed_splay_tree::splay_tree_successor,
2705 typed_splay_tree::splay_tree_min,
2706 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2707 (typed_splay_tree::root, typed_splay_tree::comp,
2708 typed_splay_tree::delete_key,
2709 typed_splay_tree::delete_value): New data members.
2710 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2711 typed_splay_tree::remove.
2713 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2715 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2716 -mginv and -mno-ginv to the assembler.
2717 * config/mips/mips.opt (-mcrc): New option.
2719 * doc/invoke.text (-mcrc): Document.
2722 2018-06-15 Nick Clifton <nickc@redhat.com>
2725 * tree.c (escaped_string): New class. Converts an unescaped
2726 string into its escaped equivalent.
2727 (warn_deprecated_use): Use the new class to convert the
2728 deprecation message, if present.
2729 (test_escaped_strings): New self test.
2730 (test_c_tests): Add test_escaped_strings.
2731 * doc/extend.texi (deprecated): Add a note that the
2732 deprecation message is affected by the -fmessage-length
2733 option, and that control characters will be escaped.
2734 (#pragma GCC error): Document this pragma.
2735 (#pragma GCC warning): Likewise.
2736 * doc/invoke.texi (-fmessage-length): Document this option's
2737 effect on the #warning and #error preprocessor directives and
2738 the deprecated attribute.
2740 2018-06-15 Richard Biener <rguenther@suse.de>
2742 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2743 here, also noting vector size used.
2744 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2745 size used in MSG_OPTIMIZED_LOCATIONS dump.
2746 (pass_slp_vectorize::execute): Adjust.
2748 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2751 * config/arc/arc.c (arc_return_address_register): Fix
2754 2018-06-15 Richard Biener <rguenther@suse.de>
2757 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2758 leave useless conversion stripping to force_gimple_operand_gsi.
2759 (gimplify_build2): Likewise.
2760 (gimplify_build1): Likewise.
2762 2018-06-15 Richard Biener <rguenther@suse.de>
2765 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2766 before adjusting their block.
2768 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2770 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2771 multilibs for *-*-rtems*.
2772 * config/riscv/t-rtems: New file.
2774 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2777 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2778 unsigned_type_for returns NULL.
2781 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2782 subregs of multi-word pseudos unless the float mode has word size.
2784 2018-06-14 Richard Biener <rguenther@suse.de>
2787 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2788 duplicate and harmful type_hash_canon.
2789 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2791 2018-06-14 Richard Biener <rguenther@suse.de>
2794 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2797 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2799 * config/rtems.h (STDINT_LONG32): Define.
2801 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2802 Prachi Godbole <prachi.godbole@imgtec.com>
2804 * config/mips/mips-cpus.def: Define P6600.
2805 * config/mips/mips-tables.opt: Regenerate.
2806 * config/mips/mips.c (mips_ucbranch_type): New enum.
2807 (mips_rtx_cost_data): Add support for P6600.
2808 (mips_issue_rate): Likewise.
2809 (mips_multipass_dfa_lookahead): Likewise.
2810 (mips_avoid_hazard): Likewise.
2811 (mips_reorg_process_insns): Likewise.
2812 (mips_classify_branch_p6600): New function.
2813 * config/mips/mips.h (TUNE_P6600): New define.
2814 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2815 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2816 * config/mips/mips.md: Include p6600.md.
2817 (processor): Add p6600.
2818 * config/mips/p6600.md: New file.
2819 * doc/invoke.texi: Add p6600 to supported architectures.
2821 2018-06-13 Martin Sebor <msebor@redhat.com>
2823 PR tree-optimization/86114
2824 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2826 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2828 2018-06-13 Richard Biener <rguenther@suse.de>
2830 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2831 Properly set vector type of the intermediate stmt.
2832 * tree-vect-stmts.c (vectorizable_operation): The destination
2833 var always has vectype_out type.
2835 2018-06-13 Jeff Law <law@redhat.com>
2837 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2838 integer 0 for argument to print_rtl_with_bb.
2839 (rl78_reorg): Likewise.
2841 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2843 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2844 from rtx to rtx_insn *.
2845 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2847 (add_sched_insns_for_speculation): Likewise for local "target",
2848 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2849 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2850 from rtx_insn ** to rtx_code_label **.
2851 (reorg_emit_nops): Likewise.
2852 (c6x_reorg): Likewise for local "call_labels".
2853 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2855 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2856 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2857 the loops over LABEL_REFS.
2858 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2860 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2861 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2862 (split_branches): Strengthen local "olabel" from rtx to
2863 rtx_insn *, adding a safe_as_a cast.
2864 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2866 (add_insn_after): Likewise for first two params.
2867 (add_insn_before): Likewise.
2868 (remove_insn): Likewise for param.
2869 (emit_pattern_before_noloc): Likewise for second and third params.
2870 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2871 (emit_call_insn_before_noloc): Likewise.
2872 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2874 (emit_barrier_before): Likewise.
2875 (emit_label_before): Strengthen "label" param from "rtx" to
2876 "rtx_code_label *". Strengthen "before" param from "rtx" to
2878 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2880 (emit_pattern_after_noloc): Likewise.
2881 (emit_insn_after_noloc): Likewise.
2882 (emit_jump_insn_after_noloc): Likewise.
2883 (emit_call_insn_after_noloc): Likewise.
2884 (emit_debug_insn_after_noloc): Likewise.
2885 (emit_barrier_after): Likewise.
2886 (emit_label_after): Likewise for both params.
2887 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2888 "loc" param from "int" to "location_t".
2889 (emit_insn_after_setloc): Likewise.
2890 (emit_jump_insn_after_setloc): Likewise.
2891 (emit_call_insn_after_setloc): Likewise.
2892 (emit_debug_insn_after_setloc): Likewise.
2893 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2894 "loc" param from "int" to "location_t".
2895 (emit_pattern_before): Convert NULL_RTX to NULL.
2896 (emit_insn_before_setloc): Convert "loc" param from "int" to
2898 (emit_jump_insn_before_setloc): Likewise.
2899 (emit_call_insn_before_setloc): Likewise.
2900 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2901 rtx_insn *. Convert "loc" param from "int" to "location_t".
2902 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2903 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2904 Convert 3rd param from "int" to "location_t".
2905 (emit_barrier_before, emit_barrier_after, next_real_insn):
2906 Strengthen param from rtx to rtx_insn *.
2907 (emit_label_before): Strengthen 1st param from "rtx" to
2908 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2910 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2911 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2912 Strengthen 2nd param from "rtx" to "rtx_insn *".
2913 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2914 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2915 Likewise. Convert 3rd param from "int" to "location_t".
2916 (emit_label_after): Strengthen 1st param from "rtx" to
2918 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2920 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2921 from "rtx" to "rtx_insn *".
2923 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2925 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2926 bodies streamed in with -Q.
2927 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2928 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2929 * gimple-streamer-out.c: Include gimple-pretty-print.h
2930 (output_bb): Dump stmts streamed.
2931 * lto-section-out.c: Include print-tree.h
2932 (lto_begin_section): Dump sections created.
2933 (lto_output_decl_index): Dump decl encoded.
2934 * lto-streamer-out.c: Include print-tree.h
2935 (create_output_block): Dump output block created.
2936 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2937 (output_function): Dump function output.
2938 (output_constructor): Dump constructor streamed.
2939 (write_global_stream): Output indexes encoded.
2940 (produce_asm_for_decls): Dump streams encoded.
2941 * lto-streamer.c (streamer_dump_file): New global var.
2942 * lto-streamer.h (streamer_dump_file): Declare.
2943 * passes.c (ipa_write_summaries): Initialize streamer dump.
2944 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2947 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2950 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2951 offsets for register save directives. Emit a second batch of save
2952 directives, if need be, when the function accesses prior frames.
2954 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2956 * config/arc/fpu.md (fmasf4): Force operand to register.
2957 (fnmasf4): Likewise.
2959 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2961 * config/arc/arc-protos.h (arc_pad_return): Remove.
2962 * config/arc/arc.c (machine_function): Remove force_short_suffix
2964 (arc_print_operand): Adjust printing of '&'.
2965 (arc_verify_short): Remove conditional printing of short suffix.
2966 (arc_final_prescan_insn): Remove reference to size_reason.
2967 (pad_return): New function.
2968 (arc_reorg): Call pad_return.
2969 (arc_pad_return): Remove.
2970 (arc_init_machine_status): Remove reference to force_short_suffix.
2971 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2972 (attr length): When attribute iscompact is true force to 2
2973 regardless; in the case of maybe check if we want to force the
2974 instruction to have 4 bytes length.
2975 (nopv): Change it to generate 4 byte long nop as well.
2976 (blockage): New pattern.
2977 (simple_return): Remove call to arc_pad_return.
2978 (p_return_i): Likewise.
2980 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2982 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2984 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2986 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2989 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2991 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2992 for ARC700 and ARCv2.
2994 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2997 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2998 operands[2] instead of operands[1].
3001 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3003 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3004 case, check whether the outer register overlaps an unallocatable
3005 register, not just whether it fits the required class.
3007 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3009 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3011 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3012 (memrefs_conflict_p): Likewise.
3013 (init_alias_analysis): Likewise.
3014 * cfgexpand.c (expand_debug_expr): Likewise.
3015 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3016 * cse.c (fold_rtx): Likewise.
3017 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3018 * expr.c (emit_block_move_hints): Likewise.
3019 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3020 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3021 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3022 (emit_group_store): Likewise.
3023 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3024 to read the PRE/POST_MODIFY increment.
3025 * calls.c (store_one_arg): Use strip_offset.
3026 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3028 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3030 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3031 (simplify_binary_operation_1): Extend CONST_INT handling to
3033 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3034 than a HOST_WIDE_INT.
3035 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3037 (adjust_mems, add_stores): Update accodingly.
3038 (vt_canonicalize_addr): Track polynomial offsets.
3039 (emit_note_insn_var_location): Likewise.
3040 (vt_add_function_parameter): Likewise.
3041 (vt_initialize): Likewise.
3043 2018-06-12 Jeff Law <law@redhat.com>
3045 * config.gcc (alpha*-*-freebsd*): Remove.
3046 * config/alpha/freebsd.h: Remove.
3048 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3051 * spellcheck-tree.c (levenshtein_distance): Rename to...
3052 (get_edit_distance): ...this, and update for underlying renaming.
3053 * spellcheck-tree.h (levenshtein_distance): Rename to...
3054 (get_edit_distance): ...this.
3055 * spellcheck.c (levenshtein_distance): Rename to...
3056 (get_edit_distance): ...this. Convert from Levenshtein distance
3057 to Damerau-Levenshtein distance by supporting transpositions of
3058 adjacent characters. Rename "v1" to "v_next" and "v0" to
3060 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3061 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3062 update for underlying renaming.
3063 (selftest::levenshtein_distance_unit_test): Rename to...
3064 (selftest::test_get_edit_distance_unit): ...this, and update for
3065 underlying renaming.
3066 (selftest::test_find_closest_string): Add example from PR 69968
3067 where transposition helps
3068 (selftest::test_metric_conditions): Update for renaming.
3069 (selftest::test_metric_conditions): Likewise.
3070 (selftest::spellcheck_c_tests): Likewise.
3071 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3072 (get_edit_distance): ...this.
3073 (best_match::consider): Update for renaming.
3075 2018-06-12 Martin Sebor <msebor@redhat.com>
3077 PR tree-optimization/85259
3078 * builtins.c (compute_objsize): Handle constant offsets.
3079 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3080 true iff a warning has been issued.
3081 * gimple.h (gimple_nonartificial_location): New function.
3082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3083 gimple_nonartificial_location and handle -Wno-system-headers.
3084 (handle_builtin_stxncpy): Same.
3086 2018-06-12 Martin Sebor <msebor@redhat.com>
3089 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3091 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3093 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3094 BUILTIN_VEC_XST entries for pointer to double and long long.
3096 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3099 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3101 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3104 2018-06-12 Martin Liska <mliska@suse.cz>
3106 * doc/options.texi: Document IntegerRange.
3108 2018-06-12 Martin Liska <mliska@suse.cz>
3110 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3111 * opt-functions.awk: Handle Deprecated flag.
3112 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3114 (read_cmdline_option): Report warning for a deprecated option.
3115 * opts.h (struct cl_option): Add new field cl_deprecated.
3116 (CL_ERR_DEPRECATED): New.
3118 2018-06-12 Martin Liska <mliska@suse.cz>
3120 * doc/options.texi: Document Deprecated option flag.
3122 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3124 * config/arc/arc-arch.h (arc_extras): New enum.
3125 (arc_cpu_t):Add field extra.
3126 (arc_cpu_types): Consider the extras.
3127 * config/arc/arc-cpus.def: Add extras info.
3128 * config/arc/arc-opts.h (processor_type): Consider extra field.
3129 * config/arc/arc.c (arc_override_options): Handle extra field.
3131 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3133 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3134 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3136 * config/arc/arc.c (arc_init): Update pic warning.
3137 (irq_range): Update irq range parsing warnings.
3138 (arc_override_options): Update various warning messages.
3139 (arc_handle_aux_attribute): Likewise.
3141 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3143 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3145 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3147 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3148 compiler messages directives.
3150 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3152 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3153 * config/mips/mips-tables.opt: Regenerate.
3154 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3156 * doc/invoke.texi: Document -march=i6500.
3158 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3160 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3161 (i6400_gpmul): Add cpu_unit.
3162 (i6400_gpdiv): Likewise.
3163 (i6400_msa_add_d): Update reservations.
3164 (i6400_msa_int_add) Likewise.
3165 (i6400_msa_short_logic3) Likewise.
3166 (i6400_msa_short_logic2) Likewise.
3167 (i6400_msa_short_logic) Likewise.
3168 (i6400_msa_move) Likewise.
3169 (i6400_msa_cmp) Likewise.
3170 (i6400_msa_short_float2) Likewise.
3171 (i6400_msa_div_d) Likewise.
3172 (i6400_msa_long_logic1) Likewise.
3173 (i6400_msa_long_logic2) Likewise.
3174 (i6400_msa_mult) Likewise.
3175 (i6400_msa_long_float2) Likewise.
3176 (i6400_msa_long_float4) Likewise.
3177 (i6400_msa_long_float5) Likewise.
3178 (i6400_msa_long_float8) Likewise.
3179 (i6400_fpu_fadd): Include frint type.
3180 (i6400_fpu_store): New define_insn_reservation.
3181 (i6400_fpu_load): Likewise.
3182 (i6400_fpu_move): Likewise.
3183 (i6400_fpu_fcmp): Likewise.
3184 (i6400_fpu_fmadd): Likewise.
3185 (i6400_int_mult): Include imul3nc type and update reservation.
3186 (i6400_int_div): Include idiv3 type and update reservation.
3187 (i6400_int_load): Update to check type not move_type.
3188 (i6400_int_store): Likewise.
3189 (i6400_int_prefetch): Set zero latency.
3191 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3193 * gcc.c: Document new %@{...} sequence.
3194 (LINK_COMMAND_SPEC): Use it for the -L switches.
3195 (cpp_unique_options): Use it for the -I switches.
3196 (at_file_argbuf): New global variable.
3197 (in_at_file): Likewise.
3198 (alloc_args): Create at_file_argbuf.
3199 (clear_args): Truncate at_file_argbuf.
3200 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3201 (open_at_file): New function.
3202 (close_at_file): Likewise.
3203 (create_at_file): Delete.
3204 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3207 (validate_switches_from_spec): Deal with %@{...} sequence.
3208 (validate_switches): Likewise.
3209 (driver::finalize): Call clear_args.
3211 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3213 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3215 2018-06-11 Martin Sebor <msebor@redhat.com>
3217 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3218 (Wno-class-memaccess): Add @opindex.
3219 (Wno-templates, Wno-multiple-inheritance): Same.
3220 (Wno-virtual-inheritance, Wno-namespaces): Same.
3221 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3222 (Wno-format-overflow, Wno-format-truncation): Same.
3223 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3224 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3225 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3226 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3227 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3228 (Wno-misspelled-isr): Same.
3230 2018-06-11 Martin Sebor <msebor@redhat.com>
3232 * PR tree-optimization/86083
3233 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3235 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3237 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3239 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3242 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3243 on the correct operand.
3244 (*movdi_internal64): Ditto.
3246 2018-06-11 Martin Liska <mliska@suse.cz>
3248 PR tree-optimization/86089
3249 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3251 2018-06-11 Julia Koval <julia.koval@intel.com>
3253 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3254 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3255 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3257 2018-06-11 Olivier Hainque <hainque@adacore.com>
3259 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3260 for Ada with strict dwarf2.
3262 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3265 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3268 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3270 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3272 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3274 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3277 2018-06-08 Martin Liska <mliska@suse.cz>
3279 * tree-cfg.h (debug_function): Fix argument type to match
3282 2018-06-08 Martin Liska <mliska@suse.cz>
3284 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3285 Remove usage of MPX-related (and removed) fields.
3286 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3288 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3290 * cfg.c (debug): Use TDF_NONE rather than 0.
3291 * cfghooks.c (debug): Likewise.
3292 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3293 (struct dump_option_value_info): Convert to...
3294 (struct kv_pair): ...this template type.
3295 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3297 (optinfo_verbosity_options): Likewise.
3298 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3300 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3301 than int for "optgroup_flags" param.
3302 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3304 (dump_dec): Likewise.
3305 (dump_finish): Use TDF_NONE rather than 0.
3306 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3307 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3308 than 0. Update for change to option_ptr.
3309 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3310 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3311 0. Update for changes to optinfo_verbosity_options and
3313 (opt_info_switch_p): Convert optgroup_flags from int to
3315 (dump_basic_block): Use dump_flags_t rather than int
3316 for "dump_kind" param.
3317 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3318 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3319 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3320 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3321 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3322 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3323 TDF_NONE): Convert from macros to...
3324 (enum dump_flag): ...this new enum.
3325 (dump_flags_t): Update to use enum.
3326 (operator|, operator&, operator~, operator|=, operator&=):
3327 Implement for dump_flags_t.
3328 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3329 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3330 Convert from macros to...
3331 (enum optgroup_flag): ...this new enum.
3332 (optgroup_flags_t): New typedef.
3333 (operator|, operator|=): Implement for optgroup_flags_t.
3334 (struct dump_file_info): Convert field "alt_flags" to
3335 dump_flags_t. Convert field "optgroup_flags" to
3337 (dump_basic_block): Use dump_flags_t rather than int for param.
3338 (dump_generic_expr_loc): Likewise.
3339 (dump_dec): Likewise.
3340 (dump_register): Convert param "optgroup_flags" to
3342 (opt_info_enable_passes): Likewise.
3343 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3345 * gimple-pretty-print.c (debug): Likewise.
3346 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3347 (merged_store_group::apply_stores): Likewise.
3348 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3349 * gimple.c (verify_gimple_pp): Likewise.
3350 * graphite-poly.c (print_pbb_body): Likewise.
3351 * passes.c (pass_manager::register_one_dump_file): Convert
3352 local "optgroup_flags" to optgroup_flags_t.
3353 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3355 (debug_body): Likewise.
3356 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3357 to optgroup_flags_t.
3358 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3360 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3361 (convert_mult_to_fma): Likewise.
3362 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3363 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3364 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3365 "dump_kind" to dump_flags_t.
3367 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3369 * config/rs6000/rs6000.c (min, max): Delete.
3371 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3373 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3374 -mabi=spe and -mabi=no-spe.
3376 2018-06-08 Martin Liska <mliska@suse.cz>
3378 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3379 where we expect an existing summary.
3381 2018-06-08 Martin Liska <mliska@suse.cz>
3383 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3384 * ipa-inline.h (estimate_edge_growth): Likewise.
3386 2018-06-08 Martin Liska <mliska@suse.cz>
3388 * cgraph.c (function_version_hasher::hash): Use
3389 cgraph_node::get_uid ().
3390 (function_version_hasher::equal):
3391 * cgraph.h (cgraph_node::get_uid): New method.
3392 * ipa-inline.c (update_caller_keys): Use
3393 cgraph_node::get_uid ().
3394 (update_callee_keys): Likewise.
3395 * ipa-utils.c (searchc): Likewise.
3396 (ipa_reduced_postorder): Likewise.
3397 * lto-cgraph.c (input_node): Likewise.
3398 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3399 * symbol-summary.h (symtab_insertion): Likewise.
3400 (symtab_removal): Likewise.
3401 (symtab_duplication): Likewise.
3402 * tree-pretty-print.c (dump_function_header): Likewise.
3403 * tree-sra.c (convert_callers_for_node): Likewise.
3405 2018-06-08 Martin Liska <mliska@suse.cz>
3407 * cgraph.c (symbol_table::create_edge): Always assign a new
3409 (symbol_table::free_edge): Do not recycle numbers.
3410 * cgraph.h (cgraph_edge::get): New method.
3411 * symbol-summary.h (symtab_removal): Use it.
3412 (symtab_duplication): Likewise.
3413 (call_summary::hashable_uid): Remove.
3415 2018-06-08 Martin Liska <mliska@suse.cz>
3417 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3418 (initialize_growth_caches): Remove.
3419 (free_growth_caches): Likewise.
3420 (do_estimate_edge_time): Use edge_growth_cache.
3421 (do_estimate_edge_size): Likewise.
3422 (do_estimate_edge_hints): Likewise.
3423 * ipa-inline.c (reset_edge_caches): Likewise.
3424 (recursive_inlining): Likewise.
3425 (inline_small_functions): Likewise.
3426 * ipa-inline.h (initialize_growth_caches): Remove.
3427 (estimate_edge_size): Likewise.
3428 (estimate_edge_time): Likewise.
3429 (estimate_edge_hints): Likewise.
3430 (reset_edge_growth_cache): Likewise.
3431 * symbol-summary.h (call_summary::remove): New method.
3433 2018-06-08 Martin Liska <mliska@suse.cz>
3435 * ipa-cp.c (class edge_clone_summary): New summary.
3436 (grow_edge_clone_vectors): Remove.
3437 (ipcp_edge_duplication_hook): Remove.
3438 (class edge_clone_summary_t): New call_summary class.
3439 (ipcp_edge_removal_hook): Remove.
3440 (edge_clone_summary_t::duplicate): New function.
3441 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3442 (create_specialized_node): Likewise.
3443 (ipcp_driver): Initialize edge_clone_summaries and do not
3446 2018-06-08 Martin Liska <mliska@suse.cz>
3448 * symbol-summary.h (get): New function.
3449 (call_summary::m_initialize_when_cloning): New class member.
3451 2018-06-08 Martin Liska <mliska@suse.cz>
3453 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3454 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3455 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3456 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3457 (remove_cgraph_node_from_order): Use the removed_nodes set.
3458 (do_per_function_toporder): Likwise.
3459 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3460 instead of summary_uid.
3461 (symtab_removal): Likewise.
3462 (symtab_duplication): Likewise.
3464 2018-06-08 Martin Liska <mliska@suse.cz>
3466 * ipa-cp.c (ipcp_store_bits_results): Use
3467 ipcp_transformation_sum.
3468 (ipcp_store_vr_results): Likewise.
3469 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3471 (ipcp_transformation_initialize): ... this.
3472 (ipa_set_node_agg_value_chain):
3473 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3474 (write_ipcp_transformation_info): Likewise.
3475 (read_ipcp_transformation_info): Likewise.
3476 (ipcp_update_bits): Likewise.
3477 (ipcp_update_vr): Likewise.
3478 (ipcp_transform_function): Likewise.
3479 * ipa-prop.h: Rename ipcp_transformation_summary to
3480 ipcp_transformation.
3481 (class ipcp_transformation_t): New function summary.
3482 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3483 (ipa_get_agg_replacements_for_node): Likewise.
3485 2018-06-08 Martin Liska <mliska@suse.cz>
3487 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3489 (class funct_state_summary_t): New function_summary class.
3490 (has_function_state): Remove.
3491 (get_function_state): Likewise.
3492 (set_function_state): Likewise.
3493 (add_new_function): Likewise.
3494 (funct_state_summary_t::insert): New function.
3495 (duplicate_node_data): Remove.
3496 (remove_node_data): Remove.
3497 (funct_state_summary_t::duplicate): New function.
3498 (register_hooks): Create new funct_state_summaries.
3499 (pure_const_generate_summary): Use it.
3500 (pure_const_write_summary): Likewise.
3501 (pure_const_read_summary): Likewise.
3502 (propagate_pure_const): Likewise.
3503 (propagate_nothrow): Likewise.
3504 (dump_malloc_lattice): Likewise.
3505 (propagate_malloc): Likewise.
3506 (execute): Do not register hooks, just remove summary
3508 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3511 2018-06-08 Martin Liska <mliska@suse.cz>
3513 * ipa-reference.c (remove_node_data): Remove.
3514 (duplicate_node_data): Likewise.
3515 (class ipa_ref_var_info_summary_t): New class.
3516 (class ipa_ref_opt_summary_t): Likewise.
3517 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3518 (get_reference_optimization_summary): Use
3519 ipa_ref_opt_sum_summaries.
3520 (set_reference_vars_info): Remove.
3521 (set_reference_optimization_summary): Likewise.
3522 (ipa_init): Create summaries.
3523 (init_function_info): Use function summary.
3524 (ipa_ref_opt_summary_t::duplicate): New function.
3525 (ipa_ref_opt_summary_t::remove): New function.
3526 (get_read_write_all_from_node): Fix GNU coding style.
3527 (propagate): Use function summary.
3528 (write_node_summary_p): Fix GNU coding style.
3529 (stream_out_bitmap): Likewise.
3530 (ipa_reference_read_optimization_summary): Use function summary.
3531 (ipa_reference_c_finalize): Do not release hooks.
3533 2018-06-08 Martin Liska <mliska@suse.cz>
3535 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3536 (analyze_function_body): Extract multiple calls of get_create.
3537 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3538 * ipa-inline.c (recursive_inlining): Use ::get method.
3539 * ipa-inline.h (estimate_edge_growth): Likewise.
3541 2018-06-08 Martin Liska <mliska@suse.cz>
3543 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3545 (hsa_function_summary::hsa_function_summary): Use the new enum
3547 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3548 * hsa-gen.c (hsa_get_host_function): Likewise.
3549 (get_brig_function_name): Likewise.
3550 * ipa-hsa.c (process_hsa_functions): Likewise.
3551 (ipa_hsa_write_summary): Likewise.
3552 * symbol-summary.h (symtab_duplication): Use ::get function/
3553 (get): New function.
3555 2018-06-08 Martin Liska <mliska@suse.cz>
3557 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3559 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3560 (hsa_register_kernel): Likewise.
3561 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3562 * hsa-gen.c (hsa_get_host_function): Likewise.
3563 (get_brig_function_name): Likewise.
3564 (generate_hsa): Likewise.
3565 (pass_gen_hsail::execute): Likewise.
3566 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3567 (devirtualization_time_bonus): Likewise.
3568 (ipcp_propagate_stage): Likewise.
3569 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3570 (edge_set_predicate): Likewise.
3571 (evaluate_conditions_for_known_args): Likewise.
3572 (evaluate_properties_for_edge): Likewise.
3573 (ipa_fn_summary::reset): Likewise.
3574 (ipa_fn_summary_t::duplicate): Likewise.
3575 (dump_ipa_call_summary): Likewise.
3576 (ipa_dump_fn_summary): Likewise.
3577 (analyze_function_body): Likewise.
3578 (compute_fn_summary): Likewise.
3579 (estimate_edge_devirt_benefit): Likewise.
3580 (estimate_edge_size_and_time): Likewise.
3581 (estimate_calls_size_and_time): Likewise.
3582 (estimate_node_size_and_time): Likewise.
3583 (inline_update_callee_summaries): Likewise.
3584 (remap_edge_change_prob): Likewise.
3585 (remap_edge_summaries): Likewise.
3586 (ipa_merge_fn_summary_after_inlining): Likewise.
3587 (ipa_update_overall_fn_summary): Likewise.
3588 (read_ipa_call_summary): Likewise.
3589 (inline_read_section): Likewise.
3590 (write_ipa_call_summary): Likewise.
3591 (ipa_fn_summary_write): Likewise.
3592 (ipa_free_fn_summary): Likewise.
3593 * ipa-hsa.c (process_hsa_functions): Likewise.
3594 (ipa_hsa_write_summary): Likewise.
3595 (ipa_hsa_read_section): Likewise.
3596 * ipa-icf.c (sem_function::merge): Likewise.
3597 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3598 (do_estimate_edge_time): Likewise.
3599 (estimate_size_after_inlining): Likewise.
3600 (estimate_growth): Likewise.
3601 (growth_likely_positive): Likewise.
3602 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3603 (inline_call): Likewise.
3604 * ipa-inline.c (caller_growth_limits): Likewise.
3605 (can_inline_edge_p): Likewise.
3606 (can_inline_edge_by_limits_p): Likewise.
3607 (compute_uninlined_call_time): Likewise.
3608 (compute_inlined_call_time): Likewise.
3609 (want_inline_small_function_p): Likewise.
3610 (edge_badness): Likewise.
3611 (update_caller_keys): Likewise.
3612 (update_callee_keys): Likewise.
3613 (recursive_inlining): Likewise.
3614 (inline_small_functions): Likewise.
3615 (inline_to_all_callers_1): Likewise.
3616 (dump_overall_stats): Likewise.
3617 (early_inline_small_functions): Likewise.
3618 (early_inliner): Likewise.
3619 * ipa-inline.h (estimate_edge_growth): Likewise.
3620 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3621 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3622 * ipa-prop.h (IPA_NODE_REF): Likewise.
3623 (IPA_EDGE_REF): Likewise.
3624 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3625 (propagate_malloc): Likewise.
3626 * ipa-split.c (execute_split_functions): Likewise.
3627 * symbol-summary.h: Rename get to get_create.
3629 (get_create): Likewise.
3630 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3632 2018-06-08 Martin Liska <mliska@suse.cz>
3634 * symbol-summary.h (release): Move definition out of class
3636 (symtab_removal): Likewise.
3637 (symtab_duplication): Likewise.
3639 2018-06-08 Martin Liska <mliska@suse.cz>
3641 * symbol-summary.h (function_summary): Move constructor
3642 implementation out of class declaration.
3643 (release): Likewise.
3644 (symtab_insertion): Likewise.
3645 (symtab_removal): Likewise.
3646 (symtab_duplication): Likewise.
3649 2018-06-08 Martin Liska <mliska@suse.cz>
3651 * Makefile.in: Remove support for MPX (macros, related functions,
3652 fields in cgraph_node, ...).
3653 * builtin-types.def (BT_BND): Likewise.
3654 (BT_FN_BND_CONST_PTR): Likewise.
3655 (BT_FN_CONST_PTR_BND): Likewise.
3656 (BT_FN_VOID_PTR_BND): Likewise.
3657 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3658 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3659 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3660 (expand_builtin_mempcpy_with_bounds): Likewise.
3661 (expand_builtin_memset_with_bounds): Likewise.
3662 (expand_builtin_memset_args): Likewise.
3663 (std_expand_builtin_va_start): Likewise.
3664 (expand_builtin): Likewise.
3665 (expand_builtin_with_bounds): Likewise.
3666 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3667 (DEF_LIB_BUILTIN_CHKP): Likewise.
3668 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3669 (DEF_CHKP_BUILTIN): Likewise.
3670 (BUILT_IN_MEMCPY): Likewise.
3671 (BUILT_IN_MEMMOVE): Likewise.
3672 (BUILT_IN_MEMPCPY): Likewise.
3673 (BUILT_IN_MEMSET): Likewise.
3674 (BUILT_IN_STPCPY): Likewise.
3675 (BUILT_IN_STRCAT): Likewise.
3676 (BUILT_IN_STRCHR): Likewise.
3677 (BUILT_IN_STRCPY): Likewise.
3678 (BUILT_IN_STRLEN): Likewise.
3679 (BUILT_IN_MEMCPY_CHK): Likewise.
3680 (BUILT_IN_MEMMOVE_CHK): Likewise.
3681 (BUILT_IN_MEMPCPY_CHK): Likewise.
3682 (BUILT_IN_MEMSET_CHK): Likewise.
3683 (BUILT_IN_STPCPY_CHK): Likewise.
3684 (BUILT_IN_STRCAT_CHK): Likewise.
3685 (BUILT_IN_STRCPY_CHK): Likewise.
3686 * calls.c (store_bounds): Likewise.
3687 (emit_call_1): Likewise.
3688 (special_function_p): Likewise.
3689 (maybe_warn_nonstring_arg): Likewise.
3690 (initialize_argument_information): Likewise.
3691 (finalize_must_preallocate): Likewise.
3692 (compute_argument_addresses): Likewise.
3693 (expand_call): Likewise.
3694 * cfgexpand.c (expand_call_stmt): Likewise.
3695 (expand_return): Likewise.
3696 (expand_gimple_stmt_1): Likewise.
3697 (pass_expand::execute): Likewise.
3698 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3699 (cgraph_node::remove): Likewise.
3700 (cgraph_node::dump): Likewise.
3701 (cgraph_node::verify_node): Likewise.
3702 * cgraph.h (chkp_function_instrumented_p): Likewise.
3703 (symtab_node::get_alias_target): Likewise.
3704 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3705 (cgraph_local_p): Likewise.
3706 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3707 (cgraph_edge::rebuild_references): Likewise.
3708 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3709 (walk_polymorphic_call_targets): Likewise.
3710 (cgraph_node::expand_thunk): Likewise.
3711 (symbol_table::output_weakrefs): Likewise.
3712 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3713 (ix86_handle_option): Likewise.
3714 * config/i386/constraints.md: Likewise.
3715 * config/i386/i386-builtin-types.def (BND): Likewise.
3719 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3720 (BDESC_FIRST): Likewise.
3722 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3723 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3724 * config/i386/i386.c (enum reg_class): Likewise.
3725 (ix86_target_string): Likewise.
3726 (ix86_option_override_internal): Likewise.
3727 (ix86_conditional_register_usage): Likewise.
3728 (ix86_valid_target_attribute_inner_p): Likewise.
3729 (ix86_set_indirect_branch_type): Likewise.
3730 (ix86_set_current_function): Likewise.
3731 (ix86_function_arg_regno_p): Likewise.
3732 (init_cumulative_args): Likewise.
3733 (ix86_function_arg_advance): Likewise.
3734 (ix86_function_arg): Likewise.
3735 (ix86_pass_by_reference): Likewise.
3736 (ix86_function_value_regno_p): Likewise.
3737 (ix86_function_value_1): Likewise.
3738 (ix86_function_value_bounds): Likewise.
3739 (ix86_return_in_memory): Likewise.
3740 (ix86_setup_incoming_vararg_bounds): Likewise.
3741 (ix86_va_start): Likewise.
3742 (indirect_thunk_need_prefix): Likewise.
3743 (print_reg): Likewise.
3744 (ix86_print_operand): Likewise.
3745 (ix86_expand_call): Likewise.
3746 (ix86_output_function_return): Likewise.
3747 (reg_encoded_number): Likewise.
3748 (BDESC_VERIFYS): Likewise.
3749 (ix86_init_mpx_builtins): Likewise.
3750 (ix86_init_builtins): Likewise.
3751 (ix86_emit_cmove): Likewise.
3752 (ix86_emit_move_max): Likewise.
3753 (ix86_expand_builtin): Likewise.
3754 (ix86_builtin_mpx_function): Likewise.
3755 (ix86_get_arg_address_for_bt): Likewise.
3756 (ix86_load_bounds): Likewise.
3757 (ix86_store_bounds): Likewise.
3758 (ix86_load_returned_bounds): Likewise.
3759 (ix86_store_returned_bounds): Likewise.
3760 (ix86_class_likely_spilled_p): Likewise.
3761 (ix86_hard_regno_mode_ok): Likewise.
3762 (x86_order_regs_for_local_alloc): Likewise.
3763 (ix86_mitigate_rop): Likewise.
3764 (ix86_bnd_prefixed_insn_p): Likewise.
3765 (ix86_mpx_bound_mode): Likewise.
3766 (ix86_make_bounds_constant): Likewise.
3767 (ix86_initialize_bounds): Likewise.
3768 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3769 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3770 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3771 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3772 (TARGET_CHKP_BOUND_MODE): Likewise.
3773 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3774 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3775 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3776 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3777 * config/i386/i386.h (TARGET_MPX): Likewise.
3778 (TARGET_MPX_P): Likewise.
3779 (VALID_BND_REG_MODE): Likewise.
3780 (FIRST_BND_REG): Likewise.
3781 (LAST_BND_REG): Likewise.
3782 (enum reg_class): Likewise.
3783 (BND_REG_P): Likewise.
3784 (BND_REGNO_P): Likewise.
3785 (BNDmode): Likewise.
3786 (ADJUST_INSN_LENGTH): Likewise.
3787 * config/i386/i386.md: Likewise.
3788 * config/i386/i386.opt: Likewise.
3789 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3790 (defined): Likewise.
3791 (LINK_MPX): Likewise.
3792 (MPX_SPEC): Likewise.
3793 (LIBMPX_SPEC): Likewise.
3794 (LIBMPXWRAPPERS_SPEC): Likewise.
3795 (CHKP_SPEC): Likewise.
3796 * config/i386/predicates.md: Likewise.
3797 * dbxout.c (dbxout_type): Likewise.
3798 * doc/extend.texi: Likewise.
3799 * doc/invoke.texi: Likewise.
3800 * doc/md.texi: Likewise.
3801 * doc/tm.texi: Likewise.
3802 * doc/tm.texi.in: Likewise.
3803 * dwarf2out.c (is_base_type): Likewise.
3804 (gen_formal_types_die): Likewise.
3805 (gen_subprogram_die): Likewise.
3806 (gen_type_die_with_usage): Likewise.
3807 (gen_decl_die): Likewise.
3808 (dwarf2out_late_global_decl): Likewise.
3809 * expr.c (expand_assignment): Likewise.
3810 (emit_storent_insn): Likewise.
3811 (store_expr_with_bounds): Likewise.
3812 (store_expr): Likewise.
3813 (expand_expr_real_1): Likewise.
3814 * expr.h (store_expr_with_bounds): Likewise.
3815 * function.c (use_register_for_decl): Likewise.
3816 (struct bounds_parm_data): Likewise.
3817 (assign_parms_augmented_arg_list): Likewise.
3818 (assign_parm_find_entry_rtl): Likewise.
3819 (assign_parm_is_stack_parm): Likewise.
3820 (assign_parm_load_bounds): Likewise.
3821 (assign_bounds): Likewise.
3822 (assign_parms): Likewise.
3823 (expand_function_start): Likewise.
3824 * gcc.c (CHKP_SPEC): Likewise.
3825 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3826 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3827 (wrestrict_dom_walker::check_call): Likewise.
3828 * gimple.c (gimple_build_call_from_tree): Likewise.
3829 * gimple.h (enum gf_mask): Likewise.
3830 (gimple_call_with_bounds_p): Likewise.
3831 (gimple_call_set_with_bounds): Likewise.
3832 * gimplify.c (gimplify_init_constructor): Likewise.
3833 * ipa-cp.c (initialize_node_lattices): Likewise.
3834 (propagate_constants_across_call): Likewise.
3835 (find_more_scalar_values_for_callers_subset): Likewise.
3836 * ipa-hsa.c (process_hsa_functions): Likewise.
3837 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3838 * ipa-icf.c (sem_function::merge): Likewise.
3839 * ipa-inline.c (early_inliner): Likewise.
3840 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3841 (warn_function_cold): Likewise.
3842 (propagate_pure_const): Likewise.
3843 * ipa-ref.h (enum GTY): Likewise.
3844 * ipa-split.c (find_retbnd): Likewise.
3845 (consider_split): Likewise.
3846 (split_function): Likewise.
3847 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3848 * ipa.c (walk_polymorphic_call_targets): Likewise.
3849 (symbol_table::remove_unreachable_nodes): Likewise.
3850 (process_references): Likewise.
3851 (cgraph_build_static_cdtor_1): Likewise.
3852 * lto-cgraph.c (lto_output_node): Likewise.
3853 (output_refs): Likewise.
3854 (compute_ltrans_boundary): Likewise.
3855 (input_overwrite_node): Likewise.
3856 (input_node): Likewise.
3857 (input_cgraph_1): Likewise.
3858 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3859 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3860 (class pass_chkp_instrumentation_passes): Likewise.
3861 (make_pass_chkp_instrumentation_passes): Likewise.
3862 * passes.def: Likewise.
3863 * rtl.h (struct GTY): Likewise.
3864 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3865 * stor-layout.c (layout_type): Likewise.
3866 * symtab.c: Likewise.
3867 * target.def: Likewise.
3868 * targhooks.c (default_chkp_bound_type): Likewise.
3869 (default_chkp_bound_mode): Likewise.
3870 (default_builtin_chkp_function): Likewise.
3871 (default_chkp_function_value_bounds): Likewise.
3872 (default_chkp_make_bounds_constant): Likewise.
3873 (default_chkp_initialize_bounds): Likewise.
3874 * targhooks.h (default_chkp_bound_type): Likewise.
3875 (default_chkp_bound_mode): Likewise.
3876 (default_builtin_chkp_function): Likewise.
3877 (default_chkp_function_value_bounds): Likewise.
3878 (default_chkp_make_bounds_constant): Likewise.
3879 (default_chkp_initialize_bounds): Likewise.
3880 * toplev.c (compile_file): Likewise.
3881 (process_options): Likewise.
3882 * tree-core.h (DEF_BUILTIN): Likewise.
3883 (DEF_BUILTIN_CHKP): Likewise.
3884 * tree-inline.c (declare_return_variable): Likewise.
3885 (remap_gimple_stmt): Likewise.
3886 (copy_bb): Likewise.
3887 (initialize_inlined_parameters): Likewise.
3888 (expand_call_inline): Likewise.
3889 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3890 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3891 (make_pass_ipa_chkp_produce_thunks): Likewise.
3892 (make_pass_chkp): Likewise.
3893 (make_pass_chkp_opt): Likewise.
3894 (make_pass_chkp_instrumentation_passes): Likewise.
3895 * tree-pretty-print.c (dump_generic_node): Likewise.
3896 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3897 * tree-ssa-dce.c (propagate_necessity): Likewise.
3898 (eliminate_unnecessary_stmts): Likewise.
3899 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3900 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3901 * tree-ssa-sccvn.h: Likewise.
3902 * tree-ssa-strlen.c (get_string_length): Likewise.
3903 (valid_builtin_call): Likewise.
3904 (adjust_last_stmt): Likewise.
3905 (handle_builtin_strchr): Likewise.
3906 (handle_builtin_strcpy): Likewise.
3907 (handle_builtin_stxncpy): Likewise.
3908 (handle_builtin_memcpy): Likewise.
3909 (handle_builtin_strcat): Likewise.
3910 (strlen_check_and_optimize_stmt): Likewise.
3911 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3912 * tree-streamer-in.c: Likewise.
3913 * tree-streamer.c (record_common_node): Likewise.
3914 * tree.c (tree_code_size): Likewise.
3915 (wide_int_to_tree_1): Likewise.
3916 (type_contains_placeholder_1): Likewise.
3917 (build_common_tree_nodes): Likewise.
3918 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3919 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3920 (POINTER_BOUNDS_P): Likewise.
3921 (BOUNDED_TYPE_P): Likewise.
3922 (BOUNDED_P): Likewise.
3923 (CALL_WITH_BOUNDS_P): Likewise.
3924 (pointer_bounds_type_node): Likewise.
3925 * value-prof.c (gimple_ic): Likewise.
3926 * var-tracking.c (vt_add_function_parameters): Likewise.
3927 * varasm.c (make_decl_rtl): Likewise.
3928 (assemble_start_function): Likewise.
3929 (output_constant): Likewise.
3930 (maybe_assemble_visibility): Likewise.
3931 * varpool.c (ctor_for_folding): Likewise.
3932 * chkp-builtins.def: Remove.
3933 * ipa-chkp.c: Remove.
3934 * ipa-chkp.h: Remove.
3935 * rtl-chkp.c: Remove.
3936 * rtl-chkp.h: Remove.
3937 * tree-chkp-opt.c: Remove.
3938 * tree-chkp.c: Remove.
3939 * tree-chkp.h: Remove.
3941 2018-06-07 Carl Love <cel@us.ibm.com>
3943 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3944 vextract_fp_from_shortl): Add BE support.
3946 2018-06-07 Paul Koning <ni1d@arrl.net>
3948 * compare-elim.c (try_merge_compare): Don't merge compare if
3949 address contains a side effect.
3950 (try_eliminate_compare): Likewise.
3952 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3954 * config.gcc: Support "tremont".
3955 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3956 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3958 * config/i386/i386.c (m_TREMONT): Define.
3959 (processor_target_table): Add "tremont".
3960 (PTA_TREMONT): Define.
3961 (ix86_lea_outperforms): Add TARGET_TREMONT.
3962 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3963 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3964 and M_INTEL_GOLDMONT_PLUS.
3965 (fold_builtin_cpu): Add "tremont".
3966 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3967 (ix86_option_override_internal): Add "tremont".
3968 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3969 (processor_type): Add PROCESSOR_TREMONT.
3970 * config/i386/x86-tune.def: Add m_TREMONT.
3971 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3973 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3975 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3976 symbol defined for msp430i* devices to be lower case.
3978 gcc/testsuite/gcc.target/msp430/
3979 * msp430i-device-symbol.c: New test.
3980 * msp430f-device-symbol.c: New test.
3981 * msp430.h: New test header file.
3983 2018-06-07 Richard Biener <rguenther@suse.de>
3985 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3986 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3987 Properly wrap signed arithmetic if overflow wraps.
3989 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3991 PR tree-optimization/69615
3992 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3993 of a cast from a same precision integral SSA_NAME in a bb dominated
3994 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3995 cast to utype if rhs2 has already a compatible type.
3997 2018-06-07 Richard Biener <rguenther@suse.de>
3999 PR tree-optimization/85935
4000 * graphite-scop-detection.c (find_params_in_bb): Analyze
4001 condition operands with respect to the correct loop. Assert
4002 the analysis doesn't fail.
4004 2018-06-04 Carl Love <cel@us.ibm.com>
4006 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4007 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4008 as it is slightly cheaper.
4009 (first_match_or_eos_index_<mode>):
4010 Calculate index using natural element order.
4011 (first_match_index_<mode>):
4012 Calculate index using natural element order.
4013 (first_match_or_eos_index_<mode>):
4014 Calculate index using natural order.
4015 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4016 for BE and LE modes.
4017 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4018 P9V_BUILTIN_VCLZLSBB_V16QI.
4019 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4022 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4024 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4025 indentation and line wrap for many prototypes. Add missing
4026 @smallexample directives around block of prototypes for vec_xl and
4029 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4031 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4032 track if we pass or return IEEE 128-bit floating point.
4033 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4034 C++ mangling that is compatible with GCC 8.1.
4035 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4036 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4037 floating point types.
4038 (rs6000_function_arg_advance_1): Likewise.
4039 (rs6000_mangle_type): Optionally generate mangled names that match
4040 what GCC 8.1 generated for IEEE 128-bit floating point types.
4041 (rs6000_globalize_decl_name): If we have an external function that
4042 passes or returns IEEE 128-bit types, generate a weak reference
4043 from the mangled name used in GCC 8.1 to the current mangled
4045 (rs6000_init_builtins): Make __ibm128 use the long double type if
4046 long double is IBM extended double. Make __float128 use the long
4047 double type if long double is IEEE 128-bit.
4050 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4051 macro for __ibm128 built-in functions.
4052 (PACK_IF): Add __ibm128 pack/unpack functions.
4053 (UNPACK_IF): Likewise.
4054 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4055 enable long double built-in functions if long double is IEEE
4056 128-bit floating point.
4057 (rs6000_invalid_builtin): Update long double built-in function
4059 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4060 functions, adjust the built-in function to use the long double
4061 built-in function if __ibm128 and long double are the same type.
4062 * doc/extend.texi (PowerPC builtins): Update documention for
4063 __builtin_{,un}pack_longdouble. Add documentation for
4064 __builtin_{,un}pack_ibm128.
4066 2018-06-06 Jim Wilson <jimw@sifive.com>
4068 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4069 (struct machine_function): New field interrupt_mode.
4070 (riscv_handle_type_attribute): New function. Add forward declaration.
4071 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4072 (riscv_expand_epilogue): Check interrupt_mode field.
4073 (riscv_set_current_function): Check interrupt attribute args and
4074 set interrupt_mode field.
4075 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4076 (riscv_sret, riscv_uret): New.
4077 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4078 new arguments to interrupt attribute.
4080 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4083 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4084 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4086 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4088 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4089 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4090 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4091 several redundant entries.
4093 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4095 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4096 type from "rtx" to "rtx_insn *".
4097 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4098 for local "call_insn", removing cast.
4099 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4101 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4103 PR tree-optimization/86066
4104 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4105 for BIT_INSERT_EXPR stores.
4107 2018-06-06 Richard Biener <rguenther@suse.de>
4109 PR tree-optimization/86062
4110 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4111 component refs ontop
4112 of to be offsetted base.
4114 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4116 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4117 to be static and remove check on interrupt attribute name.
4119 gcc/testsuite/gcc.target/msp430/
4120 * function-attributes-4.c: New test.
4121 * static-interrupts.c: New test.
4123 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4125 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4126 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4128 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4131 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4133 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4134 Remove second argument, change how error is called.
4135 (aarch64_layout_arg): Remove second argument from
4136 aarch64_err_no_fpadvsimd call.
4137 (aarch64_init_cumulative_args): Ditto.
4138 (aarch64_gimplify_va_arg_expr): Ditto.
4139 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4141 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4143 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4144 (*simple_return_indirect_internal<mode>): Rename from
4145 simple_return_indirect_internal. Use W mode iterator.
4146 (rstorssp): New expander.
4147 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4148 (clrssbsy): New expander.
4149 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4151 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4153 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4155 (cmse_check_pointed_object): Likewise.
4157 2018-06-05 Martin Liska <mliska@suse.cz>
4159 PR gcov-profile/47618
4160 * doc/invoke.texi: Document how -fprofile-dir format
4163 2018-06-05 Richard Biener <rguenther@suse.de>
4165 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4166 removal pretend DOM info isn't available so we do not update
4167 it and only remove edges, not dominated blocks. Actually free
4168 DOM info in case we removed something. Remove unreachable blocks.
4169 (mfb_keep_latches): Work with either DOM info or marked backedges.
4170 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4171 first. Mark backedges if DOM info isn't available.
4172 (Re-)compute DOM info after cleanup_control_flow_pre.
4174 2018-06-05 Richard Biener <rguenther@suse.de>
4176 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4177 (locus_discrim_hasher::hash): Adjust.
4178 (locus_discrim_hasher::equal): Likewise.
4179 (next_discriminator_for_locus): Work on line directly.
4180 (same_line_p): Pass in expanded locus1 as well.
4181 (assign_discriminators): Avoid redundant location expansions.
4183 2018-06-05 Richard Biener <rguenther@suse.de>
4185 PR tree-optimization/86046
4186 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4187 if required after clearing TREE_ADDRESSABLE.
4189 2018-06-05 Richard Biener <rguenther@suse.de>
4191 PR tree-optimization/86047
4192 * tree-ssa-loop.c (for_each_index): Glob handling of all
4193 decls and constants and really handle all of them.
4195 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4198 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4199 qualifier_void_pointer and qualifier_const_void_pointer.
4200 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4201 (arm_init_builtins): Handle the above.
4202 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4203 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4204 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4207 2018-06-05 Martin Liska <mliska@suse.cz>
4209 * auto-profile.c (read_autofdo_file): Do not use
4210 gcov_ctr_summary struct.
4211 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4212 * coverage.c (struct counts_entry): Likewise.
4213 (read_counts_file): Read just single summary entry.
4214 (get_coverage_counts): Use gcov_summary struct.
4215 * coverage.h (get_coverage_counts): Likewise.
4216 * gcov-dump.c (dump_working_sets): Likewise.
4217 (tag_summary): Dump just single summary.
4218 * gcov-io.c (gcov_write_summary): Write just histogram
4220 (gcov_read_summary): Read just single summary.
4221 (compute_working_sets): Use gcov_summary struct.
4222 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4223 of GCOV_COUNTERS_SUMMABLE.
4224 (GCOV_COUNTERS_SUMMABLE): Remove.
4225 (GCOV_FIRST_VALUE_COUNTER): Replace with
4226 GCOV_COUNTER_V_INTERVAL.
4227 (struct gcov_ctr_summary): Remove.
4228 (struct gcov_summary): Directly use fields of former
4230 (compute_working_sets): Use gcov_summary struct.
4231 * gcov.c (read_count_file): Do not use ctrs fields.
4232 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4234 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4236 * profile.c: Likewise.
4237 * profile.h: Likewise.
4239 2018-06-05 Martin Liska <mliska@suse.cz>
4241 PR gcov-profile/84846
4242 * gcov.c (output_lines): Print working directory only
4243 in intermediate format.
4245 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4247 * config/s390/s390-builtin-types.def: Add void function type.
4248 * config/s390/s390-builtins.def: Use the function type for the
4251 2018-06-04 Jim Wilson <jimw@sifive.com>
4253 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4255 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4256 and maybe_eh_return. Change regno to unsigned int. Use new args to
4257 handle EH_RETURN_DATA_REGNO registers properly.
4258 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4259 (riscv_expand_epilogue): Update comment. Change argument name and
4260 type. Update code to use new name and type. Pass new args to
4261 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4263 * config/riscv/riscv.md (NORMAL_RETURN): New.
4264 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4265 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4266 (eh_return): Call gen_eh_return_internal and emit barrier.
4267 (eh_return_internal): Call riscv_expand_epilogue.
4269 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4271 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4272 bit_insertion field and declare can_be_merged_into method.
4273 (merged_store_group::can_be_merged_into): New method.
4274 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4275 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4276 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4278 2018-06-04 Richard Biener <rguenther@suse.de>
4280 PR tree-optimization/85955
4281 * builtins.c (fold_builtin_sincos): Convert pointers to
4282 destination to appropriate type before dereferencing.
4284 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4286 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4288 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4290 * expr.c (expand_expr_real_1): Force the operand into memory if
4291 its TYPE_MODE is BLKmode and if there is no integer mode for
4292 the number of bits being extracted.
4294 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4298 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4299 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4301 2018-06-04 Richard Biener <rguenther@suse.de>
4303 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4304 (cleanup_tree_cfg_noloop): ... single caller. Do
4305 start_recording_case_labels later.
4307 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4309 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4310 to _IMMINTRIN_H_INCLUDED.
4311 * config/i386/pconfigintrin.h: Ditto.
4312 * config/i386/waitpkgintrin.h: Ditto.
4313 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4314 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4315 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4316 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4317 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4318 waitpkgintrin.h and cldemoteintrin.h.
4320 2018-06-04 Richard Biener <rguenther@suse.de>
4322 PR tree-optimization/86038
4323 * tracer.c (find_best_successor): Check probability for
4324 being initialized, bail out if not.
4326 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4329 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4330 of bits to ignore when comparing architectures.
4332 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4334 PR tree-optimization/69615
4335 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4336 maximum or minimum of the type, try to merge it also as if
4337 range1 is + [-, x - 1] or + [x + 1, -].
4340 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4342 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4344 PR tree-optimization/86034
4345 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4346 the unsigned bitfield type in a bit insertion sequence if it does not
4347 have a larger precision than the bitfield size.
4348 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4350 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4352 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4354 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4356 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4357 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4358 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4359 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4361 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4363 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4364 Disable -fdelete-null-pointer-checks for ELF toolchain.
4366 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4367 Kito Cheng <kito.cheng@gmail.com>
4369 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4370 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4371 (nds32*-*-*): Add glibc and uclibc conditions.
4372 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4373 (TARGET_EXCEPT_UNWIND_INFO): Define.
4374 * config/nds32/elf.h: New file.
4375 * config/nds32/linux.h: New file.
4376 * config/nds32/nds32-elf.opt: New file.
4377 * config/nds32/nds32-linux.opt: New file.
4378 * config/nds32/nds32-fp-as-gp.c
4379 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4380 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4382 (nds32_asm_file_end): Ditto.
4383 (nds32_print_operand): Ditto.
4384 (nds32_insert_attributes): Ditto.
4385 (nds32_init_libfuncs): New function.
4386 (TARGET_HAVE_TLS): Define.
4387 (TARGET_INIT_LIBFUNCS): Define.
4388 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4390 (TARGET_ELF): Apply different mcmodel setting.
4391 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4392 been migrated into elf.h and linux.h files.
4393 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4394 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4395 (mcmodel): The content has been migrated into nds32-elf.opt and
4396 nds32-linux.opt files.
4397 * config/nds32/t-elf: New file.
4398 * config/nds32/t-linux: New file.
4400 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4401 Shiva Chen <shiva0217@gmail.com>
4403 * config/nds32/constants.md (unspec_volatile_element): Add
4404 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4405 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4407 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4408 (make_pass_nds32_fp_as_gp): Declare.
4409 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4411 (nds32_asm_function_end_prologue): Remove unused asm output.
4412 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4413 (nds32_asm_file_start): Output necessary fp_as_gp information.
4414 (nds32_option_override): Adjust register usage.
4415 (nds32_expand_prologue): Consider fp_as_gp situation.
4416 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4417 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4420 (simple_return): Ditto.
4421 (omit_fp_begin): Output special directive for fp_as_gp.
4422 (omit_fp_end): Output special directive for fp_as_gp.
4423 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4424 mforbid-fp-as-gp): New options.
4426 2018-06-01 Mark Wielaard <mark@klomp.org>
4428 * dwarf2out.c (dwarf2out_finish): Remove generation of
4429 DW_AT_loclists_base.
4431 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4433 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4434 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4435 (struct merged_store_group): Add bit_insertion field.
4436 (dump_char_array): Use standard hexadecimal format.
4437 (merged_store_group::merged_store_group): Set bit_insertion to false.
4438 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4439 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4440 also print the mask in the dump file.
4441 (pass_store_merging::gate): Minor tweak.
4442 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4443 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4444 stores with INTEGER_CST stores.
4445 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4446 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4447 and use it throughout. Generate bit insertion sequences if need be.
4448 (pass_store_merging::process_store): Remove redundant condition.
4449 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4451 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4453 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4454 the 128-bit floating point types. Fix function comment.
4456 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4458 * config/aarch64/aarch64-simd.md
4459 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4461 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4464 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4466 PR tree-optimization/85989
4467 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4469 (backprop::intersect_uses): Check it when deciding whether this
4470 is a backedge reference.
4471 (backprop::process_block): Add each phi to m_visited_phis
4472 after visiting it, then clear it at the end.
4474 2018-06-01 Richard Biener <rguenther@suse.de>
4476 * tree-vectorizer.h (vect_dr_stmt): New function.
4477 (vect_get_load_cost): Adjust.
4478 (vect_get_store_cost): Likewise.
4479 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4480 Use vect_dr_stmt instead of DR_SMTT.
4481 (vect_record_base_alignments): Likewise.
4482 (vect_calculate_target_alignment): Likewise.
4483 (vect_compute_data_ref_alignment): Likewise and make static.
4484 (vect_update_misalignment_for_peel): Likewise.
4485 (vect_verify_datarefs_alignment): Likewise.
4486 (vector_alignment_reachable_p): Likewise.
4487 (vect_get_data_access_cost): Likewise. Pass down
4488 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4489 (vect_get_peeling_costs_all_drs): Likewise.
4490 (vect_peeling_hash_get_lowest_cost): Likewise.
4491 (vect_enhance_data_refs_alignment): Likewise.
4492 (vect_find_same_alignment_drs): Likewise.
4493 (vect_analyze_data_refs_alignment): Likewise.
4494 (vect_analyze_group_access_1): Likewise.
4495 (vect_analyze_group_access): Likewise.
4496 (vect_analyze_data_ref_access): Likewise.
4497 (vect_analyze_data_ref_accesses): Likewise.
4498 (vect_vfa_segment_size): Likewise.
4499 (vect_small_gap_p): Likewise.
4500 (vectorizable_with_step_bound_p): Likewise.
4501 (vect_prune_runtime_alias_test_list): Likewise.
4502 (vect_analyze_data_refs): Likewise.
4503 (vect_supportable_dr_alignment): Likewise.
4504 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4505 (vect_gen_prolog_loop_niters): Likewise.
4506 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4507 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4509 (vect_recog_mask_conversion_pattern): Likewise.
4510 (vect_try_gather_scatter_pattern): Likewise.
4511 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4512 to vect_get_store_cost.
4513 (vect_get_store_cost): Get stmt_info instead of DR.
4514 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4515 (vect_get_load_cost): Get stmt_info instead of DR.
4517 2018-06-01 Richard Biener <rguenther@suse.de>
4520 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4521 vars wrapped in MEM_REFs.
4523 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4525 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4526 Fix subreg tests so that we only return a choice between
4527 GENERAL_REGS and FP_REGS if the original classes included both.
4529 2018-06-01 Richard Biener <rguenther@suse.de>
4532 * tree-ssa-structalias.c (get_function_part_constraint):
4533 Handle NULL fi->decl.
4534 (find_func_aliases_for_call): Properly handle indirect
4535 fi from direct call.
4536 (find_func_clobbers): Likewise.
4537 (ipa_pta_execute): Likewise.
4538 (create_variable_info_for): For functions that are ifunc_resolver
4539 resolve to a varinfo that contains the result of the resolver call.
4540 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4543 2018-05-31 Michael Collison <michael.collison@arm.com>
4545 * config/aarch64/aarch64.md:
4546 (*fix_to_zero_extenddfdi2): New pattern.
4547 * gcc.target/aarch64/fix_extend1.c: New testcase.
4549 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4553 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4554 and BUILT_IN_STRNCMP_EQ.
4555 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4556 BUILT_IN_STRNCMP_EQ.
4557 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4558 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4559 (gimple_fold_builtin): Likewise.
4560 * tree-ssa-strlen.c (compute_string_length): New function.
4561 (determine_min_obsize): New function.
4562 (handle_builtin_string_cmp): New function to handle calls to
4563 string compare functions.
4564 (strlen_optimize_stmt): Add handling to builtin string compare
4566 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4567 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4568 * tree.c (build_common_builtin_nodes): Add new defines of
4569 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4571 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4574 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4575 functions with naked attribute.
4577 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4579 * config/i386/sse.md (avx_vec_concat<mode>):
4580 Substitute concat_tg_mode mode attribute with xtg_mode.
4581 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4582 (concat_tg_mode): Remove mode attribute.
4584 2018-05-31 Martin Sebor <msebor@redhat.com>
4587 * calls.c (alloc_max_size): Correct a logic error/typo.
4588 Treat excessive arguments as infinite. Warn for invalid arguments.
4589 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4591 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4594 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4595 and movx for Haswell.
4597 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4598 Cesar Philippidis <cesar@codesourcery.com>
4601 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4602 when emitting error on private/firstprivate reductions.
4603 * omp-low.c (lower_omp_target): Avoid reference-type processing
4604 on pointers for firstprivate clause.
4606 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4608 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4611 * config/aarch64/aarch64-simd.md
4612 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4613 (aarch64_ld1_x3_<mode>): Likewise
4614 (aarch64_st1x2<VALLDIF:mode>): Likewise
4615 (aarch64_st1_x2_<mode>): Likewise
4616 (aarch64_st1x3<VALLDIF:mode>): Likewise
4617 (aarch64_st1_x3_<mode>): Likewise
4618 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4619 (vld1_s8_x3): Likewise.
4620 (vld1_u16_x3): Likewise.
4621 (vld1_s16_x3): Likewise.
4622 (vld1_u32_x3): Likewise.
4623 (vld1_s32_x3): Likewise.
4624 (vld1_u64_x3): Likewise.
4625 (vld1_s64_x3): Likewise.
4626 (vld1_f16_x3): Likewise.
4627 (vld1_f32_x3): Likewise.
4628 (vld1_f64_x3): Likewise.
4629 (vld1_p8_x3): Likewise.
4630 (vld1_p16_x3): Likewise.
4631 (vld1_p64_x3): Likewise.
4632 (vld1q_u8_x3): Likewise.
4633 (vld1q_s8_x3): Likewise.
4634 (vld1q_u16_x3): Likewise.
4635 (vld1q_s16_x3): Likewise.
4636 (vld1q_u32_x3): Likewise.
4637 (vld1q_s32_x3): Likewise.
4638 (vld1q_u64_x3): Likewise.
4639 (vld1q_s64_x3): Likewise.
4640 (vld1q_f16_x3): Likewise.
4641 (vld1q_f32_x3): Likewise.
4642 (vld1q_f64_x3): Likewise.
4643 (vld1q_p8_x3): Likewise.
4644 (vld1q_p16_x3): Likewise.
4645 (vld1q_p64_x3): Likewise.
4646 (vst1_s64_x2): Likewise.
4647 (vst1_u64_x2): Likewise.
4648 (vst1_f64_x2): Likewise.
4649 (vst1_s8_x2): Likewise.
4650 (vst1_p8_x2): Likewise.
4651 (vst1_s16_x2): Likewise.
4652 (vst1_p16_x2): Likewise.
4653 (vst1_s32_x2): Likewise.
4654 (vst1_u8_x2): Likewise.
4655 (vst1_u16_x2): Likewise.
4656 (vst1_u32_x2): Likewise.
4657 (vst1_f16_x2): Likewise.
4658 (vst1_f32_x2): Likewise.
4659 (vst1_p64_x2): Likewise.
4660 (vst1q_s8_x2): Likewise.
4661 (vst1q_p8_x2): Likewise.
4662 (vst1q_s16_x2): Likewise.
4663 (vst1q_p16_x2): Likewise.
4664 (vst1q_s32_x2): Likewise.
4665 (vst1q_s64_x2): Likewise.
4666 (vst1q_u8_x2): Likewise.
4667 (vst1q_u16_x2): Likewise.
4668 (vst1q_u32_x2): Likewise.
4669 (vst1q_u64_x2): Likewise.
4670 (vst1q_f16_x2): Likewise.
4671 (vst1q_f32_x2): Likewise.
4672 (vst1q_f64_x2): Likewise.
4673 (vst1q_p64_x2): Likewise.
4674 (vst1_s64_x3): Likewise.
4675 (vst1_u64_x3): Likewise.
4676 (vst1_f64_x3): Likewise.
4677 (vst1_s8_x3): Likewise.
4678 (vst1_p8_x3): Likewise.
4679 (vst1_s16_x3): Likewise.
4680 (vst1_p16_x3): Likewise.
4681 (vst1_s32_x3): Likewise.
4682 (vst1_u8_x3): Likewise.
4683 (vst1_u16_x3): Likewise.
4684 (vst1_u32_x3): Likewise.
4685 (vst1_f16_x3): Likewise.
4686 (vst1_f32_x3): Likewise.
4687 (vst1_p64_x3): Likewise.
4688 (vst1q_s8_x3): Likewise.
4689 (vst1q_p8_x3): Likewise.
4690 (vst1q_s16_x3): Likewise.
4691 (vst1q_p16_x3): Likewise.
4692 (vst1q_s32_x3): Likewise.
4693 (vst1q_s64_x3): Likewise.
4694 (vst1q_u8_x3): Likewise.
4695 (vst1q_u16_x3): Likewise.
4696 (vst1q_u32_x3): Likewise.
4697 (vst1q_u64_x3): Likewise.
4698 (vst1q_f16_x3): Likewise.
4699 (vst1q_f32_x3): Likewise.
4700 (vst1q_f64_x3): Likewise.
4701 (vst1q_p64_x3): Likewise.
4703 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4705 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4706 user_label_prefix to name.
4708 * tree-core.h: Update comment about the format of NAME string
4709 passed to handler in attribute_spec.
4711 * config/msp430/msp430.md: Remove erroneous subreg expression from
4712 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4713 zero_extend{q,h}isi2.
4715 2018-05-30 Borislav Petkov <bp@suse.de>
4717 * doc/extend.texi: Document some architecture specific
4718 constraints and sort entries.
4720 2018-05-30 Martin Sebor <msebor@redhat.com>
4723 * builtins.c (expand_builtin_stpcpy_1): New function.
4724 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4725 only if the former succeeds.
4727 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4729 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4732 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4734 * doc/invoke.texi (-flinker-output): Document
4736 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4738 * passes.c (ipa_write_summaries): Only modify statements if body
4740 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4741 incrementally linking.
4742 (ipa_passes): Likewise.
4743 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4744 pass down resolution info.
4745 * common.opt (flag_incremental_link): Update info.
4746 * gcc.c (plugin specs): Turn flinker-output=* to
4747 -plugin-opt=-linker-output-known
4748 * toplev.c (compile_file): Also cut compilation when doing incremental
4750 * flag-types. (enum lto_partition_model): Add
4751 LTO_LINKER_OUTPUT_NOLTOREL.
4752 (invoke.texi): Add -flinker-output docs.
4753 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4754 link same way as WPA; do not stream in dead initializers.
4756 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4757 darf2out_register_external_decl): Support incremental link.
4759 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4761 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4763 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4765 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4766 it down to simple_object_copy_lto_debug_sections.
4767 (run_gcc): Determine incremental LTO link time and configure
4768 lto1 into non-wpa mode, disable renaming of debug sections.
4770 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4772 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4773 descriptions of various incorrectly documented functions.
4775 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4778 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4779 address check not strict.
4781 2018-05-30 Richard Biener <rguenther@suse.de>
4783 PR tree-optimization/85964
4784 * tracer.c (better_p): Drop initialized count check, we only
4785 call the function with initialized counts now.
4786 (find_best_successor): Do find a best edge if one
4787 has uninitialized count.
4788 (find_best_predecessor): Likewise. Do BB frequency check only
4789 if count is initialized.
4791 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4793 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4794 (aarch64_ldrstr_offset_compare): New.
4795 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4796 load/store orderings.
4797 (aarch64_gen_adjusted_ldpstp): Likewise.
4799 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4801 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4802 Check for subset of GENERAL_REGS and FP_REGS.
4803 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4806 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4808 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4809 and wi::to_poly_offset. Add the current offset and then check
4810 whether the sum fits, rather than using an unchecked addition of
4811 a checked term. Check for a shwi rather than a uhwi.
4812 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4813 (store_constructor): Use poly_int_tree_p.
4814 (expand_expr_real_1): Likewise.
4815 * function.c (assign_temp): Likewise.
4816 * fold-const.c (const_binop): Use poly_int_tree_p and
4818 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4820 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4821 to_poly_offset for MEM offsets.
4822 * ipa-icf.c (sem_variable::equals): Likewise.
4823 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4824 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4825 wi::to_poly_offset for BIT_FIELD_REF offsets.
4826 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4828 * var-tracking.c (emit_note_insn_var_location): Use
4829 tree_to_poly_uint64.
4831 2018-05-29 Jim Wilson <jimw@sifive.com>
4833 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4835 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4838 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4839 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4841 (sse4_1_round<mode>2): Use nonimmediate_operand
4842 for operand 1 predicate.
4844 2018-05-29 Martin Sebor <msebor@redhat.com>
4845 Richard Biener <rguenther@suse.de>
4848 * calls.c (get_size_range): Call determine_value_range instead
4849 of get_value_range..
4850 * tree-vrp.h (determine_value_range): Declared new function.
4851 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4853 2018-05-29 Richard Biener <rguenther@suse.de>
4855 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4856 sure to use non-pattern stmts for get_earlier_stmt arguments.
4857 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4858 called on pattern stmts.
4859 (get_later_stmt): Likewise.
4861 2018-05-29 Martin Liska <mliska@suse.cz>
4863 PR gcov-profile/85759
4864 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4867 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4869 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4871 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4875 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4876 VEC_PACK_FLOAT_EXPR): New tree codes.
4877 * tree-pretty-print.c (op_code_prio): Handle
4878 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4879 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4880 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4881 * tree-inline.c (estimate_operator_cost): Likewise.
4882 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4883 * fold-const.c (const_binop): Likewise.
4884 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4885 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4886 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4887 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4888 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4889 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4890 * expr.c (expand_expr_real_2): Likewise.
4891 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4892 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4893 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4895 * optabs.c (expand_widen_pattern_expr): For
4896 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4897 sign from result type rather than operand's type.
4898 (expand_binop_directly): For vec_packu_float_optab and
4899 vec_packs_float_optab allow result type to be different from operand's
4901 * optabs-tree.c (optab_for_tree_code): Handle
4902 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4903 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4904 * tree-vect-generic.c (expand_vector_operations_1): Handle
4905 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4906 VEC_PACK_FLOAT_EXPR.
4907 * tree-vect-stmts.c (supportable_widening_operation): Handle
4909 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4910 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4911 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4912 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4913 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4915 (vec_pack<floatprefix>_float_<mode>): New expander.
4916 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4918 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4919 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4920 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4921 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4922 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4924 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4925 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4926 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4927 VEC_PACK_FLOAT_EXPR): Document.
4929 2018-05-29 Richard Biener <rguenther@suse.de>
4931 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4933 (stmt_vec_info_vec): Make pointer.
4934 (init_stmt_vec_info_vec): Remove.
4935 (free_stmt_vec_info_vec): Likewise.
4936 (set_stmt_vec_info_vec): New function.
4937 (free_stmt_vec_infos): Likewise.
4938 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4939 (set_vinfo_for_stmt): Likewise.
4940 (get_earlier_stmt): Likewise.
4941 (get_later_stmt): Likewise.
4942 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4943 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4944 (vec_info::~vec_info): Free stmt_vec_infos.
4945 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4946 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4947 (pass_slp_vectorize::execute): Likewise.
4948 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4949 (free_stmt_vec_info_vec): Likewise.
4950 (set_stmt_vec_info_vec): New function.
4951 (free_stmt_vec_infos): Likewise.
4952 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4953 the global stmt_vec_info_vec.
4954 * tree-parloops.c (gather_scalar_reductions): Use
4955 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4958 2018-05-29 Richard Biener <rguenther@suse.de>
4960 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4962 2018-05-29 Martin Liska <mliska@suse.cz>
4963 David Malcolm <dmalcolm@redhat.com>
4965 * vec.c (test_reverse): New.
4966 (vec_c_tests): Add new test.
4967 * vec.h (vl_ptr>::reverse): New function.
4969 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4971 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4973 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4976 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4978 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4980 2018-05-28 Richard Biener <rguenther@suse.de>
4982 PR tree-optimization/85933
4983 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4984 look at stmts marked as vectorizable.
4986 2018-05-28 Richard Biener <rguenther@suse.de>
4988 PR tree-optimization/85934
4989 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4990 vector boolean check before scalar optimization.
4992 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4994 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4997 2018-05-28 Mark Wielaard <mark@klomp.org>
4999 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5000 if it is an expression containing a minus sign.
5002 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5004 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5006 2018-05-27 Paul Koning <ni1d@arrl.net>
5008 * config/pdp11/pdp11.md (truncsihi2): Remove.
5010 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5011 Chung-Ju Wu <jasonwucj@gmail.com>
5013 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5015 (unaligned_store_dw): Ditto.
5016 * config/nds32/nds32-memory-manipulation.c
5017 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5018 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5019 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5020 (emit_setmem_word_loop): Rename to ...
5021 (emit_setmem_doubleword_loop): ... this.
5022 (nds32_gen_dup_4_byte_to_word_value): New function.
5023 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5024 (nds32_expand_setmem_loop): Refine implementation.
5025 (nds32_expand_setmem_loop_v3m): Ditto.
5026 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5029 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5031 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5033 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5035 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5036 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5037 machine->attr_no_prologue_p.
5038 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5039 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5040 (nds32_expand_epilogue): Consider attr_naked_p.
5041 (nds32_expand_epilogue_v3pop): Likewise.
5042 (nds32_can_use_return_insn): Likewise.
5043 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5044 attr_no_prologue_p fields.
5045 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5047 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5050 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5052 * config/i386/sse.md
5053 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5055 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5057 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5059 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5061 (*<floatsuffix>floatv2div2sf2): Rename to ...
5062 (*float<floatunssuffix>v2div2sf2): ... this.
5063 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5064 (float<floatunssuffix>v2div2sf2_mask): ... this.
5065 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5066 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5067 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5069 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5071 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5073 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5075 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5077 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5079 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5080 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5081 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5082 gen_ufix_truncv8dfv8si2.
5083 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5084 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5085 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5086 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5087 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5088 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5089 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5090 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5092 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5096 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5098 2018-05-25 Jim Wilson <jimw@sifive.com>
5100 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5101 * config/riscv/riscv.c (struct machine_function): Add
5102 interrupt_handler_p and attribute_checked_p fields.
5103 (riscv_attribute_table): Add interrupt.
5104 (riscv_interrupt_type_p): New.
5105 (riscv_save_reg_p): Save extra regs for interrupt handler.
5106 (riscv_use_save_libcall): Return false for interrupt handler.
5107 (riscv_first_stack_step): Add forward declaration.
5108 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5109 for interrupt handler with large frame. Use it for saved reg list.
5110 (riscv_expand_prologue): Move flag_stack_usage_info support to
5111 eliminate duplication.
5112 (riscv_expand_epilogue): Generate mret for interrupt handler.
5113 (riscv_epilogue_uses): New.
5114 (riscv_can_use_return_insn): Return false for interrupt handler.
5115 (riscv_function_ok_for_sibcall): Likewise.
5116 (riscv_set_current_function): Add interrupt handler support.
5117 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5118 * config/riscv/riscv.md (UNSPECV_MRET): New.
5120 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5122 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5124 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5126 PR tree-optimization/85712
5127 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5128 this candidate has already been replaced in-situ by a copy.
5130 2018-05-25 Jason Merrill <jason@redhat.com>
5132 PR c++/80485 - inline function non-zero address.
5133 * symtab.c (nonzero_address): Check DECL_COMDAT.
5135 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5138 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5139 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5140 extension of SImode operation. Use const123_operand predicate.
5141 (*saddsi_1): Remove.
5142 (*saddl_se_1): Ditto.
5144 (*ssubl_se_1): Ditto.
5145 * config/alpha/predicates.md (const123_operand): New predicate.
5146 * config/alpha/constraints.md (P): Use IN_RANGE.
5148 2018-05-25 Richard Biener <rguenther@suse.de>
5150 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5152 (ref_maybe_used_by_stmt_p): Likewise.
5153 (stmt_may_clobber_ref_p): Likewise.
5154 (stmt_may_clobber_ref_p_1): Likewise.
5155 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5157 (ref_maybe_used_by_stmt_p): Likewise.
5158 (stmt_may_clobber_ref_p): Likewise.
5159 (stmt_may_clobber_ref_p_1): Likewise.
5160 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5161 the alias oracle to disambiguate DRs with stmts DR analysis
5163 (vect_analyze_data_refs): Do not give up on not analyzable
5164 DRs for BB vectorization. Remove code truncating the dataref
5167 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5170 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5171 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5172 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5174 2018-05-25 Richard Biener <rguenther@suse.de>
5176 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5177 function, combining stmt data ref gathering and fatal analysis
5179 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5180 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5181 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5182 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5183 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5185 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5187 PR tree-optimization/85720
5188 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5189 SCC if all partitions are builtins.
5190 (version_loop_by_alias_check): New parameter. Generate cancelable
5191 runtime alias check if all partitions are builtins.
5192 (distribute_loop): Update call to above function.
5194 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5196 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5197 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5198 (parm_default_def_partition_arg): Ditto.
5199 (set_parm_default_def_partition): Ditto.
5200 (get_parm_default_def_partitions): Ditto and make it static.
5201 (get_undefined_value_partitions): Ditto and make it static.
5202 (remove_ssa_form): Refactor call to init_var_map here.
5203 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5204 computation for loop region.
5205 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5206 (register_default_def): Delete.
5207 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5208 (parm_default_def_partition_arg): Ditto.
5209 (set_parm_default_def_partition): Ditto.
5210 (get_parm_default_def_partitions): Ditto and make it static.
5211 (get_undefined_value_partitions): Ditto and make it static.
5212 (coalesce_with_default, coalesce_with_default): Update comment.
5213 (create_coalesce_list_for_region): New func factored out from
5214 create_outofssa_var_map.
5215 (populate_coalesce_list_for_outofssa): New func factored out from
5216 create_outofssa_var_map and coalesce_ssa_name.
5217 (create_outofssa_var_map): Delete.
5218 (coalesce_ssa_name): Refactor to support live range computation.
5219 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5220 (get_parm_default_def_partitions): Delete.
5221 (get_undefined_value_partitions): Ditto.
5222 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5223 computation for loop region.
5224 (new_tree_live_info, loe_visit_block): Ditto.
5225 (live_worklist, set_var_live_on_entry): Ditto.
5226 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5227 * tree-ssa-live.h (struct _var_map): New fields.
5228 (init_var_map): Change decl.
5229 (region_contains_p): New.
5231 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5233 * tree-ssa-live.h (live_merge_and_clear): Delete.
5235 2018-05-25 Richard Biener <rguenther@suse.de>
5238 * tree-dump.c (dequeue_and_dump): Remove access to removed
5239 operand 2 of a SWITCH_EXPR.
5241 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5243 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5244 multiplication and division.
5245 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5246 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5247 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5248 (cond_udiv_optab, cond_umod_optab): New optabs.
5249 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5250 (IFN_COND_RDIV): New internal functions.
5251 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5252 TRUNC_MOD_EXPR and RDIV_EXPR.
5253 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5254 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5256 (SVE_INT_BINARY): Include mult.
5257 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5258 (optab, sve_int_op): Handle mult.
5259 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5261 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5262 for SVE_INT_BINARY_SD.
5264 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5266 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5267 (optab, sve_int_op): Handle div and udiv.
5268 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5269 for SVE_INT_BINARY_SD.
5270 (*<optab><mode>3): New insn for the same.
5272 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5274 * tree-vect-patterns.c: Include predict.h.
5275 (vect_recog_divmod_pattern): Restrict check for division support
5276 to when optimizing for size.
5278 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5280 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5281 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5282 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5283 (gimple_match_op::set_op): Likewise.
5284 (gimple_resimplify4): Declare.
5285 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5286 (expr::gen_transform): Likewise.
5287 (decision_tree::gen): Generate a simplification routine for 4 operands.
5288 * gimple-match-head.c (gimple_simplify): Add an overload for
5289 4 operands. In the top-level function, handle up to 4 call
5290 arguments and call gimple_resimplify4.
5291 (gimple_resimplify4): New function.
5292 (build_call_internal): Pass a fourth operand.
5293 (maybe_push_to_seq): Likewise.
5294 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5295 Fold VEC_COND_EXPRs of an operation and a default value into
5296 an IFN_COND_* function if possible.
5297 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5299 (SVE_COND_FP_BINARY): Include them.
5300 (optab, sve_fp_op): Handle them.
5301 (SVE_INT_BINARY_REV): New code iterator.
5302 (SVE_COND_FP_BINARY_REV): New int iterator.
5303 (commutative): New int attribute.
5304 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5306 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5308 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5309 (*cond_<optab><mode>): New patterns for reversed operands.
5311 2018-05-25 Richard Biener <rguenther@suse.de>
5313 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5314 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5315 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5316 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5317 * tree-vect-data-refs.c (everywhere): Adjust users.
5318 * tree-vect-loop.c (everywhere): Likewise.
5319 * tree-vect-slp.c (everywhere): Likewise.
5320 * tree-vect-stmts.c (everywhere): Likewise.
5321 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5323 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5325 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5327 (gcc_cv_as_section_exclude): ... this.
5328 Try Solaris as #exclude syntax.
5329 * configure: Regenerate.
5330 * config.in: Regenerate.
5331 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5333 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5334 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5336 * varasm.c (default_elf_asm_named_section): Don't check if
5337 HAVE_GAS_SECTION_EXCLUDE is defined.
5339 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5341 * doc/md.texi: Update the documentation of the cond_* optabs
5342 to mention the new final operand. Fix GET_MODE_NUNITS call.
5343 Describe the scalar case too.
5344 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5345 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5347 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5348 (get_conditional_internal_fn): Update comment.
5349 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5350 accumulator value as a final argument to conditional functions.
5351 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5352 a define_expand and add an "else" operand. Assert for now that
5353 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5354 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5355 (*cond_<optab><mode>): New patterns.
5356 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5357 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5358 (UNSPEC_COND_EOR): Delete.
5359 (optab): Remove associated mappings.
5360 (SVE_INT_BINARY): New code iterator.
5361 (sve_int_op): Remove int attribute and add "minus" to the code
5363 (SVE_COND_INT_OP): Delete.
5364 (SVE_COND_FP_OP): Rename to...
5365 (SVE_COND_FP_BINARY): ...this.
5367 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5369 * optabs.c (can_reuse_operands_p): New function.
5370 (maybe_legitimize_operands): Try to reuse the results for
5373 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5375 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5376 Add {q} suffix to insn mnemonic.
5378 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5380 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5381 (msp430_warn_func_return): New.
5383 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5385 * fold-const.c (tree_nonzero_bits): New function.
5386 * fold-const.h (tree_nonzero_bits): Likewise.
5387 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5388 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5390 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5394 * varasm.c (assemble_alias): Check ifunc_resolver only on
5397 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5400 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5401 when memory input operand is handled.
5403 2018-05-24 Luis Machado <luis.machado@linaro.org>
5405 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5407 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5409 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5411 * match.pd: Delay FMA folds until after vectorization.
5413 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5416 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5417 address check not strict.
5419 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5421 * gimple-match.h (gimple_match_op): New class.
5422 (mprts_hook): Replace parameters with a gimple_match_op *.
5423 (maybe_build_generic_op): Likewise.
5424 (gimple_simplified_result_is_gimple_val): Replace parameters with
5425 a const gimple_match_op *.
5426 (gimple_simplify): Replace code_helper * and tree * parameters with
5427 a gimple_match_op * parameter.
5428 (gimple_resimplify1): Replace code_helper *, tree and tree *
5429 parameters with a gimple_match_op * parameter.
5430 (gimple_resimplify2): Likewise.
5431 (gimple_resimplify3): Likewise.
5432 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5433 parameters with a gimple_match_op * parameter.
5434 * gimple-match-head.c (gimple_simplify): Change prototypes of
5435 auto-generated functions to take a gimple_match_op * instead of
5436 separate code_helper * and tree * parameters. Make the same
5437 change in the top-level overload and update calls to the
5438 gimple_resimplify routines. Update calls to the auto-generated
5439 functions and to maybe_push_res_to_seq in the publicly-facing
5440 operation-specific gimple_simplify overloads.
5441 (gimple_match_op::MAX_NUM_OPS): Define.
5442 (gimple_resimplify1): Replace rcode and ops with a single res_op
5443 parameter. Update call to gimple_simplify.
5444 (gimple_resimplify2): Likewise.
5445 (gimple_resimplify3): Likewise.
5446 (mprts_hook): Replace parameters with a gimple_match_op *.
5447 (maybe_build_generic_op): Likewise.
5448 (build_call_internal): Replace type, nargs and ops with
5449 a gimple_match_op *.
5450 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5451 with a single gimple_match_op *. Update calls to mprts_hook,
5452 build_call_internal and gimple_simplified_result_is_gimple_val.
5453 Factor out code that is common to the tree_code and combined_fn cases.
5454 * genmatch.c (expr::gen_transform): Replace tem_code and
5455 tem_ops with a gimple_match_op called tem_op. Update calls
5456 to the gimple_resimplify functions and maybe_push_res_to_seq.
5457 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5458 res_ops. Update call to the gimple_resimplify functions.
5459 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5460 (decision_tree::gen): Make the functions take a gimple_match_op *
5461 called res_op instead of separate res_code and res_ops parameters.
5462 Update call accordingly.
5463 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5464 and ops with a single res_op parameter. Update calls to
5465 maybe_build_generic_op and maybe_push_res_to_seq.
5466 (fold_stmt_1): Update calls to gimple_simplify and
5467 replace_stmt_with_simplification.
5468 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5469 and gimple_simplified_result_is_gimple_val.
5470 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5472 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5473 with a gimple_match_op *.
5474 (vn_nary_build_or_lookup): Likewise. Update call to
5475 vn_nary_build_or_lookup_1.
5476 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5477 gimple_match_op *. Update calls to the gimple_resimplify routines
5478 and to gimple_simplified_result_is_gimple_val.
5479 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5480 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5481 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5482 (visit_nary_op): Likewise.
5483 (visit_reference_op_load): Likewise.
5485 2018-05-23 Luis Machado <luis.machado@linaro.org>
5487 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5488 modifier for printing the step amount.
5490 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5493 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5496 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5498 * doc/sourcebuild.texi (Endianness): New subsubsection.
5500 2018-05-23 Luis Machado <luis.machado@linaro.org>
5502 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5503 <prefetch_dynamic_strides>: New const bool field.
5504 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5505 prefetch_dynamic_strides.
5506 (exynosm1_prefetch_tune): Likewise.
5507 (thunderxt88_prefetch_tune): Likewise.
5508 (thunderx_prefetch_tune): Likewise.
5509 (thunderx2t99_prefetch_tune): Likewise.
5510 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5512 (aarch64_override_options_internal): Update to set
5513 PARAM_PREFETCH_DYNAMIC_STRIDES.
5514 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5515 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5516 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5517 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5518 prefetch-dynamic-strides setting.
5520 2018-05-23 Luis Machado <luis.machado@linaro.org>
5522 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5523 <minimum_stride>: New const int field.
5524 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5525 minimum_stride field defaulting to -1.
5526 (exynosm1_prefetch_tune): Likewise.
5527 (thunderxt88_prefetch_tune): Likewise.
5528 (thunderx_prefetch_tune): Likewise.
5529 (thunderx2t99_prefetch_tune): Likewise.
5530 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5531 <default_opt_level>: Set to 3.
5532 (aarch64_override_options_internal): Update to set
5533 PARAM_PREFETCH_MINIMUM_STRIDE.
5534 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5535 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5536 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5537 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5538 stride is constant and is below the minimum stride threshold.
5540 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5542 * config/arm/arm-cpus.in (mode26): Delete.
5543 (armv4): Delete mode26 reference.
5544 * config/arm/arm.c (arm_configure_build_target): Delete use of
5547 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5549 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5551 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5552 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5554 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5555 (floatunsdidf2): Ditto.
5557 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5559 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5560 (fixuns_trunc<mode>si2_avx512f): Ditto.
5561 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5562 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5563 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5565 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5567 PR rtl-optimization/79985
5568 * df-scan.c (df_insn_refs_collect): Remove special case for
5569 global registers and asm statements.
5571 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5573 * extend.texi (Global Register Variables): Rewrite the bullet list.
5574 Note that the register is available for allocation. Note that access
5575 via inline asm must use constraints. Add note about async-signal
5576 handlers. Remove paragraph about automagic register selection.
5578 2018-05-23 Richard Biener <rguenther@suse.de>
5580 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5581 of fixed offset from memset VN.
5583 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5585 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5587 (alloc_cand_and_find_basis): Initialize first_interp field.
5588 (slsr_process_mul): Modify first_interp field.
5589 (slsr_process_add): Likewise.
5590 (slsr_process_cast): Modify first_interp field for each new
5592 (slsr_process_copy): Likewise.
5593 (dump_candidate): Dump first_interp field.
5594 (replace_mult_candidate): Process all interpretations, not just
5596 (replace_rhs_if_not_dup): Likewise.
5597 (replace_one_candidate): Likewise.
5599 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5601 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5603 (aarch64_needs_frame_chain): New function.
5604 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5606 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5609 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5610 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5611 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5612 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5613 as true for strict-align.
5614 (aarch64_can_inline_p): Perform checks even when callee has no
5615 attributes to check for strict alignment.
5616 * doc/extend.texi (AArch64 Function Attributes): Document
5618 * doc/invoke.texi: (AArch64 Options): Likewise.
5620 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5622 PR tree-optimization/85853
5623 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5624 the handling of the root of the node to...
5625 (vect_slp_analyze_node_operations_1): ...this new function,
5626 and run the whole thing with the child nodes' def types
5627 set according to their SLP node's def type.
5629 2018-05-23 Richard Biener <rguenther@suse.de>
5632 * tree-data-ref.c (create_runtime_alias_checks): Defer
5633 and ignore overflow warnings.
5635 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5637 PR tree-optimization/85822
5638 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5641 2018-05-23 Richard Biener <rguenther@suse.de>
5643 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5644 memset constants via native_interpret_expr.
5646 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5649 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5651 (cgraph_node::create_alias): Likewise.
5652 (cgraph_node::get_availability): Check ifunc_resolver instead
5653 of looking up ifunc attribute.
5654 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5655 * varasm.c (do_assemble_alias): Likewise.
5656 (assemble_alias): Likewise.
5657 (default_binds_local_p_3): Likewise.
5658 * cgraph.h (cgraph_node): Add ifunc_resolver.
5659 (cgraph_node::only_called_directly_or_aliased_p): Return false
5661 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5663 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5664 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5665 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5666 instead of looking up ifunc attribute.
5668 2018-05-22 Luis Machado <luis.machado@linaro.org>
5670 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5672 2018-05-22 Martin Sebor <msebor@redhat.com>
5675 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5676 only when expasion succeeds.
5677 (expand_builtin_strcmp): Same.
5678 (expand_builtin_strncmp): Same.
5680 2018-05-22 Martin Sebor <msebor@redhat.com>
5682 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5684 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5685 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5687 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5688 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5690 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5691 Add check for legitimate_address.
5692 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5693 (aarch64_swap_ldrstr_operands): New.
5694 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5697 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5698 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5700 * config/aarch64/aarch64.md: New patterns to generate stp
5702 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5703 single words and double words.
5704 (load_pair_sw, load_pair_dw): Likewise.
5705 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5707 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5709 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5710 for different mode ldpstp and add peephole for merged zero stores.
5712 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5714 (aarch64_gen_store_pair): Rename calls to match new patterns.
5715 (aarch64_gen_load_pair): Rename calls to match new patterns.
5716 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5717 (load_pair<DREG:mode><DREG2:mode>): ... This.
5718 (store_pair<mode>): Rename to...
5719 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5720 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5722 (V_INT_EQUIV): Handle SImode.
5723 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5726 2018-05-22 Martin Sebor <msebor@redhat.com>
5729 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5730 or ajust the presumed bound on an operation to avoid unnecessary
5733 2018-05-22 Martin Sebor <msebor@redhat.com>
5735 PR tree-optimization/85826
5736 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5737 assuming that a DECL necesarily has a constant size.
5739 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5742 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5744 2018-05-22 Richard Biener <rguenther@suse.de>
5746 PR tree-optimization/85834
5747 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5748 non-constant and non-zero memset arguments.
5750 2018-05-22 Martin Liska <mliska@suse.cz>
5753 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5755 2018-05-22 Richard Biener <rguenther@suse.de>
5757 PR tree-optimization/85863
5758 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5759 comparisons when vectype is specified.
5760 (vectorizable_condition): Do not specify vectype for
5761 vect_is_simple_cond when SLP vectorizing.
5763 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5766 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5767 define __ibm128 as long double.
5768 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5769 as a distinct type when IEEE 128-bit support is enabled.
5770 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5771 128-bit types to use the correct functions.
5772 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5773 convert between 128-bit floating point types that have different
5774 modes but the same representation, instead of using gen_lowpart to
5776 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5778 (IFKF_reg): New attributes to give the register constraints for
5780 (extend<mode>tf2_internal): New insns to mark an explicit
5781 conversion between 128-bit floating point types that have a
5782 different mode but share the same representation.
5784 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5786 PR tree-optimization/85814
5787 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5788 a null return from get_strinfo when unsharing the next
5789 strinfo in the chain.
5791 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5794 * varasm.c (weak_finish): Clean up weak_decls.
5796 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5798 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5799 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5800 UNSPEC_UADALP values.
5801 * config/aarch64/iterators.md (ABAL): New int iterator.
5804 (sur): Add mappings for the above.
5805 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5807 (aarch64_<sur>abal<mode>_4): Likewise.
5808 (aarch64_<sur>adalp<mode>_3): Likewise.
5809 (<sur>sadv16qi): New define_expand.
5811 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5813 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5814 (*movdf_internal): Ditto.
5815 (*rcpsf2_sse): Ditto.
5816 (*rsqrtsf2_sse): Ditto.
5817 (*sqrt<mode>2_sse): Ditto.
5819 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5821 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5823 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5824 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5825 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5826 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5828 * config/aarch64/arm_neon.h: Likewise.
5829 * config/aarch64/iterators.md (VQ_I): New.
5831 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5833 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5835 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5838 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5840 * config/nds32/constraints.md (S): New constraint.
5841 * config/nds32/nds32.md (call_internal): Use constraint S.
5842 (call_value_internal): Likewise.
5843 (sibcall_internal): Likewise.
5844 (sibcall_value_internal): Likewise.
5846 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5847 Chung-Ju Wu <jasonwucj@gmail.com>
5849 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5852 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5853 Chung-Ju Wu <jasonwucj@gmail.com>
5855 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5856 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5857 (nds32_rtx_costs_impl): Simplify.
5858 (nds32_address_cost_impl): Simplify.
5859 (nds32_init_rtx_costs): New function.
5860 (nds32_rtx_costs_speed_prefer): Likewise.
5861 (nds32_rtx_costs_size_prefer): Likewise.
5862 (nds32_address_cost_speed_prefer): Likewise.
5863 (nds32_address_cost_speed_fwprop): Likewise.
5864 (nds32_address_cost_size_prefer): Likewise.
5865 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5866 * config/nds32/nds32.c (nds32_option_override): Use
5867 nds32_init_rtx_costs function.
5869 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5871 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5872 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5873 (TARGET_PIPELINE_N8): Likewise.
5874 (TARGET_PIPELINE_N10): Likewise.
5875 (TARGET_PIPELINE_N13): Likewise.
5876 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5878 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5880 * config/nds32/nds32-fpu.md: Update copyright year.
5882 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5884 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5886 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5888 * config/nds32/nds32.c
5889 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5890 * config/nds32/nds32.opt (minline-asm-r15): New option.
5892 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5894 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5896 * config/nds32/nds32.md (abssi2): New pattern.
5898 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5900 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5901 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5902 Merge insn pattern from sse_cvtsi2ss<round_name> and
5903 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5904 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5905 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5906 using SWI48 mode iterator.
5907 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5908 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5909 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5910 pattern from sse_cvttss2si<round_saeonly_name>
5911 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5912 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5913 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5914 using SWI48 mode iterator.
5915 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5916 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5917 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5918 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5919 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5920 using SWI48 mode iterator.
5921 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5922 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5923 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5924 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5925 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5926 SWI48 mode iterator.
5927 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5928 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5929 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5930 pattern from sse_cvttsd2si<round_saeonly_name>
5931 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5933 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5935 * config/nds32/nds32-md-auxiliary.c
5936 (nds32_valid_smw_lwm_base_p): Refine.
5937 (nds32_output_smw_single_word): Refine.
5938 (nds32_output_smw_double_word): New.
5939 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5941 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5943 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5944 (nds32_output_stack_pop): Refine.
5945 (nds32_expand_unaligned_load): Refine.
5946 (nds32_expand_unaligned_store): Refine.
5948 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5949 Chung-Ju Wu <jasonwucj@gmail.com>
5951 * config/nds32/constants.md: Add TP_REGNUM constant.
5952 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5953 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5955 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5956 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5957 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5958 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5959 and PIC code generation.
5960 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5962 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5964 * config/nds32/nds32.md: Support TLS and PIC.
5965 * config/nds32/nds32.c: Support TLS and PIC.
5966 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5967 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5970 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5972 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5973 mode with E_ prefix.
5975 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5976 Chung-Ju Wu <jasonwucj@gmail.com>
5978 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5979 * config/nds32/nds32-md-auxiliary.c
5980 (symbolic_reference_mentioned_p): New.
5981 (nds32_legitimize_ict_address): New.
5982 (nds32_expand_ict_move): New.
5983 (nds32_indirect_call_referenced_p): New.
5984 (nds32_symbol_binds_local_p): Delete.
5985 (nds32_long_call_p): Modify.
5986 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5987 * config/nds32/nds32-protos.h
5988 (symbolic_reference_mentioned_p): Declare.
5989 (nds32_legitimize_ict_address): Declare.
5990 (nds32_expand_ict_move): Declare.
5991 (nds32_indirect_call_referenced_p): Declare.
5992 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5993 (nds32_relax_group): Use nds32_ict_const_p as condition.
5994 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5995 (nds32_asm_file_start): Output ict_model directive in asm code.
5996 (nds32_legitimate_address_p): Consider indirect call.
5997 (nds32_print_operand): Consider indirect call.
5998 (nds32_print_operand_address): Consider indirect call.
5999 (nds32_insert_attributes): Handle "indirect_call" attribute.
6000 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6001 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6002 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6003 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6004 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6005 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6006 (TARGET_ICT_MODEL_SMALL): Define.
6007 (TARGET_ICT_MODEL_LARGE): Define.
6008 * config/nds32/nds32.md (movsi): Consider ict model.
6009 (call, call_value): Consider ict model.
6010 (sibcall, sibcall_value): Consider ict model.
6011 * config/nds32/nds32.opt (mict-model): New option.
6012 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6015 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6016 Monk Chiang <sh.chiang04@gmail.com>
6017 Jim Wilson <jimw@sifive.com>
6019 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6020 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6021 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6022 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6023 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6024 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6025 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6026 compute save_libcall_adjustment properly.
6027 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6028 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6029 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6030 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6031 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6032 (ABI_SPEC): Handle mabi=ilp32e.
6033 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6034 (RVE): Add RVE mask.
6035 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6036 <-march>: Add rv32e as an example.
6038 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6041 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6042 (intra_create_variable_infos): Handle C++ constructors.
6044 2018-05-18 Martin Liska <mliska@suse.cz>
6046 * passes.def: Remove a redundant pass.
6048 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6051 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6053 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6055 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6057 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6059 (armv2, armv2a, armv3, armv3m): Delete architectures.
6060 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6061 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6062 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6064 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6065 (*mulsidi3adddi): Likewise.
6066 (mulsidi3): Likewise.
6067 (*mulsidi3_nov6): Likewise.
6068 (umulsidi3): Likewise.
6069 (umulsidi3_nov6): Likewise.
6070 (umaddsidi4): Likewise.
6071 (*umulsidi3adddi): Likewise.
6072 (smulsi3_highpart): Likewise.
6073 (*smulsi3_highpart_nov6): Likewise.
6074 (umulsi3_highpart): Likewise.
6075 (*umulsi3_highpart_nov6): Likewise.
6076 * config/arm/arm.h (arm_arch3m): Delete.
6077 * config/arm/arm.c (arm_arch3m): Delete.
6078 (arm_option_override_internal): Update armv3-related comment.
6079 (arm_configure_build_target): Delete use of isa_bit_mode32.
6080 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6081 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6082 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6084 (mulusa3): Likewise.
6085 * config/arm/arm-protos.h (arm_arch3m): Delete.
6086 * config/arm/arm-tables.opt: Regenerate.
6087 * config/arm/arm-tune.md: Likewise.
6088 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6089 deleted architectures.
6091 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6093 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6094 (armv5t, armv5te): New features.
6095 (ARMv5, ARMv5e): Delete fgroups.
6096 (ARMv5t, ARMv5te): Adjust for above changes.
6098 (armv5, armv5e): Delete arches.
6099 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6101 (*call_reg_arm): Likewise.
6102 (*call_value_reg_armv5): Likewise.
6103 (*call_value_reg_arm): Likewise.
6104 (*call_symbol): Likewise.
6105 (*call_value_symbol): Likewise.
6106 (*sibcall_insn): Likewise.
6107 (*sibcall_value_insn): Likewise.
6109 (prefetch): Likewise.
6110 (define_split and define_peephole2 dependent on arm_arch5):
6112 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6114 (TARGET_ARM_QBIT): Likewise.
6115 (TARGET_DSP_MULTIPLY): Likewise.
6116 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6117 (arm_arch5, arm_arch5e): Delete.
6118 (arm_arch5t, arm_arch5te): Declare.
6119 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6120 (arm_arch5t): Declare.
6121 (arm_option_reconfigure_globals): Update for the above.
6122 (arm_options_perform_arch_sanity_checks): Update comment, replace
6123 use of arm_arch5 with arm_arch5t.
6124 (use_return_insn): Likewise.
6125 (arm_emit_call_insn): Likewise.
6126 (output_return_instruction): Likewise.
6127 (arm_final_prescan_insn): Likewise.
6128 (arm_coproc_builtin_available): Likewise.
6129 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6130 arm_arch5e with arm_arch5t and arm_arch5te.
6131 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6132 (arm_arch5t, arm_arch5te): Declare.
6133 * config/arm/arm-tables.opt: Regenerate.
6134 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6135 * config/arm/t-multilib: Likewise.
6136 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6137 instead of arm_arch5.
6138 (*call_reg_thumb1): Likewise.
6139 (*call_value_reg_thumb1_v5): Likewise.
6140 (*call_value_reg_thumb1): Likewise.
6141 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6143 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6145 2018-05-18 Martin Liska <mliska@suse.cz>
6147 PR gcov-profile/84846
6148 * doc/gcov.texi: Document -t option of gcov tool.
6150 2018-05-18 Martin Liska <mliska@suse.cz>
6152 PR gcov-profile/84846
6153 * gcov.c (print_usage): Add new -t option.
6154 (process_args): Handle the option.
6155 (generate_results): Use stdout as output when requested by
6158 2018-05-18 Martin Liska <mliska@suse.cz>
6160 PR gcov-profile/84846
6161 * coverage.c (coverage_init): Write PWD to .gcno file.
6162 * doc/gcov.texi: Document how working directory is printed.
6163 * gcov-dump.c (dump_gcov_file): Print PWD.
6164 * gcov.c (output_intermediate_file): Likewise.
6165 (read_graph_file): Read PWD string.
6166 (output_lines): Print PWD.
6168 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6171 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6172 for retval and return false if all args to phi are zero.
6174 2018-05-18 Richard Biener <rguenther@suse.de>
6176 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6178 (evrp_dom_walker::before_dom_children): Call it.
6180 2018-05-18 Richard Biener <rguenther@suse.de>
6182 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6183 results when processing array refs with variable index.
6185 2018-05-18 Toon Moene <toon@moene.org>
6187 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6188 directly after that of -floop-interchange. Indicate that both
6189 options are enabled by default when specifying -O3.
6191 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6193 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6194 iterator. Delete separate integer-mode vec_set<mode> expander.
6195 (aarch64_simd_vec_setv2di): Delete.
6196 (vec_setv2di): Delete.
6197 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6198 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6199 the "w, r" alternative.
6201 2018-05-18 Martin Liska <mliska@suse.cz>
6203 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6204 * tree-pass.h (make_pass_lower_switch_O0): New function.
6205 * tree-switch-conversion.c (node_has_low_bound): Remove.
6206 (node_has_high_bound): Likewise.
6207 (node_is_bounded): Likewise.
6208 (class pass_lower_switch): Make it a template type and create
6210 (pass_lower_switch::execute): Add template argument.
6211 (make_pass_lower_switch): New function.
6212 (make_pass_lower_switch_O0): New function.
6213 (do_jump_if_equal): Remove.
6214 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6215 all the hard work to tree optimization passes.
6217 2018-05-18 Martin Liska <mliska@suse.cz>
6219 * dbgcnt.c (limit_low): Renamed from limit.
6220 (limit_high): New variable.
6221 (dbg_cnt_is_enabled): Check for upper limit.
6222 (dbg_cnt): Adjust dumping.
6223 (dbg_cnt_set_limit_by_index): Add new argument for high
6225 (dbg_cnt_set_limit_by_name): Likewise.
6226 (dbg_cnt_process_single_pair): Parse new format.
6227 (dbg_cnt_process_opt): Use strtok.
6228 (dbg_cnt_list_all_counters): Remove 'value' and add
6230 * doc/invoke.texi: Document changes.
6232 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6234 * doc/sourcebuild.texi (scalar_all_fma): Document.
6235 * tree.def (FMA_EXPR): Delete.
6236 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6237 * internal-fn.c (ternary_direct): New macro.
6238 (expand_ternary_optab_fn): Likewise.
6239 (direct_ternary_optab_supported_p): Likewise.
6240 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6241 * builtins.c (fold_builtin_fma): Delete.
6242 (fold_builtin_3): Don't call it.
6243 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6244 * expr.c (expand_expr_real_2): Likewise.
6245 * fold-const.c (operand_equal_p): Likewise.
6246 (fold_ternary_loc): Likewise.
6247 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6248 * gimple.c (DEFTREECODE): Likewise.
6249 * gimplify.c (gimplify_expr): Likewise.
6250 * optabs-tree.c (optab_for_tree_code): Likewise.
6251 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6252 * tree-eh.c (operation_could_trap_p): Likewise.
6253 (stmt_could_throw_1_p): Likewise.
6254 * tree-inline.c (estimate_operator_cost): Likewise.
6255 * tree-pretty-print.c (dump_generic_node): Likewise.
6256 (op_code_prio): Likewise.
6257 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6258 * tree-ssa-operands.c (get_expr_operands): Likewise.
6259 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6260 * fold-const-call.h (fold_fma): Delete.
6261 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6262 CFN_FNMA and CFN_FNMS.
6264 * genmatch.c (combined_fn): New enum.
6265 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6266 (commutative_op): New function.
6267 (commutate): Use it. Handle more than 2 operands.
6268 (dt_operand::gen_gimple_expr): Use commutative_op.
6269 (parser::parse_expr): Allow :c to be used with non-binary
6270 operators if the commutative operand is known.
6271 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6272 CFN_FMS, CFN_FNMA and CFN_FNMS.
6273 (backprop::process_assign_use): Remove FMA_EXPR handling.
6274 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6275 (gen_hsa_fma): New function.
6276 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6277 IFN_FNMA and IFN_FNMS.
6278 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6279 * gimple-fold.h (follow_all_ssa_edges): Declare.
6280 * gimple-fold.c (follow_all_ssa_edges): New function.
6281 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6282 gimple_build interface and use follow_all_ssa_edges to fold the result.
6283 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6284 instead of checking for optabs directly.
6285 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6286 rather than FMA_EXPRs.
6287 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6288 call to IFN_FMA instead of an FMA_EXPR.
6290 2018-05-17 Jim Wilson <jimw@sifive.com>
6292 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6293 sign extended promoted subreg, and the range does not have the sign bit
6294 set, then do a sign extend.
6296 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6297 test, check for sign extended subreg and/or constant operands, and
6298 do a sign extend in that case.
6300 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6302 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6303 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6305 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6306 Change logics_shift_reg to logics_shift_imm.
6307 (thunderx2t99_fp_loadpair_basic): Delete.
6308 (thunderx2t99_fp_storepair_basic): Delete.
6309 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6310 (thunderx2t99_asimd_polynomial): Delete.
6311 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6312 and neon_fp_mul_d_scalar_q.
6313 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6314 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6315 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6316 (thunderx2t99_asimd_lut): Add missing tbl types.
6317 (thunderx2t99_asimd_ext): Delete.
6318 (thunderx2t99_asimd_load1_1_mult): Delete.
6319 (thunderx2t99_asimd_load1_2_mult): Delete.
6320 (thunderx2t99_asimd_load1_ldp): New.
6321 (thunderx2t99_asimd_load1): New.
6322 (thunderx2t99_asimd_load2): Add missing *load2* types.
6323 (thunderx2t99_asimd_load3): New.
6324 (thunderx2t99_asimd_load4): New.
6325 (thunderx2t99_asimd_store1_1_mult): Delete.
6326 (thunderx2t99_asimd_store1_2_mult): Delete.
6327 (thunderx2t99_asimd_store2_mult): Delete.
6328 (thunderx2t99_asimd_store2_onelane): Delete.
6329 (thunderx2t99_asimd_store_stp): New.
6330 (thunderx2t99_asimd_store1): New.
6331 (thunderx2t99_asimd_store2): New.
6332 (thunderx2t99_asimd_store3): New.
6333 (thunderx2t99_asimd_store4): New.
6335 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6337 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6338 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6340 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6341 Segher Boessenkool <segher@kernel.crashing.org>
6344 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6347 2018-05-17 Richard Biener <rguenther@suse.de>
6349 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6350 for pruning loop and prune defs feeding only already visited PHIs.
6352 2018-05-17 Richard Biener <rguenther@suse.de>
6354 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6356 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6357 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/85793
6360 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6361 for VMAT_ELEMENTWISE.
6363 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6365 * internal-fn.h (lookup_internal_fn): Declare
6366 * internal-fn.c (lookup_internal_fn): New function.
6367 * gimple.c (gimple_build_call_from_tree): Handle calls to
6369 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6370 internal function names.
6371 * tree-pretty-print.c (dump_generic_node): Likewise.
6372 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6374 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6376 * gimple-fold.h (gimple_build): Make the function forms take
6377 combined_fn rather than built_in_function.
6378 (gimple_simplify): Likewise.
6379 * gimple-match-head.c (gimple_simplify): Likewise.
6380 * gimple-fold.c (gimple_build): Likewise.
6381 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6382 rather than gimple_build_call_internal.
6383 (get_initial_defs_for_reduction): Likewise.
6384 (vect_create_epilog_for_reduction): Likewise.
6385 (vectorizable_live_operation): Likewise.
6387 2018-05-17 Martin Liska <mliska@suse.cz>
6389 * gimple-ssa-sprintf.c (format_directive): Do not use
6390 space in between 'G_' and '('.
6392 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6395 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6396 even if the mask is not all ones.
6399 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6401 (ix86_gimple_fold_builtin): Likewise.
6404 * config/i386/i386.c: Include tree-vector-builder.h.
6405 (ix86_vector_shift_count): New function.
6406 (ix86_fold_builtin): Fold shift builtins by scalar count.
6407 (ix86_gimple_fold_builtin): Likewise.
6409 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6410 _mm512_setzero): New intrinsics.
6412 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6413 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6415 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6416 code generation for cases where splatting a value is not useful.
6417 * simplify-rtx.c (simplify_ternary_operation): Simplify
6418 vec_merge across a vec_duplicate and a paradoxical subreg forming
6419 a vector mode to a vec_concat.
6421 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6423 * config.gcc: Support "goldmont-plus".
6424 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6426 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6427 PROCESSOR_GOLDMONT_PLUS.
6428 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6429 (processor_target_table): Add "goldmont-plus".
6430 (PTA_GOLDMONT_PLUS): Define.
6431 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6432 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6433 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6434 (fold_builtin_cpu): Add "goldmont-plus".
6435 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6436 (ix86_option_override_internal): Add "goldmont-plus".
6437 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6438 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6439 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6440 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6442 2018-05-17 Richard Biener <rguenther@suse.de>
6444 PR tree-optimization/85757
6445 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6446 remove defs that only feed that PHI from further processing.
6448 2018-05-16 Jim Wilson <jimw@sifive.com>
6450 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6452 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6454 2018-05-16 Mark Wielaard <mark@klomp.org>
6456 * dwarf2out.c (count_index_strings): New function.
6457 (output_indirect_strings): Call count_index_strings and generate
6458 header for dwarf_version >= 5.
6460 2018-05-16 Mark Wielaard <mark@klomp.org>
6462 * dwarf2out.c (dwarf_FORM): New function.
6463 (set_indirect_string): Use dwarf_FORM.
6464 (reset_indirect_string): Likewise.
6465 (size_of_die): Likewise.
6466 (value_format): Likewise.
6467 (output_die): Likewise.
6468 (add_skeleton_AT_string): Likewise.
6469 (output_macinfo_op): Likewise.
6470 (index_string): Likewise.
6471 (output_index_string_offset): Likewise.
6472 (output_index_string): Likewise.
6473 (count_index_strings): Likewise.
6475 2018-05-16 Carl Love <cel@us.ibm.com>
6477 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6478 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6480 2018-05-16 Martin Jambor <mjambor@suse.cz>
6482 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6483 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6485 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6487 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6488 (fnma<mode>4): Likewise.
6489 (fms<mode>4): Likewise.
6490 (fnms<mode>4): Likewise.
6491 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6492 (aarch64_fnma<mode>4): Likewise.
6493 (aarch64_fms<mode>4): Likewise.
6494 (aarch64_fnms<mode>4): Likewise.
6495 (aarch64_fnmadd<mode>4): Likewise.
6497 2018-05-16 Jason Merrill <jason@redhat.com>
6499 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6501 2018-05-16 Richard Biener <rguenther@suse.de>
6503 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6504 (dump_stmt_cost): Declare.
6505 (add_stmt_cost): Dump cost we add.
6506 (add_stmt_costs): New function.
6507 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6509 (vect_analyze_stmt): Adjust prototype.
6510 (vectorizable_condition): Likewise.
6511 (vectorizable_live_operation): Likewise.
6512 (vectorizable_reduction): Likewise.
6513 (vectorizable_induction): Likewise.
6514 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6515 cost vector to pass to vectorizable_ and record afterwards.
6516 (vect_model_reduction_cost): Take cost vector argument and adjust.
6517 (vect_model_induction_cost): Likewise.
6518 (vectorizable_reduction): Likewise.
6519 (vectorizable_induction): Likewise.
6520 (vectorizable_live_operation): Likewise.
6521 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6522 SLP_TREE_NUMBER_OF_VEC_STMTS.
6523 (vect_analyze_slp_cost_1): Remove.
6524 (vect_analyze_slp_cost): Likewise.
6525 (vect_slp_analyze_node_operations): Take visited args and
6526 a target cost vector. Avoid processing already visited stmt sets.
6527 (vect_slp_analyze_operations): Use a local cost vector to gather
6528 costs and register those of non-discarded instances.
6529 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6530 (vect_schedule_slp_instance): Remove copying of
6531 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6533 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6534 adding cost. Record cost entry location.
6535 (vect_prologue_cost_for_slp_op): Function to compute cost of
6536 a constant or invariant generated for SLP vect in the prologue,
6537 split out from vect_analyze_slp_cost_1.
6538 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6539 (vect_model_promotion_demotion_cost): Likewise.
6540 (vect_model_store_cost): Likewise, make static.
6541 (vect_model_load_cost): Likewise.
6542 (vectorizable_bswap): Add cost vector arg and adjust.
6543 (vectorizable_call): Likewise.
6544 (vectorizable_simd_clone_call): Likewise.
6545 (vectorizable_conversion): Likewise.
6546 (vectorizable_assignment): Likewise.
6547 (vectorizable_shift): Likewise.
6548 (vectorizable_operation): Likewise.
6549 (vectorizable_store): Likewise.
6550 (vectorizable_load): Likewise.
6551 (vectorizable_condition): Likewise.
6552 (vectorizable_comparison): Likewise.
6553 (can_vectorize_live_stmts): Likewise.
6554 (vect_analyze_stmt): Likewise.
6555 (vect_transform_stmt): Adjust calls to vectorizable_*.
6556 * tree-vectorizer.c: Include gimple-pretty-print.h.
6557 (dump_stmt_cost): New function.
6559 2018-05-16 Richard Biener <rguenther@suse.de>
6561 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6562 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6563 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6564 (check_name): New callback.
6565 (dse_classify_store): Track cycles via a visited bitmap of PHI
6566 defs and simplify handling of in-loop and across loop dead stores
6567 and properly fail for loop-variant refs. Handle byte-tracking with
6568 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6571 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6573 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6574 (vect_get_mask_type_for_stmt): Likewise.
6575 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6577 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6578 to determine the statement's vector type and the vector type that
6579 should be used for calculating nunits. Deal with cases in which
6580 the type has to be deferred.
6581 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6582 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6583 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6584 (vect_determine_vf_for_stmt): New functions, split out from...
6585 (vect_determine_vectorization_factor): ...here.
6586 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6587 (vect_get_mask_type_for_stmt): New functions, split out from
6588 vect_determine_vectorization_factor.
6590 2018-05-16 Richard Biener <rguenther@suse.de>
6592 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6593 verify the [VEC_]COND_EXPR embedded comparison.
6595 2018-05-15 Martin Sebor <msebor@redhat.com>
6597 PR tree-optimization/85753
6598 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6599 RECORD_TYPE in addition to ARRAY_TYPE.
6601 2018-05-15 Martin Sebor <msebor@redhat.com>
6604 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6606 2018-05-15 Richard Biener <rguenther@suse.de>
6608 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6609 add by_clobber_p one. Change algorithm to collect all defs
6610 representing uses we need to walk and try reducing them to
6611 a single one before failing.
6612 (dse_dom_walker::dse_optimize_stmt): Adjust.
6614 2018-05-13 Mark Wielaard <mark@klomp.org>
6616 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6617 (size_of_loc_descr): Likewise.
6618 (output_loc_operands): Likewise.
6619 (output_loc_operands_raw): Likewise.
6620 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6621 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6622 (hash_loc_operands): Likewise.
6623 (compare_loc_operands): Likewise.
6625 2018-05-14 Mark Wielaard <mark@klomp.org>
6627 * dwarf2out.c (count_index_addrs): New function.
6628 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6630 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6632 PR tree-optimization/83648
6633 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6634 return value as malloc candidate.
6636 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6639 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6640 param as true in call to suggest_attribute.
6642 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6644 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6645 -mreadonly-in-sdata.
6647 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6649 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6651 (aarch64_crypto_aesd_fused): Likewise.
6653 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6655 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6656 (movsi_aarch64): Likewise.
6657 (load_pairsi): Likewise.
6658 (load_pairdi): Likewise.
6659 (store_pairsi): Likewise.
6660 (store_pairdi): Likewise.
6661 (load_pairsf): Likewise.
6662 (load_pairdf): Likewise.
6663 (store_pairsf): Likewise.
6664 (store_pairdf): Likewise.
6665 (zero_extend): Likewise.
6666 (trunc): Swap alternatives.
6667 (fcvt_target): Add '?' to prefer w over r.
6669 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6672 * config/i386/i386.md: Disallow non-commutative arithmetics in
6673 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6674 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6675 in the peephole2 before it.
6677 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6679 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6680 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6681 (ix86_handle_option): Handle -mcldemote.
6682 * config.gcc: New header.
6683 * config/i386/cldemoteintrin.h: New file.
6684 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6685 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6687 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6688 OPTION_MASK_ISA_CLDEMOTE.
6689 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6690 (ix86_valid_target_attribute_inner_p): Ditto.
6691 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6692 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6693 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6694 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6695 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6697 * config/i386/i386.opt: Add -mcldemote.
6698 * config/i386/x86intrin.h: New header.
6699 * doc/invoke.texi: Add -mcldemote.
6701 2018-05-14 Richard Biener <rguenther@suse.de>
6703 * doc/match-and-simplify.texi: Adjust :s documentation.
6705 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6707 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6708 intended memcpy size.
6709 (REORDER_45): Likewise.
6711 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6713 * sort.cc: New file.
6714 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6715 * vec.c (qsort_chk): Use gcc_qsort.
6716 * Makefile.in (OBJS-libcommon): Add sort.o.
6717 (build/sort.o): New target. Use it...
6718 (BUILD_RTL): ... here, and...
6719 (build/gencfn-macros): ... here, and...
6720 (build/genmatch): ... here.
6722 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6723 Chung-Ju Wu <jasonwucj@gmail.com>
6725 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6726 * config/nds32/nds32-graywolf.md: New file.
6727 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6728 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6730 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6731 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6732 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6733 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6734 * config/nds32/pipelines.md: Include n15 settings.
6736 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6737 Chung-Ju Wu <jasonwucj@gmail.com>
6739 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6740 * config/nds32/nds32-n13.md: New file.
6741 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6742 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6744 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6745 * config/nds32/nds32.md (pipeline_model): Add n13.
6746 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6747 * config/nds32/pipelines.md: Include n13 settings.
6749 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6750 Chung-Ju Wu <jasonwucj@gmail.com>
6752 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6753 * config/nds32/nds32-n10.md: New file.
6754 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6755 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6757 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6758 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6759 * config/nds32/nds32.md (pipeline_model): Add n10.
6760 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6761 * config/nds32/pipelines.md: Include n10 settings.
6763 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6764 Kito Cheng <kito.cheng@gmail.com>
6765 Chung-Ju Wu <jasonwucj@gmail.com>
6767 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6768 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6769 Add enum values for DSP extension instructions.
6770 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6772 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6773 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6775 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6776 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6777 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6778 * config/nds32/nds32-intrinsic.md: Likewise.
6779 * config/nds32/nds32_intrinsic.h: Likewise.
6780 * config/nds32/nds32-md-auxiliary.c: Likewise.
6781 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6782 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6783 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6784 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6785 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6786 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6787 TYPE_DMAC in switch statement.
6788 * config/nds32/nds32.c: New checking and implementation for DSP
6789 extension instructions.
6790 * config/nds32/nds32.h: Likewise.
6791 * config/nds32/nds32.md: Likewise.
6792 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6793 * config/nds32/predicates.md: Implement new predicates for DSP
6796 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6798 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6799 Reformat alternatives and attributes so it is easier to identify
6800 which constraints/attributes go with which instruction.
6801 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6802 (mov<mode>_softfloat32, FMOVE64): Likewise.
6803 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6804 (mov<mode>_softfloat64, FMOVE64): Likewise.
6806 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6808 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6810 (Basic PowerPC Built-in Functions): The new name of the
6811 subsection previously known as "PowerPC Built-in Functions".
6812 (Basic PowerPC Built-in Functions Available on all Configurations):
6814 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6815 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6816 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6817 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6819 2018-05-11 Martin Jambor <mjambor@suse.cz>
6822 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6825 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6828 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6830 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6832 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6833 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6834 (ix86_handle_option): Handle -mwaitpkg.
6835 * config.gcc: New header.
6836 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6837 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6838 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6840 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6841 OPTION_MASK_ISA_WAITPKG.
6842 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6843 (ix86_option_override_internal): Add PTA_WAITPKG.
6844 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6845 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6846 IX86_BUILTIN_TPAUSE.
6847 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6848 __builtin_ia32_umwait and __builtin_ia32_tpause.
6849 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6850 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6851 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6852 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6853 UNSPECV_TPAUSE): New.
6854 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6855 * config/i386/i386.opt: Add -mwaitpkg.
6856 * config/i386/waitpkgintrin.h: New file.
6857 * config/i386/x86intrin.h: New header.
6858 * doc/invoke.texi: Add -mwaitpkg.
6860 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6863 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6865 (cortex-m0): Use armv6s-m isa.
6866 (cortex-m0plus): Likewise.
6867 (cortex-m1): Likewise.
6868 (cortex-m0.small-multiply): Likewise.
6869 (cortex-m0plus.small-multiply): Likewise.
6870 (cortex-m1.small-multiply): Likewise.
6872 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6873 Jakub Jelinek <jakub@redhat.com>
6875 PR tree-optimization/85692
6876 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6877 source permute as well.
6879 2018-05-11 Martin Liska <mliska@suse.cz>
6882 * doc/extend.texi: Document LLVM style format for no_sanitize
6885 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6887 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6888 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6889 (mode_supports_vsx_dform_quad): Likewise.
6890 (mode_supports_vmx_dform): Move these functions to be next to the
6891 other mode_supports functions.
6892 (mode_supports_dq_form): Likewise.
6893 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6894 mode_supports_dq_form.
6895 (reg_offset_addressing_ok_p): Likewise.
6896 (offsettable_ok_by_alignment): Likewise.
6897 (rs6000_legitimate_offset_address_p): Likewise.
6898 (legitimate_lo_sum_address_p): Likewise.
6899 (rs6000_legitimize_address): Likewise.
6900 (rs6000_legitimize_reload_address): Likewise.
6901 (rs6000_secondary_reload_inner): Likewise.
6902 (rs6000_preferred_reload_class): Likewise.
6903 (rs6000_output_move_128bit): Likewise.
6905 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6907 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6908 Generate SImode target register for null target.
6909 <case IX86_BUILTIN_XGETBV>: Ditto.
6910 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6911 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6913 2018-05-10 Carl Love <cel@us.ibm.com>
6915 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6916 dcbtt and dcbtstt if operands[2] is 0.
6918 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6921 * config/i386/sse.md (usadv64qi): New expander.
6923 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6925 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6926 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6927 -maltivec=be support.
6928 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6929 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6930 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6931 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6932 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6933 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6934 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6935 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6936 altivec_vsumsws): Adjust.
6937 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6938 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6939 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6941 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6942 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6943 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6944 (altivec_lve<VI_char>x): Delete expand.
6945 (*altivec_lve<VI_char>x_internal): Rename to...
6946 (altivec_lve<VI_char>x): ... this.
6947 (altivec_lvxl_<mode>): Delete expand.
6948 (*altivec_lvxl_<mode>_internal): Rename to ...
6949 (altivec_lvxl_<mode>): ... this.
6950 (altivec_stvxl_<mode>): Delete expand.
6951 (*altivec_stvxl_<mode>_internal): Rename to ...
6952 (altivec_stvxl_<mode>): ... this.
6953 (altivec_stve<VI_char>x): Delete expand.
6954 (*altivec_stve<VI_char>x_internal): Rename to ...
6955 (altivec_stve<VI_char>x): ... this.
6956 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6957 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6958 reduc_plus_scal_<mode>): Adjust.
6959 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6961 (rs6000_cpu_cpp_builtins): Adjust.
6962 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6963 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6964 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6965 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6966 -maltivec=be support.
6967 (rs6000_split_vec_extract_var): Adjust.
6968 (rs6000_split_v4si_init): Adjust.
6969 (swap_selector_for_mode): Delete.
6970 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6971 altivec_expand_stvex_be): Delete.
6972 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6973 -maltivec=be support.
6974 (rs6000_gimple_fold_builtin): Ditto.
6975 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6977 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6978 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6979 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6980 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6981 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6982 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6983 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6984 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6985 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6986 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6987 anonymous split): Adjust.
6988 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6989 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6991 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6993 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6994 when --with-gxx-include-dir is also specified.
6995 * configure: Regenerate.
6997 2018-05-09 Jim Wilson <jimw@sifive.com>
7000 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7001 * config/riscv/t-withmultilib: New.
7002 * config/riscv/withmultilib.h: New.
7003 * doc/install.texi: Document RISC-V --with-multilib-list support.
7005 2018-05-09 Richard Biener <rguenther@suse.de>
7007 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7009 (vect_bb_vectorization_profitable_p): Adjust. Compute
7010 actual scalar cost using the cost vector and the add_stmt_cost
7013 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7015 PR rtl-optimization/85645
7016 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7017 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7019 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7021 PR rtl-optimization/85645
7022 * shrink-wrap.c (spread_components): Return a boolean saying if
7023 anything was changed.
7024 (try_shrink_wrapping_separate): Iterate spread_components until
7025 nothing changes anymore.
7027 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7029 PR rtl-optimization/85645
7030 * regrename.c (build_def_use): Also kill the chains that include the
7031 destination of a REG_CFA_REGISTER note.
7033 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7035 PR rtl-optimization/85645
7036 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7037 insn that has a REG_CFA_REGISTER note.
7039 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7041 * cfgexpand.c (expand_clobber): New function.
7042 (expand_gimple_stmt_1): Use it.
7043 * tree-vect-stmts.c (vect_clobber_variable): New function,
7045 (vectorizable_simd_clone_call): ...here.
7046 (vectorizable_store): Emit a clobber either side of an
7047 IFN_STORE_LANES sequence.
7048 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7050 2018-05-09 Tom de Vries <tom@codesourcery.com>
7053 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7054 (define_insn "trap_if_false"): Add exit after trap.
7056 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7058 PR rtl-optimization/85638
7059 * bb-reorder.c: Include common/common-target.h.
7060 (create_forwarder_block): New function extracted from...
7061 (fix_up_crossing_landing_pad): ...here. Rename into...
7062 (dw2_fix_up_crossing_landing_pad): ...this.
7063 (sjlj_fix_up_crossing_landing_pad): New function.
7064 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7065 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7066 from both partitions and exit the loop after one iteration.
7068 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7071 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7073 (Basic PowerPC Built-in Functions): The new name of the
7074 subsection previously known as "PowerPC Built-in Functions".
7075 (Basic PowerPC Built-in Functions Available on all Configurations):
7077 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7079 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7081 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7083 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7086 2018-05-08 Jim Wilson <jimw@sifive.com>
7088 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7089 (LD_EMUL_SUFFIX): New.
7090 (LINK_SPEC): Use it.
7092 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7094 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7096 (Basic PowerPC Built-in Functions): The new name of the
7097 subsection previously known as "PowerPC Built-in Functions".
7098 (Basic PowerPC Built-in Functions Available on all Configurations):
7100 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7102 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7104 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7106 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7109 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7112 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7113 after cmpelim optimization.
7115 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7117 * config.gcc: Support "goldmont".
7118 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7119 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7121 * config/i386/i386.c (m_GOLDMONT): Define.
7122 (processor_target_table): Add "goldmont".
7123 (PTA_GOLDMONT): Define.
7124 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7125 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7126 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7127 (fold_builtin_cpu): Add "goldmont".
7128 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7129 (ix86_option_override_internal): Add "goldmont".
7130 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7131 (processor_type): Add PROCESSOR_GOLDMONT.
7132 * config/i386/i386.md: Add CPU "glm".
7133 * config/i386/glm.md: New file.
7134 * config/i386/x86-tune.def: Add m_GOLDMONT.
7135 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7137 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7140 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7142 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7143 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7144 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7147 * config/i386/i386.c (ix86_fold_builtin): Handle
7148 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7151 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7152 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7154 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7157 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7158 (check_arch): Likewise.
7159 (check_fpu): Return the result rather than printing it.
7160 (end arch): Fix operator precedence.
7161 (end cpu): Likewise.
7162 (END): Print the result from check_fpu.
7164 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7165 Alan Hayward <alan.hayward@arm.com>
7166 David Sherwood <david.sherwood@arm.com>
7168 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7169 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7170 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7171 (*fcmuo<mode>_and): New patterns.
7173 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7175 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7176 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7177 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7178 (cmp_op, sve_imm_con): New code attributes.
7179 (SVE_COND_INT_CMP, imm_con): Delete.
7180 (cmp_op): Remove above unspecs from int attribute.
7181 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7183 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7184 comparison-specific unspecs.
7185 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7186 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7187 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7188 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7189 (*vec_fcm<cmp_op><mode>): Rename to...
7190 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7191 (*vec_fcmuo<mode>): Rename to...
7192 (*fcmuo<mode>): ...this and adjust likewise.
7193 (*pred_fcm<cmp_op><mode>): New pattern.
7194 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7195 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7197 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7199 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7200 (aarch64_emit_sve_predicated_cond): New function.
7201 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7202 (aarch64_emit_unspec_cond_or): Replace with...
7203 (aarch64_emit_sve_or_conds): ...this new function. Use
7204 aarch64_emit_sve_ptrue_op for the individual comparisons and
7205 aarch64_emit_binop to OR them together.
7206 (aarch64_emit_inverted_unspec_cond): Replace with...
7207 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7208 aarch64_emit_sve_ptrue_op for the comparison and
7209 aarch64_emit_unop to invert the result.
7210 (aarch64_expand_sve_vec_cmp_float): Update after the above
7211 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7213 2018-05-07 Nathan Sidwell <nathan@acm.org>
7215 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7216 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7217 (Backwards Compatibility): Likewise.
7219 2018-05-07 Luis Machado <luis.machado@linaro.org>
7223 2018-05-07 Luis Machado <luis.machado@linaro.org>
7225 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7226 <prefetch_dynamic_strides>: New const bool field.
7227 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7228 prefetch_dynamic_strides.
7229 (exynosm1_prefetch_tune): Likewise.
7230 (thunderxt88_prefetch_tune): Likewise.
7231 (thunderx_prefetch_tune): Likewise.
7232 (thunderx2t99_prefetch_tune): Likewise.
7233 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7235 (aarch64_override_options_internal): Update to set
7236 PARAM_PREFETCH_DYNAMIC_STRIDES.
7237 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7238 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7239 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7240 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7241 prefetch-dynamic-strides setting.
7243 2018-05-07 Luis Machado <luis.machado@linaro.org>
7245 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7246 <minimum_stride>: New const int field.
7247 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7248 minimum_stride field.
7249 (exynosm1_prefetch_tune): Likewise.
7250 (thunderxt88_prefetch_tune): Likewise.
7251 (thunderx_prefetch_tune): Likewise.
7252 (thunderx2t99_prefetch_tune): Likewise.
7253 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7254 (aarch64_override_options_internal): Update to set
7255 PARAM_PREFETCH_MINIMUM_STRIDE.
7256 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7257 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7258 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7259 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7260 stride is constant and is below the minimum stride threshold.
7262 2018-05-07 Luis Machado <luis.machado@linaro.org>
7264 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7267 2018-05-07 Luis Machado <luis.machado@linaro.org>
7269 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7270 <prefetch_dynamic_strides>: New const bool field.
7271 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7272 prefetch_dynamic_strides.
7273 (exynosm1_prefetch_tune): Likewise.
7274 (thunderxt88_prefetch_tune): Likewise.
7275 (thunderx_prefetch_tune): Likewise.
7276 (thunderx2t99_prefetch_tune): Likewise.
7277 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7279 (aarch64_override_options_internal): Update to set
7280 PARAM_PREFETCH_DYNAMIC_STRIDES.
7281 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7282 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7283 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7284 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7285 prefetch-dynamic-strides setting.
7287 2018-05-07 Luis Machado <luis.machado@linaro.org>
7289 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7290 <minimum_stride>: New const int field.
7291 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7292 minimum_stride field.
7293 (exynosm1_prefetch_tune): Likewise.
7294 (thunderxt88_prefetch_tune): Likewise.
7295 (thunderx_prefetch_tune): Likewise.
7296 (thunderx2t99_prefetch_tune): Likewise.
7297 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7298 (aarch64_override_options_internal): Update to set
7299 PARAM_PREFETCH_MINIMUM_STRIDE.
7300 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7301 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7302 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7303 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7304 stride is constant and is below the minimum stride threshold.
7306 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7309 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7310 the type is addressable. Don't force op into register if it has
7313 2018-05-05 Roland McGrath <mcgrathr@google.com>
7316 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7317 any section for which we don't know a specific type it should have,
7318 regardless of name. Previously this was done only for the exact
7319 names ".init_array", ".fini_array", and ".preinit_array".
7320 (default_elf_asm_named_section): Add comment about
7321 relationship with default_section_type_flags and SECTION_NOTYPE.
7322 (get_section): Don't consider it a type conflict if one side has
7323 SECTION_NOTYPE and the other doesn't, as long as neither has the
7324 SECTION_BSS et al used in the default_section_type_flags logic.
7326 2018-05-05 Tom de Vries <tom@codesourcery.com>
7329 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7330 (workaround_barsyncs): New function.
7331 (nvptx_reorg): Use workaround_barsyncs.
7332 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7333 (define_expand "nvptx_membar_cta"): New define_expand.
7334 (define_insn "*nvptx_membar_cta"): New insn.
7336 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7338 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7339 To improve optimization opportunities.
7340 * builtin-types.def: The new needed builtin types for the above.
7342 2018-05-04 Richard Biener <rguenther@suse.de>
7344 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7345 * gimple-ssa-store-merging.c
7346 (imm_store_chain_info::output_merged_store): Remove redundant create,
7347 release split_store vector contents on failure.
7348 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7349 scalar stmt vector on cache hit.
7351 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7353 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7355 * config.gcc (powerpc-xilinx-eabi*): Remove.
7356 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7358 (fusion_addis_mem_combo_load): Ditto.
7359 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7361 (rs6000_cpu_cpp_builtins): Ditto.
7362 * config/rs6000/rs6000-linux.c
7363 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7364 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7365 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7367 (rs6000_setup_reg_addr_masks): Ditto.
7368 (rs6000_init_hard_regno_mode_ok): Ditto.
7369 (rs6000_option_override_internal): Ditto.
7370 (legitimate_lo_sum_address_p): Ditto.
7371 (rs6000_legitimize_address): Ditto.
7372 (rs6000_legitimize_reload_address): Ditto.
7373 (rs6000_legitimate_address_p): Ditto.
7374 (abi_v4_pass_in_fpr): Ditto.
7375 (setup_incoming_varargs): Ditto.
7376 (rs6000_gimplify_va_arg): Ditto.
7377 (rs6000_split_multireg_move): Ditto.
7378 (rs6000_savres_strategy): Ditto.
7379 (rs6000_emit_prologue_components): Ditto.
7380 (rs6000_emit_epilogue_components): Ditto.
7381 (rs6000_emit_prologue): Ditto.
7382 (rs6000_emit_epilogue): Ditto.
7383 (rs6000_elf_file_end): Ditto.
7384 (rs6000_function_value): Ditto.
7385 (rs6000_libcall_value): Ditto.
7386 * config/rs6000/rs6000.h: Ditto.
7387 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7388 (TARGET_MINMAX): ... this. New.
7389 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7390 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7391 (*movsi_internal1_single): Delete.
7392 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7393 mfpu=, mxilinx-fpu): Delete.
7394 * config/rs6000/singlefp.h: Delete.
7395 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7396 * config/rs6000/t-rs6000: Ditto.
7397 * config/rs6000/t-xilinx: Delete.
7398 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7399 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7400 (VStype_simple): Delete.
7401 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7402 * config/rs6000/xfpu.h: Delete.
7403 * config/rs6000/xfpu.md: Delete.
7404 * config/rs6000/xilinx.h: Delete.
7405 * config/rs6000/xilinx.opt: Delete.
7406 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7407 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7409 2018-05-04 Tom de Vries <tom@codesourcery.com>
7412 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7415 2018-05-04 Richard Biener <rguenther@suse.de>
7418 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7419 (expand_complex_div_wide): Likewise.
7420 (expand_complex_operations_1): Likewise.
7421 (expand_complex_libcall): Preserve EH info of the original stmt.
7422 (tree_lower_complex): Handle removed blocks.
7423 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7424 on complex multiplication and division libcall builtins.
7426 2018-05-04 Richard Biener <rguenther@suse.de>
7429 * fold-const.c (negate_expr_p): Restrict negation of operand
7430 zero of a division to when we know that can happen without
7432 (fold_negate_expr_1): Likewise.
7434 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7437 * real.h (real_nextafter): Declare.
7438 * real.c (real_nextafter): New function.
7439 * fold-const-call.c (fold_const_nextafter): New function.
7440 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7441 CASE_CFN_NEXTTOWARD.
7442 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7443 even when arg1_mode is different from arg0_mode.
7445 2018-05-03 Nathan Sidwell <nathan@acm.org>
7447 * doc/extend.texi (Deprecated Features): Remove
7449 (Backwards Compatibility): Likewise.
7450 * doc/invoke.texi (C++ Language Options): Likewise.
7451 (C++ Dialect Options): Likewise.
7453 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7456 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7457 _mm512_mask_mullox_epi64): New intrinsics.
7459 2018-05-03 Tom de Vries <tom@codesourcery.com>
7462 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7463 dump files): Add offload-tree.
7465 2018-05-03 Richard Biener <rguenther@suse.de>
7467 PR tree-optimization/85615
7468 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7469 to loops not nested in BBs loop father to avoid creating multi-entry
7472 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7474 PR tree-optimization/70291
7475 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7476 arguments. Change return type to tree. Emit libcall as a new
7477 statement rather than replacing existing one when inplace_p is true.
7478 (expand_complex_multiplication_components): New function.
7479 (expand_complex_multiplication): Expand floating-point complex
7480 multiplication using the above.
7481 (expand_complex_division): Rename inner_type parameter to type.
7482 Update expand_complex_libcall call-site.
7483 (expand_complex_operations_1): Update expand_complex_multiplication
7484 and expand_complex_division call-sites.
7486 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7489 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7490 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7491 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7492 the highest significant bit of the shift count mask is clear. In
7493 check whether and[sq]i3 is needed verify that all significant bits
7494 of the shift count other than the highest are set.
7496 2018-05-02 Tom de Vries <tom@codesourcery.com>
7499 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7500 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7501 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7502 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7503 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7504 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7505 __builtin_goacc_parlevel_size.
7507 2018-05-02 Richard Biener <rguenther@suse.de>
7509 PR tree-optimization/85597
7510 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7511 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7514 2018-05-02 Tom de Vries <tom@codesourcery.com>
7517 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7518 dump files): Add ltrans-tree.
7520 2018-05-02 Tom de Vries <tom@codesourcery.com>
7523 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7524 dump files): Add wpa-ipa.
7526 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7528 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7529 powerpc*-*-linux*paired* target.
7530 * config/rs6000/750cl.h: Delete.
7531 * config/rs6000/paired.h: Delete.
7532 * config/rs6000/paired.md: Delete.
7533 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7535 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7536 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7537 comment. Remove paired float support.
7538 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7539 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7541 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7542 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7544 * config/rs6000/rs6000.c: Remove paired float support.
7545 (paired_expand_vector_init, paired_expand_vector_move,
7546 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7547 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7548 paired_expand_builtin, paired_expand_predicate_builtin,
7549 paired_init_builtins): Delete.
7550 * config/rs6000/rs6000.h: Remove paired float support.
7551 * config/rs6000/rs6000.md: Remove paired float support.
7552 (move_from_CR_ov_bit): Delete.
7553 * config/rs6000/rs6000.opt (mpaired): Delete.
7554 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7555 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7557 2018-05-02 Richard Biener <rguenther@suse.de>
7560 * gimplify.c (gimplify_save_expr): When in SSA form allow
7561 SAVE_EXPRs to compute to SSA vars.
7563 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7566 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7567 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7568 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7569 clobber operands[2], instead use a new pseudo. Formatting fixes.
7571 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7573 PR tree-optimization/85586
7574 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7575 exit early for statements in the same group if the accesses are
7578 2018-05-02 Tom de Vries <tom@codesourcery.com>
7581 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7584 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7586 PR tree-optimization/85143
7587 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7589 2018-05-01 Tom de Vries <tom@codesourcery.com>
7592 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7593 not found" error message.
7595 2018-05-01 Tom de Vries <tom@codesourcery.com>
7598 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7599 * vec.c (test_ordered_remove_if): New function.
7600 (vec_c_tests): Call test_ordered_remove_if.
7601 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7602 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7603 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7604 VEC_ORDERED_REMOVE_IF.
7606 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7608 PR tree-optimization/82665
7609 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7610 pointer subtraction where arguments come from a memchr call.
7612 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7614 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7615 --push-state --as-needed and --pop-state instead of --as-needed and
7616 --no-as-needed if ld supports it.
7617 * configure: Regenerated.
7620 * doc/install.texi2html: Replace _002d with - and _002a with * in
7621 generated html files using sed.
7623 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7626 * gcc-rich-location.c (blank_line_before_p): New function.
7627 (use_new_line): New function.
7628 (gcc_rich_location::add_fixit_insert_formatted): New function.
7629 * gcc-rich-location.h
7630 (gcc_rich_location::add_fixit_insert_formatted): New function.
7632 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7634 * selftest.c (assert_streq): Rename "expected" and "actual" to
7635 "val1" and "val2". Extend NULL-handling to cover both inputs
7636 symmetrically, while still requiring both to be non-NULL for a pass.
7637 * selftest.h (assert_streq): Rename "expected" and "actual" to
7639 (ASSERT_EQ): Likewise.
7640 (ASSERT_EQ_AT): Likewise.
7641 (ASSERT_KNOWN_EQ): Likewise.
7642 (ASSERT_KNOWN_EQ_AT): Likewise.
7643 (ASSERT_NE): Likewise.
7644 (ASSERT_MAYBE_NE): Likewise.
7645 (ASSERT_MAYBE_NE_AT): Likewise.
7646 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7647 the assertion to pass.
7648 (ASSERT_STREQ_AT): Likewise.
7650 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7652 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7653 interaction with -pie.
7655 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7657 * selftest.h: Fix alphabetization of per-source-file selftest
7660 2018-04-30 Jason Merrill <jason@redhat.com>
7662 PR c++/61982 - dead stores to destroyed objects.
7663 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7666 2018-04-30 Jason Merrill <jason@redhat.com>
7668 * tree.c (build_clobber): New.
7669 * tree.h: Declare it.
7670 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7672 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7674 * diagnostic-show-locus.c (layout::layout): Update for
7675 location_get_source_line returning a char_span.
7676 (struct char_span): Move to input.h.
7677 (struct correction): Update for fields in char_span becoming
7679 (struct source_line): Update for location_get_source_line
7680 returning a char_span.
7681 (layout::print_line): Likewise.
7682 * edit-context.c (edited_file::print_content): Likewise.
7683 (edited_file::print_diff_hunk): Likewise.
7684 (edited_file::print_run_of_changed_lines): Likewise.
7685 (edited_file::get_num_lines): Likewise.
7686 (edited_line::edited_line): Likewise.
7687 * final.c (asm_show_source): Likewise.
7688 * input.c (location_get_source_line): Convert return type
7689 from const char * to char_span, losing the final "line_len"
7691 (dump_location_info): Update for the above.
7692 (get_substring_ranges_for_loc): Likewise. Use a char_span
7693 when handling the literal within the line.
7694 (test_reading_source_line): Update for location_get_source_line
7695 returning a char_span.
7696 * input.h (class char_span): Move here from
7697 diagnostic-show-locus.c, converting from a struct to a class.
7698 Make data members private.
7699 (char_span::operator bool): New.
7700 (char_span::length): New.
7701 (char_span::get_buffer): New.
7702 (char_span::operator[]): New.
7703 (char_span::subspan): Make const.
7704 (char_span::xstrdup): New.
7705 (location_get_source_line): Convert return type from const char *
7706 to char_span, losing the final "line_size" param.
7708 2018-04-30 Jan Hubicka <jh@suse.cz>
7710 * lto-wrapper.c (ltrans_priorities): New static var.
7711 (cmp_priority): New.
7712 (run_gcc): Read priorities and if doing parallel build order
7713 the Makefile by them.
7715 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7717 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7719 2018-04-30 Richard Biener <rguenther@suse.de>
7721 * tree-cfg.c (verify_address): Remove base argument, add
7722 flag whether to check TREE_ADDRESSABLE and do that.
7723 (verify_expr): Remove.
7724 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7725 (verify_gimple_assign_single): Likewise.
7726 (verify_gimple_switch): Likewise.
7727 (verify_expr_location_1): Dereference tp once. Add (disabled)
7728 piece from verify_expr.
7729 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7731 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7733 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7735 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7737 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7738 (small_data_pattern): Likewise.
7739 (arc_rewrite_small_data): Likewise.
7740 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7741 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7742 (get_symbol_alignment): New function.
7743 (legitimate_small_data_address_p): Likewise.
7744 (legitimate_scaled_address): Update, call
7745 legitimate_small_data_address_p.
7746 (output_sdata): New static variable.
7747 (arc_print_operand): Update how we handle small data operands.
7748 (arc_print_operand_address): Likewise.
7749 (arc_legitimate_address_p): Update, use
7750 legitimate_small_data_address_p.
7751 (arc_rewrite_small_data_p): Remove.
7752 (arc_rewrite_small_data_1): Likewise.
7753 (arc_rewrite_small_data): Likewise.
7754 (small_data_pattern): Likewise.
7755 (compact_sda_memory_operand): Update to use
7756 legitimate_small_data_address_p and get_symbol_alignment.
7757 (prepare_move_operands): Don't rewite sdata pattern.
7758 (prepare_extend_operands): Remove.
7759 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7761 (zero_extendqisi2): Likewise.
7762 (zero_extendhisi2): Likewise.
7763 (extendqihi2): Likewise.
7764 (extendqisi2): Likewise.
7765 (extendhisi2): Likewise.
7769 * config/arc/constraints.md (Usd): Change it to memory constraint.
7771 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7773 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7774 as source of std instructions.
7775 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7776 allow 6-bit constants as source for store instructions.
7777 (movdi_insn): Update instruction pattern to allow 6-bit constants
7778 as source for store instructions.
7780 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7782 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7784 2018-04-30 Nathan Sidwell <nathan@acm.org>
7785 Sandra Loosemore <sandra@codesourcery.com>
7787 * dumpfile.c (dump_open): Allow '-' for stdout.
7788 * doc/invoke.texi (Developer Options): Document dump filename
7789 determination early. Document stdin/stdout selection.
7791 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7793 Microblaze Target: PIC data text relative
7795 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7796 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7798 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7799 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7800 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7801 New addressing mode for data-text relative position indepenedent code.
7802 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7803 'ADDRESS_SYMBOLIC_TXT_REL'.
7804 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7805 (microblaze_legitimate_pic_operand): Exclude function calls from
7806 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7807 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7809 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7810 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7811 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7812 for 'address + offset'.
7813 (microblaze_expand_prologue): Add new function prologue call for
7815 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7816 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7817 table in case of TARGET_PIC_DATA_TEXT_REL.
7818 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7819 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7820 Add new macros 'UNSPEC_TEXT',
7821 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7822 + exclude function calls from 'UNSPEC_PLT' in case of data text
7824 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7825 new target hook for generating address diff vector tables in case of
7827 * doc/tm.texi : Regenerate.
7828 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7829 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7830 of addr diff vector generation.
7831 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7832 target hook definition.
7833 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7834 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7835 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7836 option for data text relative.
7838 2018-04-30 Richard Biener <rguenther@suse.de>
7840 * tree-chrec.h (evolution_function_is_constant_p): Remove
7842 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7844 2018-04-30 Richard Biener <rguenther@suse.de>
7847 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7849 2018-04-30 Richard Biener <rguenther@suse.de>
7851 PR tree-optimization/28364
7852 PR tree-optimization/85275
7853 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7854 copying first exit test.
7856 2018-04-28 Mark Wielaard <mark@klomp.org>
7858 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7860 (dwarf_AT): Handle DW_AT_addr_base.
7861 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7863 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7866 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7867 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7868 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7869 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7871 2018-04-28 Richard Biener <rguenther@suse.de>
7873 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7874 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7875 to reflect use. Only add interesting stmts.
7877 2018-04-27 Martin Jambor <mjambor@suse.cz>
7880 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7881 the jump function allows for passing through aggregate values.
7883 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7885 * input.h (in_system_header_at): Convert from macro to inline
7887 (from_macro_expansion_at): Likewise.
7888 (from_macro_definition_at): Likewise.
7890 2018-04-27 Jeff Law <law@redhat.com>
7892 * config.gcc: Mark tile* targets as deprecated/obsolete.
7894 2018-04-27 Richard Biener <rguenther@suse.de>
7896 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7899 2018-04-27 Richard Biener <rguenther@suse.de>
7901 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7903 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7905 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7906 with Yd constraint. Set "preferred_for_speed" attribute from
7907 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7909 (*movdi_internal): Ditto.
7910 (movti_interunit splitters): Remove
7911 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7912 (movdi_interunit splitters): Ditto.
7913 * config/i386/constraints.md (Ye): Remove.
7914 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7916 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7919 * config/aarch64/constraints.md (Usg): Limit to 31.
7922 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7924 PR tree-optimization/85529
7925 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7926 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7927 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7928 zero extension or masking of the MSB bit.
7929 (optimize_range_tests): Add FIRST_BB argument, pass it through
7930 to optimize_range_tests_var_bound.
7931 (maybe_optimize_range_tests, reassociate_bb): Adjust
7932 optimize_range_tests callers.
7934 2018-04-26 Richard Biener <rguenther@suse.de>
7935 Jakub Jelinek <jakub@redhat.com>
7937 * cgraph.h (symbol_table): Just declare debug method here.
7938 * symtab.c (symbol_table::debug): Define.
7940 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7942 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7944 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7946 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7947 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7948 (*movdi_internal): Substitute Yi and Yj constraint with x
7949 and Ym and Yn constraint with y constraint. Update "isa"
7950 attribute and set "preferred_for_speed" attribute from
7951 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7952 (*movsi_internal): Ditto.
7953 (*movdf_internal): Ditto.
7954 (*movsf_internal): Ditto.
7955 (*zero_extendsidi2): Ditto.
7956 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7957 (sse2_loadld): Ditto.
7958 (*vec_extract<ssevecmodelower>_0): Ditto.
7959 (*vec_extractv4si_0_zext_sse4): Ditto.
7960 (vec_concatv2di): Ditto.
7961 (*vec_dup<mode>): Ditto.
7962 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7963 * config/i386/constraints.md (Yi): Remove.
7968 2018-04-26 Nathan Sidwell <nathan@acm.org>
7970 * dumpfile.c (dump_open): New.
7971 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7972 (dump_finish): Detect stdio/stderr by value not name.
7974 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7976 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7978 2018-04-26 Tom de Vries <tom@codesourcery.com>
7981 * config/nvptx/nvptx.c (verify_neutering_jumps)
7982 (verify_neutering_labels): New function
7983 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7985 2018-04-26 Tom de Vries <tom@codesourcery.com>
7988 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7989 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7992 2018-04-26 Richard Biener <rguenther@suse.de>
7993 Tom de Vries <tom@codesourcery.com>
7996 * lto-streamer-out.c (output_function): Fixup loops if required to match
7997 discovery done in the reader.
7999 2018-04-26 Richard Biener <rguenther@suse.de>
8001 PR tree-optimization/85116
8002 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8003 have a loop exit from the single latch predecessor. Remove
8004 case of header with just condition.
8005 (ch_base::copy_headers): Exclude infinite loops from any
8007 (pass_ch::execute): Record exits.
8009 2018-04-26 Richard Biener <rguenther@suse.de>
8011 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8012 prologue cost vector and pass it to vect_get_load_cost.
8013 (vect_get_peeling_costs_all_drs): Likewise.
8014 (vect_peeling_hash_get_lowest_cost): Likewise.
8015 (vect_enhance_data_refs_alignment): Likewise.
8017 2018-04-26 Richard Biener <rguenther@suse.de>
8020 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8021 checking of integer<->pointer conversions.
8022 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8023 sign-/zero-extending pointer types.
8024 (expand_omp_for_static_chunk): Likewise.
8026 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8027 Jean Lee <xiaoyur347@gmail.com>
8029 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8030 (TARGET_ASAN_SHADOW_OFFSET): Define.
8031 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8032 true for -fsanitize=address.
8034 2018-04-25 Mark Wielaard <mark@klomp.org>
8036 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8039 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8041 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8042 than "alu", remove explicit "memory" and "imm_disp" attributes.
8043 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8046 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8047 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8048 gen_lowpart_no_emit.
8050 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8053 * config/i386/i386.c (ix86_expand_builtin): Change memory
8054 operand to XI, extend p0 to Pmode.
8055 * config/i386/i386.md: Change unspec volatile and operand
8056 1 mode to XI, change operand 0 mode to P.
8058 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8060 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8061 GET_MODE_MASK before any checking.
8062 (nds32_can_use_bset_p): Likewise.
8063 (nds32_can_use_btgl_p): Likewise.
8065 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8067 * config/nds32/nds32-doubleword.md: New define_split pattern for
8068 illegal register number.
8070 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8072 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8074 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8076 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8078 2018-04-25 Richard Biener <rguenther@suse.de>
8080 * lto-streamer.h (LTO_major_version): Bump to 8.
8082 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8084 * BASE-VER: Set to 9.0.0.
8086 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8088 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8089 in __abskf2 and __powikf2.
8091 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8094 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8095 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8096 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8097 Use the above on operand 2. Reindent.
8098 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8100 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8103 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8105 * config/i386/i386.opt (mcet): Removed.
8106 * doc/install.texi: Remove -mcet documentation.
8107 * doc/invoke.texi: Likewise.
8109 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8112 * doc/install.texi: Remove -mcet from bootstrap-cet.
8114 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8117 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8118 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8122 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8123 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8124 containing a CONST_VECTOR.
8126 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8128 * doc/install.texi: Update newlib dependency for nvptx.
8130 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8133 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8134 instead of INTVAL when shifting x left.
8136 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8138 PR tree-optimization/85478
8139 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8140 vect_grouped_store_supported for single element vectors.
8142 2018-04-24 Richard Biener <rguenther@suse.de>
8145 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8146 load cost increase to the case of non-constant step.
8148 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8151 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8152 destination if any_malformed_asm.
8154 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8157 * expr.c (store_field): In the bitfield case, if the value comes from
8158 a function call and is returned in registers by means of a PARALLEL,
8159 do not change the mode of the temporary unless BLKmode and VOIDmode.
8161 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8163 PR rtl-optimization/85423
8164 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8165 dependencies to debug insns when the previous insn is non-debug.
8167 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8169 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8170 enums into a single definition.
8171 (fls): Fix predicates and printing.
8174 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8176 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8177 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8178 and short u6 immediate.
8179 (check_if_valid_sleep_operand): Remove.
8180 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8182 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8184 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8185 flag_always_save_lp condition.
8186 * config/nds32/nds32.opt (malways-save-lp): New option.
8188 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8190 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8191 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8192 * config/nds32/nds32.h
8193 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8194 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8196 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8198 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8199 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8201 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8202 Chung-Ju Wu <jasonwucj@gmail.com>
8204 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8206 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8207 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8209 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8211 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8213 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8215 * config/nds32/nds32-protos.h (nds32_data_alignment,
8216 nds32_local_alignment): Declare.
8217 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8218 nds32_local_alignment): New functions.
8219 (TARGET_CONSTANT_ALIGNMENT): Define.
8220 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8222 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8224 * config/nds32/nds32.c
8225 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8226 (TARGET_MODES_TIEABLE_P): Likewise.
8228 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8230 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8233 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8234 Chung-Ju Wu <jasonwucj@gmail.com>
8236 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8237 for unaligned access.
8238 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8240 * config/nds32/nds32-intrinsic.md: Likewise.
8241 * config/nds32/nds32_intrinsic.h: Likewise.
8242 * config/nds32/nds32.h (nds32_builtins): Likewise.
8243 * config/nds32/nds32.opt (munaligned-access): New option.
8244 * config/nds32/nds32.c (nds32_asm_file_start): Display
8245 flag_unaligned_access status.
8247 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8249 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8250 -mno-relax is present.
8251 * config/riscv/linux.h (LINK_SPEC): Ditto.
8253 2018-04-20 Martin Sebor <msebor@redhat.com>
8256 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8258 (gimple_fold_builtin_stxcpy_chk): Same.
8259 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8261 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8264 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8265 __powikf2 when long double is IEEE 128-bit.
8267 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8269 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8270 step to make sure stack always aligned.
8272 2018-04-20 Carl Love <cel@us.ibm.com>
8275 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8276 size check for arg0.
8278 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8279 Tom de Vries <tom@codesourcery.com>
8282 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8283 Emit insns for calls too.
8284 (nvptx_find_par): Always look for worker-level predecessor insn.
8285 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8287 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8288 (nvptx_process_pars): Propagate frames for calls.
8290 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8293 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8295 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8296 (ix86_handle_option): Don't handle OPT_mibt.
8297 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8299 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8301 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8302 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8303 (ix86_target_macros): Define __CET__ with flag_cf_protection
8304 for -fcf-protection.
8305 * config/i386/i386.c (isa2_opts): Remove -mibt.
8306 * config/i386/i386.h (TARGET_IBT): Removed.
8307 (TARGET_IBT_P): Likewise.
8308 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8309 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8310 * config/i386/i386.opt (mcet): Update help message.
8313 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8314 -mcet as an alias for -mshstk.
8316 2018-04-20 Richard Biener <rguenther@suse.de>
8319 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8320 complexity by forcing a single use of the multiply operand.
8322 2018-04-20 Martin Jambor <mjambor@suse.cz>
8325 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8326 recursion dependency to only apply to non-clones.
8328 2018-04-20 Martin Jambor <mjambor@suse.cz>
8331 * ipa-cp.c (create_specialized_node): Check that clones of
8332 self-recursive edges exist during IPA-CP.
8334 2018-04-19 Toon Moene <toon@moene.org>
8336 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8339 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8341 PR tree-optimization/85467
8342 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8343 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8344 VECTOR_CST element to type.
8346 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8349 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8350 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8351 (builtin_longjmp): Likewise.
8352 (save_stack_nonlocal): New pattern.
8353 (restore_stack_nonlocal): Likewise.
8355 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8358 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8359 Replace ASM_OUTPUT_LABEL with fprintf.
8361 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8364 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8365 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8366 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8367 define __IBT__ and __SHSTK__ for -fcf-protection.
8368 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8370 (ix86_trampoline_init): Likewise.
8371 (x86_output_mi_thunk): Likewise.
8372 (ix86_notrack_prefixed_insn_p): Likewise.
8373 (ix86_option_override_internal): Don't disallow -fcf-protection.
8374 * config/i386/i386.md (rdssp<mode>): Also enable for
8376 (incssp<mode>): Likewise.
8377 (nop_endbr): Likewise.
8378 * config/i386/i386.opt (mcet): Change help message to built-in
8382 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8383 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8384 enable CET built-in functions.
8386 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8388 * common/config/i386/i386-common.c
8389 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8390 OPTION_MASK_ISA_MOVDIRI_UNSET,
8391 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8392 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8393 * config.gcc (movdirintrin.h): New header.
8394 * config/i386/cpuid.h (bit_MOVDIRI,
8395 bit_MOVDIR64B): New bits.
8396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8398 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8399 (VOID, PVOID, PCVOID)): New function types.
8400 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8401 __builtin_ia32_directstoreu_u64,
8402 __builtin_ia32_movdir64b): New builtins.
8403 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8404 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8406 (ix86_valid_target_attribute_inner_p): Ditto.
8407 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8408 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8409 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8410 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8411 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8412 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8413 (movdiri<mode>, movdir64b_<mode>): New.
8414 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8415 * config/i386/immintrin.h: Include movdirintrin.h.
8416 * config/i386/movdirintrin.h: New file.
8417 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8419 2018-04-19 Richard Biener <rguenther@suse.de>
8422 * cfg.c (clear_bb_flags): When loop state says we have
8423 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8425 2018-04-19 Richard Biener <rguenther@suse.de>
8427 PR tree-optimization/84737
8428 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8429 copying restrict info.
8430 (vect_setup_realignment): Use it.
8431 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8432 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8433 the first DR to all generated stores.
8434 (vectorizable_load): Likewise for loads.
8436 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8438 PR tree-optimization/85446
8439 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8440 the integral and pointer types to have the same precision.
8442 * doc/install.texi: Document --disable-cet being the default and
8445 2018-04-18 Martin Liska <mliska@suse.cz>
8447 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8450 2018-04-18 Martin Liska <mliska@suse.cz>
8453 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8456 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8457 arguments if they are comparable.
8459 2018-04-18 Martin Liska <mliska@suse.cz>
8462 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8465 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8468 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8471 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8472 ENDBR after calling __morestack.
8474 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8477 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8478 by using gcc_base_ver to generate a gcc_driver_version, and use
8479 it when generating GCC_DRIVER_NAME.
8480 * configure: Regenerate.
8482 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8485 * config.gcc: Obsolete powerpc*-*-*spe*.
8487 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8490 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8491 (stabstr_D): Change type of unum from unsigned int to
8492 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8495 2018-04-17 Jim Wilson <jimw@sifive.com>
8498 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8499 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8500 Set arg_pointer_offset after using pretend_args_size.
8502 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8504 PR rtl-optimization/85431
8505 * dse.c (record_store): Ignore zero width stores.
8508 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8509 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8510 __builtin_stack_restore rather than after it.
8511 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8512 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8513 argument instead of virtual_dynamic_stack_rtx.
8515 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8517 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8519 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8520 Add note to error message to explain internal mapping of overloaded
8521 built-in function name to non-overloaded built-in function name.
8522 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8525 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8528 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8529 where the inputs overlap with the output.
8531 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8533 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8534 (=v, v) alternative and explicit "memory" attribute.
8535 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8536 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8538 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8539 "sselog1" type instead of "sselog".
8540 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8541 "sselog". Remove explicit "memory" attribute.
8542 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8543 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8545 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8546 "sselog1" type instead of "sselog". Remove explicit "memory"
8548 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8549 use "sselog1" type instead of "sselog". Remove explicit "memory"
8551 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8552 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8554 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8555 "sselog1" type instead of "sselog". Remove explicit "memory"
8557 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8558 use "sselog1" type instead of "sselog". Remove explicit "memory"
8562 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8565 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8568 2018-04-17 Martin Jambor <mjambor@suse.cz>
8571 * ipa-cp.c (create_specialized_node): Call
8572 expand_all_artificial_thunks if necessary.
8574 2018-04-17 Martin Liska <mliska@suse.cz>
8577 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8578 in message, remote space in between '_G' and '('.
8580 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8583 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8584 avx512f_vmcmp<mode>3<round_saeonly_name>,
8585 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8586 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8587 avx512f_rndscale<mode><round_saeonly_name>,
8588 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8589 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8590 Use %<iptr>2 instead of %2 for -masm=intel.
8591 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8592 avx512f_vcvttss2usi<round_saeonly_name>,
8593 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8595 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8596 avx512f_vcvttsd2usi<round_saeonly_name>,
8597 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8598 Use %q1 instead of %1 for -masm=intel.
8599 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8600 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8601 of %3 for -masm=intel.
8602 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8604 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8606 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8608 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8609 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8611 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8613 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8614 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8615 %g1 and one with %0 and %1.
8616 (avx512er_vmrcp28<mode><round_saeonly_name>,
8617 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8619 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8620 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8621 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8622 of %0 and %{%4%} for -masm=intel.
8623 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8624 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8625 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8626 order of %0 and %{%5%}%{z%} for -masm=intel.
8628 2018-04-17 Jan Hubicka <jh@suse.cz>
8631 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8633 2018-04-17 Martin Liska <mliska@suse.cz>
8636 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8637 for target_clone error message. Make default implementation
8638 clone to be a local declaration.
8639 (separate_attrs): Add new argument and check for an empty
8641 (expand_target_clones): Handle it.
8642 (ipa_target_clone): Make redirection just for target_clones
8645 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8646 Tom de Vries <tom@codesourcery.com>
8649 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8650 tiled basic blocks without omp continue statements.
8652 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8655 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8656 vec_extract expression as having side effects to make sure it gets
8659 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8662 * config/i386/i386.c (get_builtin_code_for_version): Check
8665 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8668 * gcc/config.gcc: Support "skylake".
8669 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8671 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8672 (processor_target_table): Add "skylake".
8673 (ix86_option_override_internal): Add "skylake".
8674 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8675 PROCESSOR_CANNONLAKE.
8676 (get_builtin_code_for_version): Fix priority for
8677 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8678 PROCESSOR_SKYLAKE-AVX512.
8679 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8680 (processor_type): Add PROCESSOR_SKYLAKE.
8682 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8683 Jason Merrill <jason@redhat.com>
8686 * convert.c (convert_to_integer_1): Use direct recursion for
8687 enumeral types and types with a precision less than the number
8688 of bits in their mode.
8690 2018-04-16 Julia Koval <julia.koval@intel.com>
8693 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8694 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8696 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8699 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8700 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8701 and -mno-direct-move.
8703 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8706 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8707 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8708 Ensure negative shifts result in {0}.
8710 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8712 PR rtl-optimization/79916
8713 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8714 regs (if any) to define how to gnerate SD moves when LRA is in
8717 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8719 PR rtl-optimization/85393
8720 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8721 * except.c (expand_dw2_landing_pad_for_region): Make static.
8722 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8723 a label and unconditional jump to old_bb, rather than
8724 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8727 PR rtl-optimization/85376
8728 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8729 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8730 instead of a specific value.
8732 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8733 Bin Cheng <bin.cheng@arm.com>
8735 PR tree-optimization/82965
8736 PR tree-optimization/83991
8737 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8738 by_profile_only parameter.
8739 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8740 information if the loop was predicted to iterate too many times.
8741 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8743 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8746 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8749 2018-04-13 Martin Liska <mliska@suse.cz>
8750 Jakub Jelinek <jakub@redhat.com>
8753 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8754 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8755 * builtins.c (expand_builtin_memory_copy_args): Use
8756 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8757 handle dest_addr == pc_rtx.
8759 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8762 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8763 asked to not generate direct moves.
8764 (fix_trunc<mode>si2_stfiwx): Similar.
8765 (fix_trunc<mode>si2_internal): Similar.
8767 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8770 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8771 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8772 lookup if dest in some wider mode is known to be const0_rtx and
8773 if so, record permanent equivalence for it to be ZERO_EXTEND of
8774 the narrower mode destination.
8776 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8778 * lto-streamer-out.c (output_function): Revert 259346.
8779 * omp-expand.c (expand_oacc_for): Likewise.
8781 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8783 PR rtl-optimization/85354
8784 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8785 * sel-sched.c (sel_global_init): ... here.
8787 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8790 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8791 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8792 mode for PE-COFF targets.
8793 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8794 (i386_pe_asm_lto_end): Likewise.
8795 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8796 (TARGET_ASM_LTO_END): Likewise.
8797 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8798 (i386_pe_asm_lto_start): New function.
8799 (i386_pe_asm_lto_end): Likewise.
8801 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8802 Richard Biener <rguenther@suse.de>
8805 * lto-streamer-out.c (output_function): Fix CFG loop state before
8807 * omp-expand.c (expand_oacc_for): Handle calls to internal
8808 functions like regular functions.
8810 2018-04-12 Richard Biener <rguenther@suse.de>
8813 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8814 for the early LTO debug to properly generate references to it
8815 during DIE emission. Do not re-use that for the skeleton for
8817 (dwarf2out_early_finish): Likewise.
8819 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8822 * config/i386/sse.md
8823 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8824 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8825 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8826 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8827 and output is a reg, avoid creating invalid lowpart subreg, but
8828 instead split into a 512-bit move. Don't split if not AVX512VL,
8829 input is xmm16+ reg and output is a mem.
8830 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8831 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8832 xmm16+ reg and output is a mem.
8834 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8836 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8837 also for flag_dwarf2_cfi_asm.
8839 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8841 PR rtl-optimization/85342
8842 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8843 a bool scalar var inside of the loop instead. Don't try to update
8844 recog_data.operand after failed apply_change_group.
8846 2018-04-12 Tom de Vries <tom@codesourcery.com>
8849 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8850 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8851 array with flexible array member as array without given dimension.
8852 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8853 argument for undefined param to true.
8855 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8858 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8859 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8860 from PowerPC section.
8861 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8862 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8863 help text that is too long.
8864 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8865 help text that is too long.
8866 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8867 help text that is too long.
8869 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8871 * config/alpha/alpha.md (stack_probe_internal): Rename
8872 from "probe_stack". Update all callers.
8874 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8876 PR rtl-optimization/84566
8877 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8878 sched_macro_fuse_insns.
8880 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8883 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8884 (compute_block_dependences): ... from here.
8886 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8888 PR tree-optimization/85331
8889 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8890 from int to HOST_WIDE_INT.
8892 2018-04-11 Martin Jambor <mjambor@suse.cz>
8895 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8896 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8897 not the same as the source val.
8898 (cgraph_edge_brings_value_p): New parameter.
8899 (gather_edges_for_value): Pass destination value to
8900 cgraph_edge_brings_value_p.
8901 (perhaps_add_new_callers): Likewise.
8902 (get_info_about_necessary_edges): Likewise and exclude values brought
8903 only by self-recursive edges.
8904 (create_specialized_node): Redirect only clones of self-calling edges.
8905 (+self_recursive_pass_through_p): New function.
8906 (find_more_scalar_values_for_callers_subset): Use it.
8907 (find_aggregate_values_for_callers_subset): Likewise.
8908 (known_aggs_to_agg_replacement_list): Removed.
8909 (decide_whether_version_node): Re-calculate known constants for all
8910 remaining context clones.
8912 2018-04-11 Richard Biener <rguenther@suse.de>
8915 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8916 from early DWARF output.
8917 (dwarf2out_early_finish): Output line info unconditionally into
8918 early DWARF and add reference to it.
8920 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8923 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8924 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8925 other than V2DFmode using iptr mode attribute.
8926 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8928 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8930 PR rtl-optimization/84659
8931 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8933 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8936 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8938 (output_loc_list): Pass address of a dummy size variable even in the
8939 locview handling loop.
8940 (index_location_lists): Add comment on why skip_loc_list_entry can't
8943 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8946 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8949 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8952 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8953 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8954 and -mstring-compare-inline-limit.
8956 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8959 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8960 for stack clash protection in a register whenever we need it to be in
8963 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8965 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8966 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8968 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8971 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8973 (mlong-double-): Ditto.
8974 * config/rs6000/sysv4.opt (msdata=): Ditto.
8975 (mtls-size=): Ditto.
8977 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8979 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8980 erroneous entries for
8981 "vector int vec_ldl (int, long int *)", and
8982 "vector unsigned int vec_ldl (int, unsigned long int *)".
8983 Add comments and entries for
8984 "vector bool char vec_ldl (int, bool char *)",
8985 "vector bool short vec_ldl (int, bool short *)",
8986 "vector bool int vec_ldl (int, bool int *)",
8987 "vector bool long long vec_ldl (int, bool long long *)",
8988 "vector pixel vec_ldl (int, pixel *)",
8989 "vector long long vec_ldl (int, long long *)",
8990 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8991 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8992 type tree bool_long_long_type_node and correct definition of
8993 bool_V2DI_type_node to make reference to this new type tree.
8994 (rs6000_mangle_type): Replace erroneous reference to
8995 bool_long_type_node with bool_long_long_type_node.
8996 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8997 comments to emphasize sign distinctions for char and int types and
8998 replace RS6000_BTI_bool_long constant with
8999 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9000 use of RS6000_BTI_pixel.
9001 (bool_long_type_node): Remove this macro definition.
9002 (bool_long_long_type_node): New macro definition
9004 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9006 PR rtl-optimization/85300
9007 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9008 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9009 simplify_unary_operation fails.
9011 2018-04-10 Martin Liska <mliska@suse.cz>
9013 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9014 cgraph_edge and ipa_ref.
9016 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9020 * config/i386/sse.md
9021 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9022 computation of the VEC_MERGE selector from mask.
9023 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9024 Fix decoding of the VEC_MERGE selector into mask.
9026 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9028 PR tree-optimization/85286
9029 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9031 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9033 * final.c (final_1): Set insn_last_address as well as
9034 insn_current_address.
9036 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9039 * explow.c (emit_stack_probe): Call validize_mem on memory location
9040 before passing it to gen_probe_stack. Create address operand and
9041 legitimize it for the probe_stack_address case.
9043 2018-04-09 Jan Hubicka <jh@suse.cz>
9046 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9047 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9048 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9050 (free_lang_data): Rebuild type inheritance graph.
9052 2018-04-09 Martin Sebor <msebor@redhat.com>
9054 * invoke.texi (-finline-small-functions): Mention other optimization
9056 (-findirect-inlining, -fpartial-inlining): Same.
9057 (-finline-functions-called-once): Same.
9058 (-freorder-blocks-and-partition): Same.
9060 2018-04-09 Jan Hubicka <jh@suse.cz>
9063 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9064 jumps; choose last target that matches the criteria (i.e.
9065 no partition changes for non-crossing jumps).
9066 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9067 support for redirecting crossing jumps to non-crossing.
9069 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9071 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9072 also for naked functions.
9074 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9076 * config/arc/arc.md (add_shift): New pattern.
9077 (add_shift2): Likewise.
9078 (sub_shift): Likewise.
9079 (sub_shift_cmp0_noout): Likewise.
9080 (compare_si_ashiftsi): Likewise.
9081 (xbfu_cmp0_noout): New combine pattern.
9082 (xbfu_cmp0"): Likewise.
9083 (movsi_set_cc_insn): Place the predicable variant first.
9084 (commutative_binary_cmp0_noout): Remove clobber.
9085 (commutative_binary_cmp0): New pattern.
9086 (noncommutative_binary_cmp0): Likewise.
9087 (noncommutative_binary_cmp0_noout): Likewise.
9088 (noncommutative_binary_comparison_result_used): Removed.
9089 (rsub_cmp0): New pattern.
9090 (rsub_cmp0_noout): Likewise.
9091 (extzvsi): Changed, keep only meaningful variants.
9092 (SQH, SEZ): New iterators.
9093 (SQH_postfix): New mode attribute.
9094 (SEZ_prefix): New code attribute.
9095 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9096 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9097 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9099 (noncommutative_operator): Check the availability of barrel
9102 2018-04-09 Richard Biener <rguenther@suse.de>
9104 PR tree-optimization/85284
9105 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9106 Only use the niter constraining form of simple_iv when the exit
9109 2018-04-09 Tom de Vries <tom@codesourcery.com>
9112 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9113 (define_expand "*memory_barrier"): New define_expand.
9114 (define_insn "memory_barrier"): New insn.
9116 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9118 PR rtl-optimization/80463
9119 PR rtl-optimization/83972
9120 PR rtl-optimization/83480
9122 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9123 correct producer for the insn.
9124 (tidy_control_flow): Fixup seqnos in case of debug insns.
9126 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9128 PR rtl-optimization/83913
9130 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9131 different sched-times when merging exprs.
9133 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9135 PR rtl-optimization/83962
9137 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9138 tidy_fallthru_edge and tidy_control_flow.
9140 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9142 PR rtl-optimization/83530
9144 * sel-sched.c (force_next_insn): New global variable.
9145 (remove_insn_for_debug): When force_next_insn is true, also leave only
9146 next insn in the ready list.
9147 (sel_sched_region): When the region wasn't scheduled, make another pass
9148 over it with force_next_insn set to 1.
9150 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9152 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9154 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9155 for interrupt control.
9156 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9157 functions for interrupt control.
9158 * config/nds32/nds32-intrinsic.md: Likewise.
9159 * config/nds32/nds32_intrinsic.h: Likewise.
9160 * config/nds32/nds32.h (nds32_builtins): Likewise.
9162 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9164 * config/nds32/nds32.c (nds32_init_machine_status,
9165 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9166 strict_aligned_p field.
9167 (nds32_expand_to_rtl_hook): New function.
9168 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9169 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9171 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9172 Chung-Ju Wu <jasonwucj@gmail.com>
9174 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9175 * config/nds32/nds32-n7.md: New file.
9176 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9177 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9179 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9180 * config/nds32/nds32.md (pipeline_model): Add n7.
9181 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9182 * config/nds32/pipelines.md: Include n7 settings.
9184 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9185 Chung-Ju Wu <jasonwucj@gmail.com>
9187 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9188 * config/nds32/nds32-e8.md: New file.
9189 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9190 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9192 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9193 * config/nds32/nds32.md (pipeline_model): Add e8.
9194 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9195 * config/nds32/pipelines.md: Include e8 settings.
9197 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9198 Chung-Ju Wu <jasonwucj@gmail.com>
9200 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9201 * config/nds32/nds32-n8.md: New file.
9202 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9203 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9205 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9206 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9207 * config/nds32/nds32.md (pipeline_model): Add n8.
9208 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9209 * config/nds32/pipelines.md: Include n8 settings.
9211 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9212 Chung-Ju Wu <jasonwucj@gmail.com>
9214 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9215 * config/nds32/nds32-n9-2r1w.md: New file.
9216 * config/nds32/nds32-n9-3r2w.md: New file.
9217 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9218 nds32_register_ports): New or modify for cpu n9.
9219 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9221 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9222 * config/nds32/nds32-utils.c: New file.
9223 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9224 TARGET_MUL_SLOW): Define.
9225 * config/nds32/nds32.md (pipeline_model): New attribute.
9226 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9227 New options that support cpu n9.
9228 * config/nds32/pipelines.md: Include n9 settings.
9229 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9231 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9233 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9234 information if necessary.
9235 (output_cond_branch_compare_zero): Likewise.
9236 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9237 (nds32_target_alignment): Refine for alignment.
9238 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9239 (FUNCTION_BOUNDARY): Modify.
9240 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9242 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9244 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9246 * config/nds32/constants.md (unspec_volatile_element): Add values for
9247 TLB operation and data prefetch.
9248 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9249 functions for TLB operation and data prefetch.
9250 * config/nds32/nds32-intrinsic.md: Likewise.
9251 * config/nds32/nds32_intrinsic.h: Likewise.
9252 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9253 (nds32_print_operand): Likewise.
9254 * config/nds32/nds32.h (nds32_builtins): Likewise.
9256 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9257 Andrew Pinski <pinsika@gcc.gnu.org>
9260 * match.pd: Use constant_boolean_node of correct type instead of
9261 boolean_true_node or boolean_false_node for simplifying
9262 pointer comparisons to zero.
9264 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9266 PR tree-optimization/80021
9267 * tree.c (verify_type_variant): Make error call in verify_variant_match
9268 translatable and remove final full stop.
9270 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9272 * config/nds32/constants.md (unspec_volatile_element): Add
9273 UNSPEC_VOLATILE_EH_RETURN.
9274 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9275 nds32_output_stack_pop): Support dwarf exception handling process.
9276 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9277 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9278 exception handling process.
9279 (nds32_compute_stack_frame): Likewise.
9280 (nds32_return_addr_rtx): Likewise.
9281 (nds32_initial_elimination_offset): Likewise.
9282 (nds32_expand_prologue): Likewise.
9283 (nds32_expand_epilogue): Likewise.
9284 (nds32_dynamic_chain_address): New function.
9285 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9287 (DYNAMIC_CHAIN_ADDRESS): Define.
9288 (EH_RETURN_DATA_REGNO): Define.
9289 (EH_RETURN_STACKADJ_RTX): Define.
9290 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9291 patterns for dwarf exception handling.
9293 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9295 * config/nds32/nds32.h: Clean up obsolete macros.
9297 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9299 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9300 Add enum values for particular instructions.
9301 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9302 particular intrinsic functions.
9303 * config/nds32/nds32-intrinsic.md: Likewise.
9304 * config/nds32/nds32_intrinsic.h: Likewise.
9305 * config/nds32/nds32.h (nds32_builtins): Likewise.
9306 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9307 (btst, ave): New patterns for particular instructions.
9309 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9311 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9312 Add enum values for atomic load/store and memory sync.
9313 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9315 * config/nds32/nds32-intrinsic.md: Likewise.
9316 * config/nds32/nds32_intrinsic.h: Likewise.
9317 * config/nds32/nds32.h (nds32_builtins): Likewise.
9319 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9321 PR tree-optimization/85257
9322 * fold-const.c (native_encode_vector): If not all elts could fit
9323 and off is -1, return 0 rather than offset.
9324 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9325 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9326 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9327 adjust buffer in native_interpret_expr call.
9329 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9331 * config/nds32/constants.md (unspec_volatile_element): Add cache
9332 control enum values.
9333 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9334 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9335 * config/nds32/nds32.c (nds32_cctl_names): New.
9336 (nds32_print_operand): Handle cache control register names.
9337 * config/nds32/nds32.h (nds32_builtins): New enum values.
9338 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9340 * config/nds32/nds32.md (type): Add mmu.
9341 * config/nds32/pipelines.md (simple_insn): Add mmu.
9343 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9345 * config/nds32/nds32.md (type): Remove call.
9346 * config/nds32/pipelines.md (simple_insn): Likewise.
9348 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9350 * config/nds32/constants.md (unspec_volatile_element): Add
9351 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9352 UNSPEC_VOLATILE_FMFCFG.
9353 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9354 description for fmfcfg and fmfcsr.
9355 (bdesc_1arg): Add fmtcsr.
9356 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9357 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9358 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9359 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9360 unspec_fmfcfg): New patterns.
9361 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9362 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9363 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9364 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9365 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9366 __nds32__fmfcfg): Define.
9368 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9370 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9371 intrinsic register names.
9372 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9373 intrinsic register enum values and macros.
9375 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9377 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9378 for load/store addressing form.
9379 (nds32_print_operand_address): Likewise.
9381 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9384 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9385 based on LABEL_REF. Remove useless assertion.
9386 (pic_address_needs_scratch): Fix formatting.
9387 (sparc_legitimize_pic_address): Minor tweaks.
9388 (sparc_delegitimize_address): Adjust assertion accordingly.
9389 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9390 into symbolic_operand.
9391 (movsi_high_pic_label_ref): Likewise.
9392 (movsi_lo_sum_pic_label_ref): Likewise.
9393 (movdi_pic_label_ref): Likewise.
9394 (movdi_high_pic_label_ref): Likewise.
9395 (movdi_lo_sum_pic_label_ref): Likewise.
9397 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9399 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9400 custom LIB_SPEC setup.
9402 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9403 Kito Cheng <kito.cheng@gmail.com>
9405 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9406 * config/riscv/freebsd.h: New.
9408 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9410 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9411 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9414 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9415 Kito Cheng <kito.cheng@gmail.com>
9417 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9418 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9419 * config/nds32/nds32-protos.h (nds32_output_call,
9420 nds32_output_return): Declare.
9421 * config/nds32/nds32.md: Refine all the call and return patterns.
9423 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9426 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9427 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9429 PR rtl-optimization/84872
9430 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9431 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9434 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9436 * expr.c (copy_blkmode_to_reg): Revert 254862.
9437 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9439 2018-04-06 Richard Biener <rguenther@suse.de>
9442 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9443 after seeing a component reference with an adjacent field. Treat
9444 refs to arrays at struct end of external decls similar to
9445 refs to unconstrained commons.
9447 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9450 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9451 look through SAVE_EXPRs with non-side-effects argument. Adjust
9453 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9456 2018-04-06 Richard Biener <rguenther@suse.de>
9459 * alias.c (find_base_term): New wrapper around find_base_term
9460 unwinding CSELIB_VAL_PTR changes.
9461 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9464 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9466 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9468 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9469 constant definitions.
9470 ("nop"): lr 0,0 -> nopr r0
9471 ("nop_lr0", "nop_lr1"): New insn definitions.
9473 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9475 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9476 NDS32_V3PUSH_AVAILABLE_P macro.
9478 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9479 Chung-Ju Wu <jasonwucj@gmail.com>
9481 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9482 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9483 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9484 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9485 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9486 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9487 * config/nds32/constraints.md: New constraints and checking for hard
9488 float configuration.
9489 * config/nds32/iterators.md: New mode iterator and attribute for hard
9490 float configuration.
9491 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9493 * config/nds32/nds32-fpu.md: New file.
9494 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9495 deal with hard float code generation.
9496 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9498 (abi_type, float_reg_number): New enum type.
9499 * config/nds32/nds32-predicates.c: New predicates for hard float.
9500 * config/nds32/nds32-protos.h: Declare functions for hard float.
9501 * config/nds32/nds32.c: Implementation for hard float configuration.
9502 * config/nds32/nds32.h: Definitions for hard float configuration.
9503 * config/nds32/nds32.md: Include hard float machine description and
9504 modify patterns for hard float configuration.
9505 * config/nds32/nds32.opt: New options for hard float configuration.
9506 * config/nds32/predicates.md: New predicates for hard float
9509 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9511 * common/config/nds32/nds32-common.c
9512 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9514 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9517 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9518 CONSTRUCTOR_ELT (ctor, ...)->value.
9520 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9523 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9525 2018-04-05 Tom de Vries <tom@codesourcery.com>
9528 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9531 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9532 Kito Cheng <kito.cheng@gmail.com>
9534 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9535 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9536 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9537 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9539 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9540 Kito Cheng <kito.cheng@gmail.com>
9542 * config/nds32/constraints.md (Ufe): New memory constraint.
9543 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9544 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9545 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9547 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9548 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9550 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9552 * config/nds32/nds32.md: Use optimize_size in the condition for
9553 alu-shift instructions.
9555 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9557 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9559 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9561 * config/nds32/nds32.md (negsi2): Refine pattern.
9563 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9564 Chung-Ju Wu <jasonwucj@gmail.com>
9566 * config/nds32/iterators.md (shift_rotate): New code iterator.
9567 (shift): New code attribute.
9568 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9569 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9570 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9571 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9572 bit-wise operations.
9573 (andsi3, *andsi3): Ditto.
9574 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9575 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9576 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9577 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9578 nds32_ior_operand, nds32_xor_operand): New predicates.
9580 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9582 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9583 (addsi3, subsi3): ... this.
9585 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9587 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9589 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9591 * config/nds32/nds32.md: Adjust indention.
9593 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9595 * config/nds32/nds32.md (feature): New attribute.
9597 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9599 * config/nds32/nds32.md (subtype): New attribute.
9601 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9604 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9605 expansion to perform a bitwise AND of the argument followed by a
9606 boolean negation of the result.
9608 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9610 PR rtl-optimization/84878
9611 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9612 the basic block. Assert the use reference is not artificial and that
9613 it has an associated insn.
9615 2018-04-04 Michael Matz <matz@suse.de>
9617 * builtins.c (compute_objsize): Pass correct operand
9618 to array_at_struct_end_p.
9620 2018-04-04 Richard Biener <rguenther@suse.de>
9623 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9624 from contexts for DINFO_LEVEL_TERSE and below.
9626 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9628 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9629 resiter_operand condition.
9630 * config/nds32/nds32.md (*move<mode>): Ditto.
9632 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9633 Monk Chiang <sh.chiang04@gmail.com>
9635 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9637 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9641 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9642 Kito Cheng <kito.cheng@gmail.com>
9644 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9645 nds32_cond_code_str, output_cond_branch,
9646 output_cond_branch_compare_zero, nds32_expand_cbranch,
9647 nds32_expand_cstore, nds32_expand_movcc,
9648 nds32_output_cbranchsi4_equality_zero,
9649 nds32_output_cbranchsi4_equality_reg,
9650 nds32_output_cbranchsi4_equality_reg_or_const_int,
9651 nds32_output_cbranchsi4_greater_less_zero: New functions.
9652 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9653 nds32_expand_cstore, nds32_expand_movcc,
9654 nds32_output_cbranchsi4_equality_zero,
9655 nds32_output_cbranchsi4_equality_reg,
9656 nds32_output_cbranchsi4_equality_reg_or_const_int,
9657 nds32_output_cbranchsi4_greater_less_zero): Declare.
9658 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9659 nds32_rimm11s_operand): New predicates.
9660 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9661 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9664 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9666 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9667 * config/nds32/nds32.md: Ditto.
9668 * config/nds32/pipelines.md: Ditto.
9670 2018-04-04 Richard Biener <rguenther@suse.de>
9672 PR tree-optimization/85168
9673 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9674 propagating abnormals.
9676 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9678 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9680 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9681 Kito Cheng <kito.cheng@gmail.com>
9683 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9684 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9685 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9686 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9687 * config/nds32/nds32.md (sibcall_internal): New.
9688 (sibcall_register): Remove.
9689 (sibcall_immediate): Remove.
9690 (sibcall_value_internal): New.
9691 (sibcall_value_register): Remove.
9692 (sibcall_value_immediate): Remove.
9693 * config/nds32/predicates.md (nds32_general_register_operand): New.
9694 (nds32_call_address_operand): New.
9696 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9698 PR rtl-optimization/85167
9699 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9700 bb_defs if *split_p, instead preinitialize it to NULL.
9702 PR tree-optimization/85156
9703 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9704 evaluating the argument multiple times.
9706 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9708 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9710 (_mm_cvtpd_ps): Likewise.
9711 (_mm_cvttpd_epi32): Likewise.
9712 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9713 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9714 vector, pixel, and bool following altivec.h include.
9716 2018-04-03 Martin Sebor <msebor@redhat.com>
9718 * doc/extend.texi (Common Function Attributes): Clarify.
9719 (const attribute): Likewise.
9720 (pure attribute): Likewise.
9722 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9725 * config/i386/i386.c (ix86_expand_vector_set): Use
9726 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9728 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9730 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9731 instructions when changing rounding bits to preserve precision bits
9732 in the x87 control word.
9734 2018-04-03 Martin Liska <mliska@suse.cz>
9736 PR tree-optimization/82491
9737 * rtl.h (strip_offset_and_add): Replace += suboffset with
9738 poly_uint64 () + suboffset.
9740 2018-03-29 Martin Liska <mliska@suse.cz>
9741 Martin Jambor <mjambor@suse.cz>
9744 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9745 param_type is not an integral or pointer type.
9747 2018-04-03 Richard Biener <rguenther@suse.de>
9749 * sese.h (recompute_all_dominators): Remove.
9751 2018-04-02 Martin Sebor <msebor@redhat.com>
9753 * doc/invoke.texi (-Wrestrict): Fix typos.
9755 2018-04-02 Jim Wilson <jimw@sifive.com>
9757 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9758 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9759 (<optab>di3, <optab>si3_extend): Likewise.
9760 (<optab>si3_mask, <optab>si3_mask_1): New.
9761 (<optab>di3_mask, <optab>di3_mask_1): New.
9762 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9763 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9764 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9766 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9768 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9771 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9773 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9774 (nds32_canonicalize_comparison): New function.
9776 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9777 Kito Cheng <kito.cheng@gmail.com>
9778 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9780 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9781 * config/nds32/constants.md (unspec_volatile_element): Add
9782 UNSPEC_VOLATILE_RELAX_GROUP.
9783 * config/nds32/nds32-relax-opt.c: New file.
9784 * config/nds32/nds32-predicates.c
9785 (nds32_symbol_load_store_p): New function.
9786 * config/nds32/nds32-protos.h
9787 (nds32_symbol_load_store_p): Declare function.
9788 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9789 * config/nds32/nds32.c
9790 (nds32_register_pass): New function to register pass.
9791 (nds32_register_passes): New function to register passes.
9792 * config/nds32/nds32.md (relax_group): New pattern.
9793 * config/nds32/nds32.opt (mrelax-hint): New option.
9794 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9796 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9798 * config/nds32/t-nds32: Modify files dependency.
9800 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9802 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9803 (PROFILE_HOOK): Define its implementation.
9805 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9807 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9808 type and 32-bit size.
9810 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9813 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9814 (V_128_256): New mode iterator.
9815 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9816 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9817 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9819 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9820 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9822 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9825 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9826 NaN inputs correctly.
9828 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9831 * config/rs6000/vsx.md (??r): New mode attribute.
9832 (*vsx_mov<mode>_64bit): Use it.
9833 (*vsx_mov<mode>_32bit): Likewise.
9835 2018-03-30 Martin Sebor <msebor@redhat.com>
9837 PR tree-optimization/84818
9838 * builtins.c (check_access): Use warning_n.
9840 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9843 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9845 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9848 2018-03-30 Julia Koval <julia.koval@intel.com>
9851 * x86-tune.def (movx, partial_reg_dependency): Enable for
9854 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9857 * lra-constraints.c (process_alt_operands): Move setting
9858 this_alternative_matches below.
9860 2018-03-29 Martin Liska <mliska@suse.cz>
9863 * doc/invoke.texi: Document how LTO works with debug info.
9864 Describe auto-load support of binutils. Mention 'x86-64'
9865 as valid option value of -march option.
9867 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9869 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9872 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9873 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9874 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9877 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9880 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9881 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9882 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9883 for RS6000_BTM_POWERPC64.
9884 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9885 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9886 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9891 2018-03-28 Carl Love <cel@us.ibm.com>
9894 2017-09-27 Carl Love <cel@us.ibm.com>
9896 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9897 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9898 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9901 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9903 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9904 instead of __vector bool.
9905 (_mm_max_pu8): Likewise.
9906 (_mm_min_pi16): Likewise.
9908 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9911 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9912 (DIVWEUO): Likewise.
9914 (DIVDEUO): Likewise.
9915 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9916 DIVWEUO and DIVDEUO.
9917 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9918 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9919 (div_extend): Likewise.
9920 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9922 (__builtin_divweuo): Likewise.
9923 (__builtin_divdeo): Likewise.
9924 (__builtin_divdeuo): Likewise.
9926 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9929 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9930 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9932 PR tree-optimization/82004
9933 * gimple-match-head.c (optimize_pow_to_exp): New function.
9934 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9935 Don't fold to exp if optimize_pow_to_exp is false.
9937 2018-03-28 Martin Liska <mliska@suse.cz>
9940 * calls.c (initialize_argument_information): Fix trailing space.
9941 * common.opt: Fix typo and provide better explanation for
9942 -fsanitize-coverage option.
9943 * config/i386/i386.opt: Fix typo.
9945 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9946 Martin Liska <mliska@suse.cz>
9949 * gimplify.c (asan_poison_variable): Don't do the check for
9950 gimplify_omp_ctxp here.
9951 (gimplify_decl_expr): Do it here.
9952 (gimplify_target_expr): Likewise.
9954 2018-03-28 Martin Liska <mliska@suse.cz>
9957 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9958 chkp_type_bounds_count if MPX is not enabled.
9960 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9962 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9964 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9967 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9968 function to create the function decl for complex long double
9969 multiply and divide for -mabi=ieeelongdouble.
9970 (init_float128_ieee): Call it.
9972 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9975 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9976 -fcf-protection=branch -mibt.
9977 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9979 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9982 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9983 arm_word_relocations.
9985 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9988 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9989 extern array declarations.
9991 2018-03-27 Richard Biener <rguenther@suse.de>
9994 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9995 explicit single_use checks.
9997 2018-03-27 Richard Biener <rguenther@suse.de>
9999 PR tree-optimization/85082
10000 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10003 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10005 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10006 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10007 Turn on fasynchronous-unwind-tables and funwind-tables.
10009 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10012 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10013 (*bmi_blsr_<mode>_ccz): Ditto.
10015 2018-03-26 Tom de Vries <tom@codesourcery.com>
10017 PR tree-optimization/85063
10018 * omp-general.c (offloading_function_p): New function. Factor out
10020 * omp-offload.c (pass_omp_target_link::gate): ... here.
10021 * omp-general.h (offloading_function_p): Declare.
10022 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10023 with attribute omp declare target for offloading functions.
10025 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10027 PR tree-optimization/84005
10028 * tree-data-ref.h (get_base_for_alignment): Declare.
10029 * tree-data-ref.c (get_base_for_alignment_1): New function.
10030 (get_base_for_alignment): Likewise.
10031 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10032 get_base_for_alignment to find a suitable base object, instead
10033 of always using drb->base_address.
10035 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10037 PR inline-asm/85022
10038 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10039 known size by default.
10041 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10043 PR inline-asm/85030
10044 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10045 and non BLKmode operands.
10047 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10050 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10051 Clean up attributes.
10053 2018-03-23 Richard Biener <rguenther@suse.de>
10056 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10057 we are going to emit early debug for LTO.
10059 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10061 PR inline-asm/85034
10062 * function.c (match_asm_constraints_1): Don't optimize if input
10063 doesn't satisfy general_operand predicate for output's mode.
10065 PR inline-asm/85022
10066 * alias.c (write_dependence_p): Don't require for x_canonicalized
10067 non-VOIDmode if x has VOIDmode.
10070 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10071 just don't try to optimize it rather than assert it never happens.
10073 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10075 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10076 macro expansions for definition of ST_INTERNAL_<mode> and
10077 LD_INTERNAL_<mode> builtins.
10078 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10080 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10082 (altivec_expand_st_builtin): Likewise.
10083 (altivec_expand_builtin): Remove calls to deleted functions.
10084 (rs6000_address_for_altivec): Delete this function.
10085 * config/rs6000/vector.md: Remove expands for
10086 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10088 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10091 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10092 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10093 re-computing once computed.
10094 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10095 (arm_init_machine_status): Initialize
10096 machine->static_chain_stack_bytes.
10098 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10101 * doc/extend.texi: Add four new prototypes for vec_ld.
10102 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10103 definitions for more logical presentation.
10104 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10105 entries for V1TI variants of __builtin_altivec_ld builtin.
10106 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10107 handling of V1TI variant of LVX icode pattern.
10108 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10109 (rs6000_gimple_fold_builtin): Likewise.
10110 (altivec_init_builtins): Add code to define
10111 __builtin_altivec_lvx_v1ti function.
10113 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10115 PR inline-asm/84941
10116 * function.c (match_asm_constraints_1): Don't do the optimization
10117 if input isn't a REG, SUBREG, MEM or constant.
10119 2018-03-22 Tom de Vries <tom@codesourcery.com>
10121 PR tree-optimization/84956
10122 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10123 bb_has_abnormal_pred.
10125 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10128 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10129 DECL_INITIAL (decl) to decl at the end.
10130 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10131 adjust the comment.
10133 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10135 * doc/extend.texi (__builtin_tgmath): Document when complex
10136 integer types are treated as _Complex _Float64.
10138 2018-03-21 Tom de Vries <tom@codesourcery.com>
10140 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10142 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10144 PR tree-optimization/84960
10145 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10146 if it is ENTRY block, move them into single succ of ENTRY in that case.
10148 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10150 PR tree-optimization/84811
10151 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10152 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10153 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10154 (known_subrange_p): Update accordingly. Cast each value involved
10155 in the size comparison, rather than casting the result of the
10158 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10160 PR tree-optimization/84982
10161 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10162 by flipping the least significant bit rather than all bits from
10163 bitpos to bitpos + bitsize - 1.
10165 2018-03-21 Nathan Sidwell <nathan@acm.org>
10167 * doc/extend.texi (Deprecated Features): Remove mention of
10168 long-deleted deprecations.
10170 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10173 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10174 * configure: Regenerate.
10176 2018-03-21 Tom de Vries <tom@codesourcery.com>
10178 PR tree-optimization/83126
10179 * tree-parloops.c (num_phis): New function.
10180 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10182 2018-03-21 Nathan Sidwell <nathan@acm.org>
10184 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10185 mention anon-struct/union members and trailing attributes.
10187 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10189 PR tree-optimization/84969
10190 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10191 builtin memset partitions if they set different rhs values.
10193 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10195 PR rtl-optimization/84989
10196 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10197 VEC_DUPLICATE with scalar result mode.
10199 2018-03-21 Martin Liska <mliska@suse.cz>
10202 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10203 not intended return statement.
10205 2018-03-21 Martin Liska <mliska@suse.cz>
10208 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10209 (chkp_find_bound_slots_1): Limit number of iterations.
10211 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10214 * Minor grammar fixes for x86 options.
10216 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10219 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10220 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10222 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10225 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10226 (altivec_lvx_<mode>_1op): Likewise.
10227 (altivec_stvx_<mode>_2op): Likewise.
10228 (altivec_stvx_<mode>_1op): Likewise.
10229 (altivec_lvx_<VM2:mode>): New define_expand.
10230 (altivec_stvx_<VM2:mode>): Likewise.
10231 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10232 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10233 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10234 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10235 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10236 (rs6000_gen_lvx): Likewise.
10237 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10238 (altivec_expand_stv_builtin): Likewise.
10239 (altivec_expand_builtin): Likewise.
10240 * config/rs6000/vector.md: Likewise.
10242 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10245 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10248 2018-03-20 Richard Biener <rguenther@suse.de>
10251 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10252 sign-conversions as zero, fall back to standard scalar_stmt
10255 2018-03-20 Martin Liska <mliska@suse.cz>
10258 * predict.c (rebuild_frequencies): Handle case when we have
10259 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10261 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10264 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10265 flag_section_anchors.
10266 * varasm.c (use_blocks_for_decl_p): Remove hack for
10267 dw2_force_const_mem.
10270 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10272 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10273 be created, use lowpart_subreg of operands[0] rather than operands[0]
10275 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10276 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10277 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10278 and n constraint instead of aarch64_shift_imm_di and Usd.
10279 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10280 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10282 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10285 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10286 to favor GPR over NEON registers.
10287 (<shift>di3_neon): Likewise.
10289 2018-03-20 Tom de Vries <tom@codesourcery.com>
10292 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10293 (nvptx_process_pars): Emit bar.sync asap and alap.
10295 2018-03-20 Tom de Vries <tom@codesourcery.com>
10298 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10299 seen_label if seen_label is already set.
10301 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10304 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10305 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10306 Use 1U instead of 1. Formatting fixes.
10309 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10310 instead of TREE_TYPE (s1) for the return value.
10312 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10314 PR tree-optimization/84946
10315 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10316 bitsize + bitsize in poly_uint64 rather than poly_int64.
10319 * dwarf2asm.c: Include fold-const.c.
10320 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10321 of decl rather than decl itself.
10323 PR rtl-optimization/84643
10324 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10326 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10329 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10330 calling assemble_variable.
10332 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10335 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10336 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10338 2018-03-19 Jim Wilson <jimw@sifive.com>
10341 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10342 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10343 (riscv_first_stack_step): Likewise.
10344 (riscv_option_override): Use STACK_BOUNDARY instead of
10345 MIN_STACK_BOUNDARY.
10346 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10347 MIN_STACK_BOUNDARY.
10348 (BIGGEST_ALIGNMENT): Set to 128.
10349 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10350 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10353 2018-03-19 Richard Biener <rguenther@suse.de>
10355 PR tree-optimization/84933
10356 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10357 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10359 2018-03-19 Richard Biener <rguenther@suse.de>
10361 PR tree-optimization/84859
10362 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10363 (cond_if_else_store_replacement): Perform sinking operation on
10364 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10365 Generalize what a BB with a single eligible store is.
10367 2018-03-19 Richard Biener <rguenther@suse.de>
10369 PR tree-optimization/84929
10370 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10371 chrec_is_positive against non-chrec arg.
10373 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10376 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10378 2018-03-18 Martin Liska <mliska@suse.cz>
10380 PR rtl-optimization/84635
10381 * regrename.c (build_def_use): Use matches_mode only when
10384 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10386 PR tree-optimization/84913
10387 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10388 vectorize chains of COND_EXPRs.
10390 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10392 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10394 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10396 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10398 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10400 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10402 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10403 Kito Cheng <kito.cheng@gmail.com>
10405 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10406 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10407 (nds32_adjust_reg_alloc_order): New function.
10408 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10410 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10412 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10413 nds32_print_operand, nds32_print_operand_address): Use
10414 HOST_WIDE_INT_PRINT_DEC instead.
10416 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10418 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10420 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10423 * config/i386/i386.c (initial_ix86_tune_features,
10424 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10425 unsigned long long.
10426 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10427 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10428 rather than 1u << ix86_tune. Formatting fix.
10429 (ix86_option_override_internal): Change ix86_arch_mask from
10430 unsigned int to unsigned HOST_WIDE_INT, initialize to
10431 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10432 (ix86_function_specific_restore): Likewise.
10434 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10437 * postreload.c (reload_combine_recognize_pattern): Perform
10438 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10439 truncate_int_for_mode the result for the destination's mode.
10442 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10443 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10445 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10448 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10451 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10452 regno_allocno_class_array and sorted_pseudos.
10453 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10454 insns where regno is used.
10456 2018-03-16 Martin Liska <mliska@suse.cz>
10459 * multiple_target.c (create_dispatcher_calls): Redirect
10460 reference in the symbol table.
10462 2018-03-16 Martin Liska <mliska@suse.cz>
10465 * multiple_target.c (create_dispatcher_calls): Redirect also
10468 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10472 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10473 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10474 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10476 2018-03-16 Julia Koval <julia.koval@intel.com>
10478 * doc/invoke.texi (Skylake Server): Add CLWB.
10479 Cannonlake): Remove CLWB.
10481 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10483 PR tree-optimization/84841
10484 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10486 (FLOAT_ONE_CONST_TYPE): Define.
10487 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10488 (sort_by_operand_rank): Put entries with higher constant_type last
10489 rather than first to match comments.
10491 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10493 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10496 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10499 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10500 check and use address of target always.
10502 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10505 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10506 (indirect_thunk_bnd_needed): Likewise.
10507 (indirect_thunks_used): Likewise.
10508 (indirect_thunks_bnd_used): Likewise.
10509 (indirect_return_needed): New.
10510 (indirect_return_bnd_needed): Likewise.
10511 (output_indirect_thunk_function): Add a bool argument for
10513 (output_indirect_thunk_function): Don't generate alias for
10514 function return thunk.
10515 (ix86_code_end): Call output_indirect_thunk_function to generate
10516 function return thunks.
10517 (ix86_output_function_return): Set indirect_return_bnd_needed
10518 and indirect_return_needed instead of indirect_thunk_bnd_needed
10519 and indirect_thunk_needed.
10521 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10523 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10524 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10525 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10527 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10528 Paul Hua <paul.hua.gm@gmail.com>
10531 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10533 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10535 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10536 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10537 resp. SFmode cases.
10539 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10542 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10543 instead of GET_MODE_SIZE when comparing Units.
10545 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10548 * varasm.c (hash_section): Return an unchangeble hash value
10549 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10550 Return !aarch64_can_use_per_function_literal_pools_p ().
10552 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10555 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10556 rather than address of cmode as last argument to prepare_cmp_insn.
10558 2018-03-15 Julia Koval <julia.koval@intel.com>
10560 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10561 F_AVX512VNNI, F_AVX512BITALG): New.
10563 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10566 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10567 insn for floating-point loads and stores.
10569 2018-03-14 Carl Love <cel@us.ibm.com>
10571 * config/rs6000/rs6000-c.c: Add macro definitions for
10572 ALTIVEC_BUILTIN_VEC_PERMXOR.
10573 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10574 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10575 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10576 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10578 * config/doc/extend.texi: Add prototypes for vec_permxor.
10580 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10583 * diagnostic-show-locus.c (class layout_point): Convert m_line
10584 from int to linenum_type.
10585 (line_span::comparator): Use linenum "compare" function when
10586 comparing line numbers.
10587 (test_line_span): New function.
10588 (layout_range::contains_point): Convert param "row" from int to
10590 (layout_range::intersects_line_p): Likewise.
10591 (layout::will_show_line_p): Likewise.
10592 (layout::print_source_line): Likewise.
10593 (layout::should_print_annotation_line_p): Likewise.
10594 (layout::print_annotation_line): Likewise.
10595 (layout::print_leading_fixits): Likewise.
10596 (layout::annotation_line_showed_range_p): Likewise.
10597 (struct line_corrections): Likewise for field m_row.
10598 (line_corrections::line_corrections): Likewise for param "row".
10599 (layout::print_trailing_fixits): Likewise.
10600 (layout::get_state_at_point): Likewise.
10601 (layout::get_x_bound_for_row): Likewise.
10602 (layout::print_line): Likewise.
10603 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10604 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10605 * input.c (selftest::test_linenum_comparisons): New function.
10606 (selftest::input_c_tests): Call it.
10607 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10608 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10609 * selftest.h (ASSERT_GT): New macro.
10610 (ASSERT_GT_AT): New macro.
10611 (ASSERT_LT): New macro.
10612 (ASSERT_LT_AT): New macro.
10614 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10616 PR rtl-optimization/84780
10617 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10619 2018-03-14 Martin Liska <mliska@suse.cz>
10621 * tree.c (record_node_allocation_statistics): Use
10622 get_stats_node_kind.
10623 (get_stats_node_kind): New function extracted from
10624 record_node_allocation_statistics.
10625 (free_node): Use get_stats_node_kind.
10627 2018-03-14 Richard Biener <rguenther@suse.de>
10629 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10630 that the value-set of ANTIC_IN doesn't grow.
10633 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10635 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10636 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10638 2018-03-14 Julia Koval <julia.koval@intel.com>
10640 * config.gcc (icelake-client, icelake-server): New.
10642 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10643 (initial_ix86_arch_features): Ditto.
10644 (PTA_SKYLAKE): Add SGX.
10645 (PTA_ICELAKE): Remove.
10646 (PTA_ICELAKE_CLIENT): New.
10647 (PTA_ICELAKE_SERVER): New.
10648 (ix86_option_override_internal): Split up icelake on icelake client and
10650 (get_builtin_code_for_version): Ditto.
10651 (fold_builtin_cpu): Ditto.
10652 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10653 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10654 * config/i386/i386.h (processor_type): Ditto.
10655 * doc/invoke.texi: Ditto.
10657 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10660 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10661 INTEGER_CST offset, add it together with bitpos / 8 and
10662 sign extend based on POINTER_SIZE.
10666 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10669 * config/i386/constraints.md (Yc): New register constraint.
10670 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10671 Use Yc constraint for alternative 2 of operand 0. Remove
10672 preferred_for_speed attribute.
10674 2018-03-14 Richard Biener <rguenther@suse.de>
10676 PR tree-optimization/84830
10677 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10678 with the old one to avoid oscillations.
10680 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10683 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10685 (assign_by_spills): Return a flag of reload assignment failure.
10686 Do not process the reload assignment failures. Do not spill other
10687 reload pseudos if they has the same reg class. Update n if
10689 (lra_assign): Add a return arg. Set up from the result of
10690 assign_by_spills call.
10691 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10692 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10693 usage_insns if it is not NULL.
10694 (spill_hard_reg_in_range): New function.
10695 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10696 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10697 function prototypes.
10698 (lra_assign): Change prototype.
10699 * lra.c (lra): Add code to deal with fails by splitting hard reg
10702 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10704 * config/riscv/riscv.opt (mrelax): New option.
10705 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10706 "norelax" when riscv_mrelax is disabled.
10707 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10709 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10712 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10713 reassociation for int modes.
10715 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10717 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10718 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10720 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10721 * config/aarch64/aarch64-sve.md
10722 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10723 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10724 (*extend<mode><Vwide>2): Rename to...
10725 (aarch64_sve_extend<mode><Vwide>2): ...this.
10726 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10727 renaming the old pattern to...
10728 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10730 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10731 define_expand, renaming the old pattern to...
10732 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10733 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10734 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10735 account when deciding which SVE instruction the optab should use.
10736 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10738 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10740 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10741 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10742 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10743 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10744 (tlsdesc_small_<mode>): Turn a define_expand and use
10745 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10746 (tlsdesc_small_advsimd_<mode>): ...this.
10747 (tlsdesc_small_sve_<mode>): New pattern.
10749 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10751 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10752 (UNSPEC_UMUL_HIGHPART): New constants.
10753 (MUL_HIGHPART): New int iteraor.
10754 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10755 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10757 (*<su>mul<mode>3_highpart): New define_insn.
10759 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10762 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10765 2018-03-13 Martin Liska <mliska@suse.cz>
10768 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10770 (sem_item_optimizer::~sem_item_optimizer): Release it.
10771 (sem_item_optimizer::merge_classes): Register variable aliases.
10772 (sem_item_optimizer::fixup_pt_set): New function.
10773 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10774 * ipa-icf.h: Declare new variables and functions.
10776 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10778 PR middle-end/84834
10779 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10780 integer_pow2p@2 and test integer_pow2p in condition.
10781 (A < 0 ? C : 0): Similarly for @1.
10783 PR middle-end/84831
10784 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10785 characters starting at p contain '\0' character, don't look beyond
10789 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10790 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10793 * reg-stack.c (change_stack): Change update_end var from int to
10794 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10795 also call set_block_for_insn on the newly added insns and rescan.
10798 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10799 on the last operand.
10802 * tree.c (stabilize_reference_1): Return save_expr (e) for
10803 STATEMENT_LIST even if it doesn't have side-effects.
10805 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10807 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10809 2018-03-12 Renlin Li <renlin.li@arm.com>
10811 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10812 aarch64_output_scalar_simd_mov_immediate.
10814 2018-03-12 Martin Sebor <msebor@redhat.com>
10816 PR tree-optimization/83456
10817 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10818 for perfectly overlapping calls to memcpy.
10819 (gimple_fold_builtin_memory_chk): Same.
10820 (gimple_fold_builtin_strcpy): Handle no-warning.
10821 (gimple_fold_builtin_stxcpy_chk): Same.
10822 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10824 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10826 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10827 parameter. Use it for SFmode.
10828 (rs6000_function_arg_advance_1): Adjust.
10829 (rs6000_function_arg): Adjust.
10830 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10832 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10834 PR rtl-optimization/84169
10835 PR rtl-optimization/84780
10836 * combine.c (can_combine_p): Check for a 2-insn combination whether
10837 the destination register is used between the two insns, too.
10839 2018-03-12 Richard Biener <rguenther@suse.de>
10841 PR tree-optimization/84803
10842 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10843 for refs DR analysis didn't process.
10845 2018-03-12 Richard Biener <rguenther@suse.de>
10847 PR tree-optimization/84777
10848 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10849 force-vectorize loops ignore whether we are optimizing for size.
10851 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10853 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10854 (TARGET_MD_ASM_ADJUST): Define.
10856 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10857 Kito Cheng <kito.cheng@gmail.com>
10858 Chung-Ju Wu <jasonwucj@gmail.com>
10860 * config/nds32/nds32.c (nds32_compute_stack_frame,
10861 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10862 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10863 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10864 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10865 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10866 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10867 * config/nds32/nds32.md (prologue, epilogue): Use macro
10868 NDS32_V3PUSH_AVAILABLE_P to do checking.
10870 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10873 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10874 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10875 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10876 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10877 addition of most attributes on !orig_type_die or the attribute not
10878 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10880 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10881 Chung-Ju Wu <jasonwucj@gmail.com>
10883 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10884 __NDS32_VH__ macro.
10885 * config/nds32/nds32.opt (mvh): New option.
10887 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10888 Chung-Ju Wu <jasonwucj@gmail.com>
10890 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10892 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10893 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10896 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10897 Chung-Ju Wu <jasonwucj@gmail.com>
10899 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10901 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10902 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10904 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10905 Kito Cheng <kito.cheng@gmail.com>
10906 Chung-Ju Wu <jasonwucj@gmail.com>
10908 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10909 UNSPEC_FFMISM and UNSPEC_FLMISM.
10910 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10911 for ffb, ffmism and flmism.
10912 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10913 (unspec_ffmism): Ditto.
10914 (unspec_flmism): Ditto.
10915 (nds32_expand_builtin_impl): Check if string extension is available.
10916 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10917 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10919 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10922 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10925 * lra-assigns.c (assign_by_spills): Return a flag of reload
10926 assignment failure. Do not process the reload assignment
10927 failures. Do not spill other reload pseudos if they has the same
10929 (lra_assign): Add a return arg. Set up from the result of
10930 assign_by_spills call.
10931 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10932 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10933 usage_insns if it is not NULL.
10934 (spill_hard_reg_in_range): New function.
10935 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10936 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10937 function prototypes.
10938 (lra_assign): Change prototype.
10939 * lra.c (lra): Add code to deal with fails by splitting hard reg
10942 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10945 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10947 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10950 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10951 (dw_val_node): Add val_symbolic_view.
10952 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10953 (symview_upper_bound): New.
10954 (new_line_info_table): Initialize symviews_since_reset.
10955 (dwarf2out_source_line): Count symviews_since_reset and set
10956 symview_upper_bound.
10957 (dw_val_equal_p): Handle symview.
10958 (add_AT_symview): New.
10959 (print_dw_val): Handle symview.
10960 (attr_checksum, attr_checksum_ordered): Likewise.
10961 (same_dw_val_p, size_of_die): Likewise.
10962 (value_format, output_die): Likewise.
10963 (add_high_low_attributes): Use add_AT_symview for entry_view.
10964 (dwarf2out_finish): Reset symview_upper_bound, clear
10967 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10970 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10971 Add strict argument and use it.
10972 (rs6000_split_multireg_move): Update for new strict argument.
10973 (mem_operand_gpr): Disallow all non-offsettable addresses.
10974 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10976 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10979 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10980 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10981 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10984 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10985 decl, use remap_type if we want to use the type.
10987 2018-03-09 Martin Sebor <msebor@redhat.com>
10989 PR tree-optimization/84526
10990 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10992 (builtin_access::generic_overlap): Be prepared to handle non-array
10995 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10997 PR rtl-optimization/84682
10998 * lra-constraints.c (process_address_1): Check is_address flag
10999 for address constraints.
11000 (process_alt_operands): Likewise.
11001 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11002 preprocess_constraints.
11003 * recog.h (preprocess_constraints): Add oploc parameter.
11005 * recog.c (preprocess_constraints): Test address_operand for
11006 CT_ADDRESS constraints.
11008 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11011 * lra-assigns.c (assign_by_spills): Return a flag of reload
11012 assignment failure. Do not process the reload assignment
11013 failures. Do not spill other reload pseudos if they has the same
11015 (lra_assign): Add a return arg. Set up from the result of
11016 assign_by_spills call.
11017 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11018 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11019 usage_insns if it is not NULL.
11020 (spill_hard_reg_in_range): New function.
11021 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11022 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11023 function prototypes.
11024 (lra_assign): Change prototype.
11025 * lra.c (lra): Add code to deal with fails by splitting hard reg
11028 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11031 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11032 Accept complain bool parameter. Only emit errors if it is true.
11033 (arm_parse_cpu_option_name): Likewise.
11034 (arm_target_thumb_only): Adjust callers of the above.
11035 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11036 prototype to take a default true bool parameter.
11037 (arm_parse_arch_option_name): Likewise.
11039 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11040 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11044 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11045 * configure: Regenerate.
11046 * configure.ac ("linker --version-script option"): New.
11047 ("linker soname option"): New.
11049 2018-03-09 Richard Biener <rguenther@suse.de>
11051 PR tree-optimization/84775
11052 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11053 immediate uses of predicate stmts and mark them modified.
11056 PR tree-optimization/84178
11057 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11059 (version_loop_for_if_conversion): Delay update_ssa call.
11060 (tree_if_conversion): Delay update_ssa until after predicate
11063 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11066 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11067 when the function accesses prior frames.
11069 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11072 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11073 gen_llsym, otherwise call maybe_gen_llsym.
11075 PR inline-asm/84742
11076 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11077 has ',' character inside of it.
11079 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11082 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11083 as clobbering CC_REGNUM.
11085 2018-03-08 Richard Biener <rguenther@suse.de>
11087 PR middle-end/84552
11088 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11089 (follow_copies_to_constant): Do not follow SSA names registered
11092 2018-03-08 Richard Biener <rguenther@suse.de>
11094 PR tree-optimization/84178
11095 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11097 (version_loop_for_if_conversion): Delay update_ssa call.
11098 (tree_if_conversion): Delay update_ssa until after predicate
11101 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11103 PR tree-optimization/84178
11104 * tree-if-conv.c (release_bb_predicate): Remove the
11105 the assertion that the stmts have NULL use_ops.
11106 Discard the statements, asserting that they haven't
11107 yet been added to a BB.
11109 2018-03-08 Richard Biener <rguenther@suse.de>
11111 PR tree-optimization/84746
11112 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11113 (phi_translate): Pass in destination ANTIC_OUT set.
11114 (phi_translate_1): Likewise. For a simplified result lookup
11115 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11116 (phi_translate_set): Adjust.
11117 (do_pre_regular_insertion): Likewise.
11118 (do_pre_partial_partial_insertion): Likewise.
11120 2018-03-08 Martin Liska <mliska@suse.cz>
11122 PR gcov-profile/84735
11123 * doc/gcov.texi: Document usage of profile files.
11124 * gcov-io.h: Document changes in the format.
11126 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11130 * dwarf2out.c (struct dw_line_info_table): Update comments for
11132 (FORCE_RESET_NEXT_VIEW): New.
11133 (FORCE_RESETTING_VIEW_P): New.
11134 (RESETTING_VIEW_P): Check for -1 too.
11135 (ZERO_VIEW_P): Likewise.
11136 (new_line_info_table): Force-reset next view.
11137 (dwarf2out_begin_function): Likewise.
11138 (dwarf2out_source_line): Simplify zero_view_p initialization.
11139 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11140 view directly. Omit view when omitting .loc at line 0.
11142 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11144 PR tree-optimization/84740
11145 * tree-switch-conversion.c (process_switch): Call build_constructors
11146 only if info.phi_count is non-zero.
11148 PR tree-optimization/84739
11149 * tree-tailcall.c (find_tail_calls): Check call arguments against
11150 DECL_ARGUMENTS (current_function_decl) rather than
11151 DECL_ARGUMENTS (func) when checking for tail recursion.
11153 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11155 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11156 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11157 Volker Reichelt's entry and add entries for people that perform
11158 GCC fuzzy testing and report numerous bugs.
11160 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11163 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11164 readonly data in sdata, if that is disabled.
11165 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11166 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11167 -mreadonly-in-sdata option.
11169 2018-03-07 Martin Sebor <msebor@redhat.com>
11171 PR tree-optimization/84468
11172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11173 basic block when looking for nul assignment.
11175 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11178 * except.h (output_function_exception_table): Adjust prototype.
11179 * except.c (output_function_exception_table): Remove FNNAME parameter
11180 and add SECTION parameter. Ouput one part of the table at a time.
11181 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11182 the first part of the exception table and emit unwind directives.
11183 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11184 (i386_pe_seh_cold_init): Likewise.
11185 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11186 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11187 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11188 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11189 * config/i386/winnt.c: Include except.h.
11190 (struct seh_frame_state): Add reg_offset, after_prologue and
11191 in_cold_section fields.
11192 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11193 (i386_pe_seh_cold_init): New function.
11194 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11195 to seh->in_cold_section.
11196 (seh_emit_push): Record the offset of the push.
11197 (seh_emit_save): Record the offet of the save.
11198 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11199 Test seh->after_prologue to disregard the epilogue.
11200 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11201 (i386_pe_end_cold_function): New function.
11203 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11206 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11207 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11210 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11212 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11213 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11215 PR middle-end/84723
11216 * multiple_target.c: Include tree-inline.h and intl.h.
11217 (expand_target_clones): Diagnose and fail if node->definition and
11218 !tree_versionable_function_p (node->decl).
11220 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11222 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11224 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11225 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11226 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11228 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11231 * combine.c (try_combine): Use reg_or_subregno instead of handling
11232 just paradoxical SUBREGs and REGs.
11234 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11236 * config/arc/arc.c (arc_finalize_pic): Remove function.
11237 (arc_must_save_register): We use single base PIC register, remove
11238 checks to save/restore the PIC register.
11239 (arc_expand_prologue): Likewise.
11240 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11242 (arc_verify_short): Likewise.
11243 (arc_attr_type): Likewise.
11244 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11245 (walk_stores): Likewise.
11246 (arc_address_cost): Make it static.
11247 (arc_verify_short): Likewise.
11248 (branch_dest): Likewise.
11249 (arc_attr_type): Likewise.
11250 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11251 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11252 (arc_final_prescan_insn): Remove inserting the nops due to
11253 hardware hazards. It is done in reorg step.
11254 (insn_length_variant_t): Remove.
11255 (insn_length_parameters_t): Likewise.
11256 (arc_insn_length_parameters): Likewise.
11257 (arc_get_insn_variants): Likewise.
11258 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11260 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11262 PR inline-asm/84683
11263 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11266 PR tree-optimization/84687
11267 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11269 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11271 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11273 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11274 Rename to ppc_speculation_barrier.
11275 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11276 __builtin_ppc_speculation_barrier.
11278 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11281 * combine.c (combine_simplify_rtx): Don't try to simplify if
11282 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11285 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11287 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11288 to 32 bytes when compiling for POWER9.
11290 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11293 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11294 regparm >= 3 with no arg reg available also for calls with
11295 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11298 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11300 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11302 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11305 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11307 2018-03-05 Richard Biener <rguenther@suse.de>
11309 PR tree-optimization/84486
11310 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11311 When inserting a __builtin_assume_aligned call set the LHS
11312 SSA name alignment info accordingly.
11314 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11316 PR tree-optimization/84114
11317 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11318 Avoid reassociation of FLOAT_MODE addition.
11320 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11322 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11323 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11324 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11325 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11326 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11327 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11328 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11330 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11331 __builtin_ia32_wbinvd): New builtins.
11332 (SPECIAL_ARGS2): New.
11333 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11334 (SPECIAL_ARGS2): New.
11335 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11336 (ix86_valid_target_attribute_inner_p): Ditto.
11337 (ix86_init_mmx_sse_builtins): Add special_args2.
11338 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11339 TARGET_WBNOINVD_P): New.
11340 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11341 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11342 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11343 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11344 * config/i386/pconfigintrin.h: New file.
11345 * config/i386/wbnoinvdintrin.h: Ditto.
11346 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11348 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11350 2018-03-05 Richard Biener <rguenther@suse.de>
11352 PR tree-optimization/84670
11353 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11355 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11356 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11357 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11358 doesn't grow if all successors (recursively) were visited at least
11361 2018-03-05 Richard Biener <rguenther@suse.de>
11363 PR tree-optimization/84650
11364 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11365 if executed in the loop pipeline.
11367 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11369 * doc/configfiles.texi (Configuration Files): Move info about
11370 conditionalizing $target-protos.h to...
11371 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11372 differs from $target-protos.h.
11374 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11375 Chung-Ju Wu <jasonwucj@gmail.com>
11377 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11378 * config/nds32/nds32-multiple.md (setmemsi): Define.
11379 * config/nds32/nds32-memory-manipulation.c
11380 (nds32_gen_dup_4_byte_to_word_value): New.
11381 (emit_setmem_word_loop): New.
11382 (emit_setmem_byte_loop): New.
11383 (nds32_expand_setmem_loop): New.
11384 (nds32_expand_setmem_loop_v3m): New.
11385 (nds32_expand_setmem_unroll): New.
11386 (nds32_expand_setmem): New.
11388 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11389 Chung-Ju Wu <jasonwucj@gmail.com>
11391 * config/nds32/nds32-memory-manipulation.c
11392 (nds32_emit_load_store): New.
11393 (nds32_emit_post_inc_load_store): New.
11394 (nds32_emit_mem_move): New.
11395 (nds32_emit_mem_move_block): New.
11396 (nds32_expand_movmemsi_loop_unknown_size): New.
11397 (nds32_expand_movmemsi_loop_known_size): New.
11398 (nds32_expand_movmemsi_loop): New.
11399 (nds32_expand_movmemsi_unroll): New.
11400 (nds32_expand_movmemqi): Rename ...
11401 (nds32_expand_movmemsi): ... to this.
11402 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11403 (movmemsi): ... to this.
11404 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11405 (nds32_expand_movmemsi): ... to this.
11407 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11408 Monk Chiang <sh.chiang04@gmail.com>
11409 Chung-Ju Wu <jasonwucj@gmail.com>
11411 * config/nds32/nds32-protos.h
11412 (nds32_expand_load_multiple): New arguments.
11413 (nds32_expand_store_multiple): Ditto.
11414 (nds32_valid_multiple_load_store): Rename ...
11415 (nds32_valid_multiple_load_store_p): ... to this.
11416 * config/nds32/nds32-memory-manipulation.c
11417 (nds32_expand_load_multiple): Refine implementation.
11418 (nds32_expand_store_multiple): Ditto.
11419 * config/nds32/nds32-multiple.md
11420 (load_multiple): Update nds32_expand_load_multiple interface.
11421 (store_multiple): Update nds32_expand_store_multiple interface.
11422 * config/nds32/nds32-predicates.c
11423 (nds32_valid_multiple_load_store): Rename ...
11424 (nds32_valid_multiple_load_store_p): ... to this and refine
11426 * config/nds32/predicates.md
11427 (nds32_load_multiple_and_update_address_operation): New predicate.
11428 (nds32_store_multiple_and_update_address_operation): New predicate.
11430 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11431 Chung-Ju Wu <jasonwucj@gmail.com>
11433 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11434 (combo): New attribute.
11435 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11437 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11439 * config/nds32/nds32.opt: Change -mcmodel= default value.
11441 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11442 Monk Chiang <sh.chiang04@gmail.com>
11443 Chung-Ju Wu <jasonwucj@gmail.com>
11445 * config/nds32/constants.md (unspec_element): New enum.
11446 * config/nds32/constraints.md (Umw): New constraint.
11447 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11448 * config/nds32/nds32-intrinsic.md: Likewise.
11449 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11450 (nds32_valid_smw_lwm_base_p): New.
11451 (nds32_output_smw_single_word): New.
11452 (nds32_output_lmw_single_word): New.
11453 (nds32_expand_unaligned_load): New.
11454 (nds32_expand_unaligned_store): New.
11455 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11456 (nds32_output_smw_single_word): Declare.
11457 (nds32_output_lmw_single_word): Declare.
11458 (nds32_expand_unaligned_load): Declare.
11459 (nds32_expand_unaligned_store): Declare.
11460 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11461 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11462 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11463 NDS32_BUILTIN_UASTORE_DW.
11464 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11467 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11468 Kito Cheng <kito.cheng@gmail.com>
11469 Chung-Ju Wu <jasonwucj@gmail.com>
11471 * config/nds32/nds32-intrinsic.c
11472 (nds32_expand_builtin_null_ftype_reg): Delete.
11473 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11474 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11475 (nds32_read_argument): New.
11476 (nds32_legitimize_target): Ditto.
11477 (nds32_legitimize_argument): Ditto.
11478 (nds32_check_constant_argument): Ditto.
11479 (nds32_expand_unop_builtin): Ditto.
11480 (nds32_expand_unopimm_builtin): Ditto.
11481 (nds32_expand_binop_builtin): Ditto.
11482 (nds32_builtin_decl_impl): Ditto.
11483 (builtin_description): Ditto.
11484 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11485 (nds32_init_builtins_impl): Ditto.
11486 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11487 (nds32_builtin_decl): New.
11488 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11489 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11491 2018-03-02 Jeff Law <law@redhat.com>
11493 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11494 (redundant_insn, fill_simple_delay_slots): Likewise.
11495 (fill_slots_from_thread): Likewise.
11496 * resource.c (mark_referenced_resources): Likewise.
11497 (mark_set_resources, find_dead_or_set_registers): Likewise.
11499 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11501 * substring-locations.h (format_warning_va): Formatting fix for
11502 ATTRIBUTE_GCC_DIAG.
11503 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11505 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11506 * substring-locations.c: Include intl.h.
11507 (format_warning_va): Turned into small wrapper around
11508 format_warning_n_va, renamed to ...
11509 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11510 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11512 (format_warning_at_substring_n): New function.
11513 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11514 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11515 format_warning_at_substring with just a shorter name instead of
11516 const function pointer.
11517 (fmtwarn_n): New function.
11518 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11519 appropriate, get rid of all the fmtstr temporaries, move conditionals
11520 with G_() wrapped string literals directly into fmtwarn arguments,
11521 cast dir.len to (int), formatting fixes.
11523 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11525 * doc/invoke.texi: Remove "Cilk Plus" references.
11527 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11528 Richard Biener <rguenther@suse.de>
11531 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11532 for error or warning attributes if CALL_FROM_THUNK_P is set.
11535 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11538 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11539 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11542 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11543 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11545 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11546 instead of -1U in last predictors element's probability member.
11548 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11551 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11552 arguments if they are comparable.
11554 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11556 PR tree-optimization/84634
11557 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11558 masks and masked_loop_p with a single loop_masks, making sure it's
11559 null for bb vectorization.
11561 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11563 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11564 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11565 loop->force_vectorize to check whether there is no alias.
11567 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11570 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11572 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11574 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11575 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11576 instead of a loop around prev_real_insn.
11577 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11580 PR inline-asm/84625
11581 * config/i386/i386.c (ix86_print_operand): Use conditional
11582 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11585 2018-03-02 Richard Biener <rguenther@suse.de>
11587 PR tree-optimization/84427
11588 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11589 (bitmap_set_subtract_values): Rewrite to handle multiple
11592 (prune_clobbered_mems): Likewise.
11593 (phi_translate): Take edge instead of pred/phiblock.
11594 (phi_translate_1): Likewise.
11595 (phi_translate_set): Likewise. Insert all translated
11596 exprs for a value into the set, keeping possibly multiple
11597 expressions per value.
11598 (compute_antic_aux): Adjust for phi_translate changes.
11599 When intersecting union the expressions and prune those
11600 not in the final value set, keeping possibly multiple
11601 expressions per value. Do not use value-insertion
11602 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11603 all expressions. Add verification that the value-sets
11604 only shrink during iteration.
11605 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11606 (do_pre_regular_insertion): Likewise.
11607 (do_pre_partial_partial_insertion): Likewise.
11609 2018-03-02 Richard Biener <rguenther@suse.de>
11612 * config/darwin.c (saved_debug_info_level): New static global.
11613 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11614 (darwin_asm_lto_end): Restore debug info generation settings.
11616 2018-03-01 Martin Liska <mliska@suse.cz>
11619 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11620 volatile arguments.
11622 2018-03-01 Richard Biener <rguenther@suse.de>
11625 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11626 type annotation with LTO when debug was disabled at compile-time.
11628 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11630 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11632 (mips_final_postscan_insn): Likewise.
11634 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11636 PR rtl-optimization/84528
11637 * alias.c (init_alias_target): Add commentary.
11638 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11639 a unique base value if the frame pointer is not eliminated
11640 to the stack pointer.
11642 2018-03-01 Tom de Vries <tom@codesourcery.com>
11644 PR rtl-optimization/83327
11645 * lra-int.h (hard_regs_spilled_into): Declare.
11646 * lra.c (hard_regs_spilled_into): Define.
11647 (init_reg_info): Init hard_regs_spilled_into.
11648 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11649 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11650 (process_bb_lives): Handle hard_regs_spilled_into.
11651 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11652 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11654 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11656 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11657 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11658 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11659 * config/rs6000/aix72.h: New file.
11661 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11663 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11664 instead of warning_at with conditional singular and plural messages
11668 * stor-layout.c (update_alignment_for_field): For
11669 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11670 && !DECL_PACKED (field), do the alignment update, just use
11671 only desired_align instead of MAX (type_align, desired_align)
11673 (place_field): Don't do known_align < desired_align handling
11674 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11675 is non-NULL, instead do it after rli->prev_field handling and
11676 only if not within a bitfield word. For DECL_PACKED (field)
11677 use type_align of BITS_PER_UNIT.
11679 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11681 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11682 superfluous parentheses and trailing spaces.
11684 2018-02-28 Richard Biener <rguenther@suse.de>
11686 PR tree-optimization/84584
11687 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11688 SCoPs with fake exit edge.
11690 2018-02-28 Martin Liska <mliska@suse.cz>
11693 * timevar.c (timer::print): Fix format to properly print 100%
11696 2018-02-28 Richard Biener <rguenther@suse.de>
11698 PR middle-end/84607
11699 * genmatch.c (capture_info::walk_match): Do not mark
11700 captured expressions without operands as expr_p given
11701 they act more like predicates and should be subject to
11702 "lost tail" side-effect preserving.
11704 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11706 PR rtl-optimization/81611
11707 * auto-inc-dec.c (attempt_change): Move dead note from
11708 mem_insn if it's the next use of regno
11709 (find_address): Take address use of reg holding
11710 non-incremented value. Add parm to limit search to the named
11712 (merge_in_block): Attempt to use a mem insn that is the next
11713 use of the original regno.
11715 2018-02-27 Martin Sebor <msebor@redhat.com>
11718 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11719 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11721 2018-02-27 Martin Sebor <msebor@redhat.com>
11723 PR translation/84207
11724 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11725 n argument to unsigned HOST_WIDE_INT.
11726 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11727 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11728 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11729 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11731 2018-02-27 Richard Biener <rguenther@suse.de>
11733 PR tree-optimization/84512
11734 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11735 Do not use the estimate returned from record_stmt_cost for
11736 the scalar iteration cost but sum properly using add_stmt_cost.
11738 2018-02-27 Richard Biener <rguenther@suse.de>
11740 PR tree-optimization/84466
11741 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11742 Adjust last change to less strictly validate use operands.
11744 2018-02-27 Martin Liska <mliska@suse.cz>
11746 PR gcov-profile/84548
11747 * gcov.c (process_file): Allow partial overlap and consider it
11748 also as group functions.
11749 (output_lines): Properly calculate range of lines for a group.
11751 2018-02-27 Martin Liska <mliska@suse.cz>
11753 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11754 'ggc' suffixes. Change first column width.
11755 (timer::print): Fix formatting of the column.
11757 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11759 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11760 preserve inline entry blocks for the sake of debug inline
11761 entry point markers alone.
11762 (remove_unused_locals): Suggest in comments a better place to
11763 force the preservation of inline entry blocks that are
11764 otherwise unused, but do not preserve them.
11766 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11768 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11770 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11773 * config/i386/constraints.md (Bs): Replace
11774 ix86_indirect_branch_register with
11775 TARGET_INDIRECT_BRANCH_REGISTER.
11777 * config/i386/i386.md (indirect_jump): Likewise.
11778 (tablejump): Likewise.
11779 (*sibcall_memory): Likewise.
11780 (*sibcall_value_memory): Likewise.
11781 Peepholes of indirect call and jump via memory: Likewise.
11782 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11783 (*sibcall_value_GOT_32): Likewise.
11784 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11785 (GOT_memory_operand): Likewise.
11786 (call_insn_operand): Likewise.
11787 (sibcall_insn_operand): Likewise.
11788 (GOT32_symbol_operand): Likewise.
11789 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11791 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11793 PR rtl-optimization/83496
11794 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11795 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11796 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11797 redundant insn, if any.
11798 (relax_delay_slots): Likewise.
11799 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11801 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11803 PR tree-optimization/83965
11804 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11805 that grouped statements are part of a reduction chain. Return
11806 true if the statement is not marked as a reduction itself but
11807 is part of a group.
11808 (vect_recog_dot_prod_pattern): Don't check whether the statement
11809 is part of a group here.
11810 (vect_recog_sad_pattern): Likewise.
11811 (vect_recog_widen_sum_pattern): Likewise.
11813 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11816 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11818 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11821 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11823 (ix86_output_indirect_function_return): New prototype.
11824 (ix86_split_simple_return_pop_internal): Likewise.
11825 * config/i386/i386.c (indirect_return_via_cx): New.
11826 (indirect_return_via_cx_bnd): Likewise.
11827 (indirect_thunk_name): Handle return va CX_REG.
11828 (output_indirect_thunk_function): Create alias for
11829 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11830 (ix86_output_indirect_jmp): Remove the bool argument.
11831 (ix86_output_indirect_function_return): New function.
11832 (ix86_split_simple_return_pop_internal): Likewise.
11833 * config/i386/i386.md (*indirect_jump): Don't pass false
11834 to ix86_output_indirect_jmp.
11835 (*tablejump_1): Likewise.
11836 (simple_return_pop_internal): Change it to define_insn_and_split.
11837 Call ix86_split_simple_return_pop_internal to split it for
11838 -mfunction-return=.
11839 (simple_return_indirect_internal): Call
11840 ix86_output_indirect_function_return instead of
11841 ix86_output_indirect_jmp.
11843 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11846 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11847 memset and value initialization afterwards.
11849 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11851 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11853 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11856 * common/config/aarch64/aarch64-common.c
11857 (aarch_option_optimization_table[]): Switch
11858 off fomit-frame-pointer
11860 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11861 Chung-Ju Wu <jasonwucj@gmail.com>
11863 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11865 (store_multiple): Ditto.
11867 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11869 * config.gcc: Add --with-cpu support for nds32 target.
11870 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11871 * config/nds32/nds32.opt: Add -mcpu= option.
11873 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11875 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11876 isel=yes): Warn for these deprecated options.
11878 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11880 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11881 ISA_2_5_MASKS_EMBEDDED.
11883 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11885 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11886 p->max as pointers rather than using iterative_hash_expr.
11888 2018-02-23 Carl Love <cel@us.ibm.com>
11890 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11891 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11893 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11894 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11895 P8V_BUILTIN_VEC_VUNSIGNED2.
11897 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11900 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11901 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11903 * lra-constraints.c (curr_insn_transform): Set up
11904 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11906 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11907 * lra-eliminations.c (spill_pseudos): Ditto.
11908 (process_insn_for_elimination): Ditto.
11909 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11910 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11911 LRA_NON_CLOBBERED_ALT.
11913 2018-02-22 Martin Sebor <msebor@redhat.com>
11915 PR tree-optimization/84480
11916 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11917 to maybe_diag_stxncpy_trunc. Call it.
11918 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11919 from gimple_fold_builtin_strcpy. Print inlining stack.
11920 (handle_builtin_stxncpy): Print inlining stack.
11921 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11923 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11926 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11927 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11928 and -fcheck-pointer-bounds are used together.
11929 (indirect_thunk_prefix): New enum.
11930 (indirect_thunk_need_prefix): New function.
11931 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11932 "_nt" instead of "_bnd" for NOTRACK prefix.
11933 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11934 (output_indirect_thunk_function): Likewise.
11936 (ix86_code_end): Update output_indirect_thunk_function calls.
11937 (ix86_output_indirect_branch_via_reg): Replace
11938 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11939 (ix86_output_indirect_branch_via_push): Likewise.
11940 (ix86_output_function_return): Likewise.
11941 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11942 incompatible with -fcf-protection=branch and
11943 -fcheck-pointer-bounds.
11945 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11948 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11949 Change gcc_assert call to output_operand_lossage.
11951 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11953 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11955 2018-02-22 DJ Delorie <dj@redhat.com>
11956 Sebastian Perta <sebastian.perta@renesas.com>
11957 Oleg Endo <olegendo@gcc.gnu.org>
11959 * config/rx/rx.c (rx_rtx_costs): New function.
11960 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11962 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11964 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11966 2018-02-22 Martin Liska <mliska@suse.cz>
11969 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11970 Add "native" as a possible value.
11972 2018-02-22 Martin Liska <mliska@suse.cz>
11975 * config/i386/i386.c (ix86_option_override_internal):
11976 Add "native" as a possible value for -march and -mtune.
11978 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11981 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11982 to all type variants.
11984 PR tree-optimization/84503
11985 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11986 width as info->bitpos + info->bitsize - start.
11987 (merged_store_group::merge_overlapping): Simplify width computation.
11988 (check_no_overlap): New function.
11989 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11990 start + width and last_order of the group, fail if check_no_overlap
11992 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11993 to group if check_no_overlap fails.
11995 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11997 * config/rs6000/altivec.md: Delete contraint arguments to
11998 define_expand, define_split, and define_peephole2, and in
11999 define_insn_and_split if always unused.
12000 * config/rs6000/darwin.md: Ditto.
12001 * config/rs6000/dfp.md: Ditto.
12002 * config/rs6000/rs6000.md: Ditto.
12003 * config/rs6000/sync.md: Ditto.
12004 * config/rs6000/vector.md: Ditto.
12005 * config/rs6000/vsx.md: Ditto.
12007 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12009 * config/rs6000/altivec.md: Write output control strings as braced
12010 blocks instead of double-quoted strings.
12011 * config/rs6000/darwin.md: Ditto.
12012 * config/rs6000/rs6000.md: Ditto.
12013 * config/rs6000/vector.md: Ditto.
12014 * config/rs6000/vsx.md: Ditto.
12016 2018-02-21 Jason Merrill <jason@redhat.com>
12018 PR c++/84314 - ICE with templates and fastcall attribute.
12019 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12021 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12023 * ipa-cp.c (determine_versionability): Fix comment typos.
12025 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12028 * ipa-cp.c (determine_versionability): Do not version functions caling
12031 2018-02-21 Martin Liska <mliska@suse.cz>
12034 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12035 Add "native" as a possible value.
12036 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12037 the macro when native cpu detection is available.
12039 2018-02-21 Martin Liska <mliska@suse.cz>
12042 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12043 Add "native" as a possible value.
12044 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12045 when native cpu detection is available.
12047 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12048 Martin Sebor <msebor@redhat.com>
12050 PR tree-optimization/84478
12051 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12053 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12054 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12055 support which is conservatively correct, for 2 only stay conservative
12056 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12057 argument to the 2 argument get_range_strlen, adjust 6 arg
12058 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12060 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12061 (gimple_fold_builtin_strlen): Pass true as last argument to
12064 2018-02-20 Martin Sebor <msebor@redhat.com>
12066 PR middle-end/84095
12067 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12068 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12069 (builtin_memref::builtin_memref): Factor out parts into
12070 set_base_and_offset and call it.
12072 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12074 PR middle-end/84406
12075 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12076 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12077 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12078 search at the associated MODE_INT.
12080 2018-02-20 Jeff Law <law@redhat.com>
12082 PR middle-end/82123
12083 PR tree-optimization/81592
12084 PR middle-end/79257
12085 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12086 for range data rather than using global data.
12087 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12088 range data rather than using global data.
12089 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12090 pass it to children as needed.
12091 (struct directive::fmtresult): Similarly.
12092 (struct directive::set_width): Similarly.
12093 (struct directive::set_precision): Similarly.
12094 (format_integer, format_directive, parse_directive): Similarly.
12095 (format_none): Accept unnamed vr_values parameter.
12096 (format_percent, format_floating, format_character): Similarly.
12097 (format_string, format_plain): Similarly.
12098 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12099 the EVRP range analyzer for range data rather than using global data.
12100 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12101 gimple-ssa-evrp-analyze.h
12102 (class sprintf_dom_walker): Add after_dom_children member function.
12103 Add evrp_range_analyzer member.
12104 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12105 range analyzer as needed.
12106 (sprintf_dom_walker::after_dom_children): New member function.
12107 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12109 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12110 (evrp_range_analyzer::pop_to_marker): Likewise.
12112 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12114 PR tree-optimization/84419
12115 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12116 with the required type if its current type is compatible but
12119 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12121 PR middle-end/82004
12122 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12123 after vectorization.
12125 2018-02-20 Martin Liska <mliska@suse.cz>
12128 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12129 possible values if we don't have a hint.
12131 2018-02-20 Martin Liska <mliska@suse.cz>
12135 * final.c (shorten_branches): Build align_tab array with one
12137 * opts.c (finish_options): Add alignment option limit check.
12138 (MAX_CODE_ALIGN): Likewise.
12139 (MAX_CODE_ALIGN_VALUE): Likewise.
12140 * doc/invoke.texi: Document maximum allowed option value for
12141 all -falign-* options.
12143 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12146 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12147 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12148 * var-tracking.c (emit_note_insn_var_location): Remove all references
12149 to NOTE_INSN_CALL_ARG_LOCATION.
12150 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12151 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12152 Use copy_rtx_if_shared.
12153 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12154 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12155 (dwarf2out_var_location): Remove handling of
12156 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12158 * final.c (final_scan_insn): Remove all references to
12159 NOTE_INSN_CALL_ARG_LOCATION.
12160 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12161 before dumping final insns.
12162 * except.c (emit_note_eh_region_end): Remove all references to
12163 NOTE_INSN_CALL_ARG_LOCATION.
12164 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12165 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12166 * config/arc/arc.c (hwloop_optimize): Likewise.
12167 * config/arm/arm.c (create_fix_barrier): Likewise.
12168 * config/s390/s390.c (s390_chunkify_start): Likewise.
12169 * config/sh/sh.c (find_barrier): Likewise.
12170 * config/i386/i386.c (rest_of_insert_endbranch,
12171 ix86_seh_fixup_eh_fallthru): Likewise.
12172 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12173 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12174 * config/frv/frv.c (frv_function_prologue): Likewise.
12175 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12177 (note_outside_basic_block_p): Remove all references to
12178 NOTE_INSN_CALL_ARG_LOCATION.
12179 * gengtype.c (adjust_field_rtx_def): Likewise.
12180 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12182 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12183 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12186 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12189 PR tree-optimization/84452
12190 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12191 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12194 2018-02-19 Martin Liska <mliska@suse.cz>
12197 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12199 2018-02-19 Martin Liska <mliska@suse.cz>
12200 Richard Sandiford <richard.sandiford@linaro.org>
12202 PR tree-optimization/82491
12203 * gimple-fold.c (get_base_constructor): Make earlier bail out
12206 2018-02-19 Carl Love <cel@us.ibm.com>
12208 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12209 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12211 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12212 P8V_BUILTIN_VEC_NEG.
12214 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12216 * config/rl78/rl78.md (movdf): New define expand.
12218 2018-02-19 Martin Liska <mliska@suse.cz>
12221 * doc/invoke.texi: Fix typo.
12222 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12224 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12226 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12227 handle rs6000_single_float and rs6000_double_float specially for
12230 2018-02-16 Jeff Law <law@redhat.com>
12232 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12233 (pop_regs): Use it.
12235 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12238 * ipa-inline.c (inline_small_functions): Fix a typo.
12240 2018-02-16 Nathan Sidwell <nathan@acm.org>
12242 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12244 2018-02-16 Carl Love <cel@us.ibm.com>
12246 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12247 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12248 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12249 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12250 expansion to P8V_BUILTIN_VEC_FLOAT2.
12252 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12254 PR rtl-optimization/70023
12255 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12256 src_regno into account.
12258 2018-02-16 Carl Love <cel@us.ibm.com>
12260 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12261 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12262 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12263 * config/rs6000/rs6000.c: Remove case statements for
12264 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12265 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12266 and P9V_BUILTIN_VEC_VINSERT4B.
12267 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12268 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12269 * config/rs6000/vsx.md:
12270 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12273 2018-02-16 Carl Love <cel@us.ibm.com>
12275 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12277 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12279 * config/rs6000/rs6000-c.c: Add the definitions for
12280 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12281 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12282 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12283 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12284 definition for insert4b and define insn *insert3b_internal.
12285 * doc/extend.texi: Add documentation for vec_extract4b.
12287 2018-02-16 Nathan Sidwell <nathan@acm.org>
12289 * doc/extend.texi (Backwards Compatibility): Mention friend
12290 injection. Note for-scope is deprecated.
12291 * doc/invoke.texi (-ffriend-injection): Deprecate.
12293 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12295 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12296 that moved to I2, also allow destinations that are a paradoxical
12297 subreg (instead of a normal reg).
12299 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12302 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12305 2018-02-16 Richard Biener <rguenther@suse.de>
12307 PR tree-optimization/84037
12308 PR tree-optimization/84016
12310 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12311 Adjust vec_construct for the fact we need additional higher latency
12312 128bit inserts for AVX256 and AVX512 vector builds.
12313 (ix86_add_stmt_cost): Scale vector construction cost for
12316 2018-02-16 Richard Biener <rguenther@suse.de>
12318 PR tree-optimization/84417
12319 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12320 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12321 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12323 2018-02-16 Martin Liska <mliska@suse.cz>
12326 * internal-fn.def (ASAN_CHECK): Set proper flags.
12327 (ASAN_MARK): Likewise.
12329 2018-02-16 Julia Koval <julia.koval@intel.com>
12331 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12332 from PTA_CANNONLAKE.
12334 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12337 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12338 Use ++iter rather than iter++ for std::list iterators.
12339 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12340 defer deleting them until all nodes in the forest are processed. Do
12341 free even leaf nodes. Change to_process into auto_vec.
12344 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12345 * vec.h (vec_default_construct): Use memset instead of placement new
12346 if BROKEN_VALUE_INITIALIZATION is defined.
12347 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12348 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12351 PR rtl-optimization/83723
12352 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12353 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12354 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12356 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12358 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12360 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12362 PR rtl-optimization/81443
12363 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12364 from inner REGs to paradoxical SUBREGs.
12366 2018-02-16 Richard Biener <rguenther@suse.de>
12368 PR tree-optimization/84399
12369 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12370 For operands we can analyze at their definition make sure we can
12371 analyze them at each use as well.
12373 2018-02-16 Richard Biener <rguenther@suse.de>
12375 PR tree-optimization/84190
12376 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12377 volatile accesses if the decl isn't volatile.
12379 2018-02-15 Jason Merrill <jason@redhat.com>
12381 PR c++/84314 - ICE with templates and fastcall attribute.
12382 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12383 TYPE_CANONICAL on an existing type.
12385 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12387 PR tree-optimization/84383
12388 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12389 dstoff nor call operand_equal_p if dstbase is NULL.
12391 PR tree-optimization/84334
12392 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12393 also a CONSTANT_CLASS_P, punt.
12395 2018-02-14 Jim Wilson <jimw@sifive.com>
12397 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12398 first SMALL_OPERAND check. New local min_second_step. Move assert
12399 to where locals are set. Add TARGET_RVC support.
12400 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12402 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12404 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12406 2018-02-14 Martin Sebor <msebor@redhat.com>
12408 PR tree-optimization/83698
12409 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12410 arrays constrain the offset range to their bounds.
12411 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12412 (builtin_access::overlap): Avoid setting the size of overlap if it's
12414 (maybe_diag_overlap): Also consider arrays when deciding what values
12415 of offsets to include in diagnostics.
12417 2018-02-14 Martin Sebor <msebor@redhat.com>
12420 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12421 that correspond to the kind of a declaration.
12423 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12426 * config/pa/pa.md: Load address of PIC label using the linkage table
12427 if the label is nonlocal.
12429 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12431 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12432 warning message if user requests -maltivec=be.
12433 * doc/invoke.texi: Document deprecation of -maltivec=be.
12435 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12438 * config/rs6000/rs6000-c.c: Update definitions for
12439 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12440 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12442 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12445 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12446 add _get_ssp intrinsics. Remove argument from
12447 __builtin_ia32_rdssp[d|q].
12448 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12449 * config/i386/i386-builtin.def: Remove argument from
12450 __builtin_ia32_rdssp[d|q].
12451 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12452 ix86_expand_special_args_builtin for _rdssp[d|q].
12453 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12454 Clear register before usage.
12455 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12456 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12458 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12460 PR tree-optimization/84357
12461 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12462 operand 1 of an ARRAY_REF too.
12464 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12467 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12468 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12470 (set_of_reg): New struct.
12471 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12472 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12473 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12475 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12476 Split into bitclr, bitset, bitinvert patterns if appropriate.
12477 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12478 use rx_fuse_in_memory_bitop.
12479 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12480 to named insn, correct maximum insn length.
12482 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12485 * machmode.def: Define a complex mode for PARTIAL_INT.
12486 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12488 * doc/rtl.texi: Document CSPImode.
12489 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12491 (msp430_hard_regno_nregs_with_padding): Likewise.
12493 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12496 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12498 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12500 PR rtl-optimization/84169
12501 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12502 we generated a parallel as new i3 and we split that to new i2 and i3
12503 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12504 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12505 those to i2, not i1. Partially rewrite this scan code.
12507 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12510 * stor-layout.c (place_field): For variable length fields, adjust
12511 offset_align afterwards not just based on the field's alignment,
12512 but also on the size.
12514 PR middle-end/84309
12515 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12516 of exps and logs in the use_exp2 case.
12518 2018-02-13 Jeff Law <law@redhat.com>
12520 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12521 entry for "vector".
12523 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12526 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12530 * common.opt (gas-loc-support, gas-locview-support): New.
12531 (ginline-points, ginternal-reset-location-views): New.
12532 * doc/invoke.texi: Document them. Use @itemx where intended.
12533 (gvariable-location-views): Adjust.
12534 * target.def (reset_location_view): New.
12535 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12536 (TARGET_RESET_LOCATION_VIEW): New.
12537 * doc/tm.texi: Rebuilt.
12538 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12539 (dwarf2out_default_as_locview_support): New.
12540 (output_asm_line_debug_info): Use option variables.
12541 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12542 (output_loc_list): Likewise.
12543 (add_high_low_attributes): Check option variables.
12544 Don't output entry view attribute in strict mode.
12545 (gen_inlined_subroutine_die): Check option variables.
12546 (dwarf2out_inline_entry): Likewise.
12547 (init_sections_and_labels): Likewise.
12548 (dwarf2out_early_finish): Likewise.
12549 (maybe_reset_location_view): New, from...
12550 (dwarf2out_var_location): ... here. Call it.
12551 * debug.h (dwarf2out_default_as_loc_support): Declare.
12552 (dwarf2out_default_as_locview_support): Declare.
12553 * hooks.c (hook_int_rtx_insn_0): New.
12554 * hooks.h (hook_int_rtx_insn_0): Declare.
12555 * toplev.c (process_options): Take -gas-loc-support and
12556 -gas-locview-support from dwarf2out. Enable
12557 -gvariable-location-views by default only with locview
12558 assembler support. Enable -ginternal-reset-location-views by
12559 default only if the target defines the corresponding hook.
12560 Enable -ginline-points by default if location views are
12561 enabled; force it disabled if statement frontiers are
12563 * tree-inline.c (expand_call_inline): Check option variables.
12564 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12566 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12568 PR tree-optimization/84321
12569 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12570 handling. Also check whether the anti-range contains any values
12571 that satisfy the mask; switch to a VR_RANGE if not.
12573 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12576 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12578 2018-02-13 Martin Jambor <mjambor@suse.cz>
12581 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12582 of call statements, also set location of a load to a temporary.
12584 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12586 * config/rl78/rl78.c (add_vector_labels): New function.
12587 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12588 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12589 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12590 which checks that no arguments are passed.
12591 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12592 * doc/extend.texi: Documentation for the new attribute.
12594 2018-02-13 Andreas Schwab <schwab@suse.de>
12596 * config/riscv/linux.h (CPP_SPEC): Define.
12598 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12601 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12602 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12603 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12604 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12605 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12606 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12607 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12608 temporarily for AES and PCLMUL builtins.
12610 PR tree-optimization/84339
12611 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12612 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12615 PR middle-end/84309
12616 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12617 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12618 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12620 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12622 * omp-simd-clone.h: New file.
12623 * omp-simd-clone.c: Include omp-simd-clone.h.
12624 (expand_simd_clones): No longer static.
12625 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12626 cgraph.h and omp-simd-clone.h.
12627 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12628 (vect_recog_widen_shift_pattern): Formatting fix.
12629 (vect_pattern_recog_1): Don't check optab for calls.
12632 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12633 operands[2] into a REG before using gen_lowpart on it.
12635 2018-02-12 Jeff Law <law@redhat.com>
12638 * config/sh/sh.c (find_barrier): Consider a sibling call
12641 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12642 successfully back substituting a reg.
12644 2018-02-12 Richard Biener <rguenther@suse.de>
12646 PR tree-optimization/84037
12647 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12648 parameter, move visited init to caller.
12649 (vect_slp_analyze_operations): Separate cost from validity
12650 check, initialize visited once for all instances.
12651 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12653 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12655 (vect_model_promotion_demotion_cost): Likewise.
12656 (vectorizable_bswap): Guard cost modeling with !slp_node
12657 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12659 (vectorizable_call): Likewise.
12660 (vectorizable_conversion): Likewise.
12661 (vectorizable_assignment): Likewise.
12662 (vectorizable_shift): Likewise.
12663 (vectorizable_operation): Likewise.
12664 (vectorizable_store): Likewise.
12665 (vectorizable_load): Likewise.
12666 (vectorizable_condition): Likewise.
12667 (vectorizable_comparison): Likewise.
12669 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12672 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12673 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12674 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12676 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12678 PR middle-end/83665
12679 * params.def (inline-min-speedup): Increase from 8 to 15.
12680 (max-inline-insns-auto): Decrease from 40 to 30.
12681 * ipa-split.c (consider_split): Add some buffer for function to
12682 be considered inlining candidate.
12683 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12686 2018-02-12 Richard Biener <rguenther@suse.de>
12688 PR tree-optimization/84037
12689 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12690 matched stmts if we cannot swap the non-matched ones.
12692 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12694 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12695 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12696 _mm_maskz_scalef_round_ss): New intrinsics.
12697 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12698 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12699 __builtin_ia32_scalefss_round): Remove.
12700 (__builtin_ia32_scalefsd_mask_round,
12701 __builtin_ia32_scalefss_mask_round): New intrinsics.
12702 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12703 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12704 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12705 "<round_constraint>")): Changed to ...
12706 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12707 "<round_scalar_constraint>")): ... this.
12708 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12709 %0, %1, %2<round_op3>}"): Changed to ...
12710 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12711 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12712 %2<round_scalar_mask_op3>}"): ... this.
12713 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12715 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12717 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12718 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12719 (_mm_maskz_sqrt_round_ss): New intrinsics.
12720 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12721 (__builtin_ia32_sqrtsd_mask_round)
12722 (__builtin_ia32_sqrtss_mask_round): New builtins.
12723 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12724 (__builtin_ia32_sqrtss_round): Remove.
12725 (__builtin_ia32_sqrtsd_mask_round)
12726 (__builtin_ia32_sqrtss_mask_round): New builtins.
12727 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12728 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12729 ((match_operand:VF_128 1 "vector_operand"
12730 "xBm,<round_constraint>")): Changed to ...
12731 ((match_operand:VF_128 1 "vector_operand"
12732 "xBm,<round_scalar_constraint>")): ... this.
12733 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12734 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12735 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12736 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12737 %<iptr>1<round_scalar_mask_op3>}): ... this.
12738 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12739 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12741 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12744 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12745 Cast vec_cmpeq result to correct type.
12746 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12747 Cast vec_cmpgt result to correct type.
12749 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12751 * final.c (final_scan_insn_1): Renamed from...
12752 (final_scan_insn): ... this. New wrapper, to recover
12753 seen from the outermost call in recursive ones.
12754 * config/sparc/sparc.c (output_return): Drop seen from call.
12755 (output_sibcall): Likewise.
12756 * config/visium/visium.c (output_branch): Likewise.
12758 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12760 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12763 2018-02-10 Alan Modra <amodra@gmail.com>
12766 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12767 Specify LR as an input.
12769 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12772 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12773 remove_member_access_dummy_vars): New functions.
12774 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12775 lower_omp_1, execute_lower_omp): Use them.
12777 PR rtl-optimization/84308
12778 * shrink-wrap.c (spread_components): Release todo vector.
12780 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12782 PR rtl-optimization/57193
12783 * ira-color.c (struct allocno_color_data): Add member
12784 conflict_allocno_hard_prefs.
12785 (update_conflict_allocno_hard_prefs): New.
12786 (bucket_allocno_compare_func): Add a preference based on
12787 conflict_allocno_hard_prefs.
12788 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12789 (color_allocnos): Remove a dead code. Initiate
12790 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12792 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12795 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12796 constraint from =wa to wa. Avoid a subreg on the output operand,
12797 instead use a pseudo and subreg it in a move.
12798 (p9_xxbrd_<mode>): Changed to ...
12799 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12800 (p9_xxbrd_v2df): New expander.
12801 (p9_xxbrw_<mode>): Changed to ...
12802 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12803 (p9_xxbrw_v4sf): New expander.
12805 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12807 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12809 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12812 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12813 multiply in 32-bit mode.
12814 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12815 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12818 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12820 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12821 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12822 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12823 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12825 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12828 * dwarf2out.c (is_trivial_indirect_ref): New function.
12829 (dwarf2out_late_global_decl): Do not generate a location
12830 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12831 and that are not defined in the current unit.
12833 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12835 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12836 instead of a libcall for UNORDERED.
12838 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12841 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12842 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12844 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12847 * config/s390/s390.c (s390_set_current_function): Invoke
12848 s390_indirect_branch_settings also if fndecl didn't change.
12850 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12852 * config/rs6000/rs6000.md (blockage): Set length to zero.
12854 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12856 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12858 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12861 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12862 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12866 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12867 PARALLEL incoming that failed vt_get_decl_and_offset check.
12869 PR middle-end/84237
12870 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12871 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12873 (get_variable_section): For decls in named .bss* sections pass true as
12874 second argument to bss_initializer_p.
12876 2018-02-09 Marek Polacek <polacek@redhat.com>
12877 Jakub Jelinek <jakub@redhat.com>
12880 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12881 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12882 Sync some changes from cxx_fold_indirect_ref.
12884 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12886 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12888 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12889 (BLOCK_INLINE_ENTRY_LABEL): New.
12890 (dwarf2out_var_location): Disregard inline entry markers.
12891 (inline_entry_data): New struct.
12892 (inline_entry_data_hasher): New hashtable type.
12893 (inline_entry_data_hasher::hash): New.
12894 (inline_entry_data_hasher::equal): New.
12895 (inline_entry_data_table): New variable.
12896 (add_high_low_attributes): Add DW_AT_entry_pc and
12897 DW_AT_GNU_entry_view attributes if a pending entry is found
12898 in inline_entry_data_table. Add old entry_pc attribute only
12899 if debug nonbinding markers are disabled.
12900 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12901 markers are enabled.
12902 (block_within_block_p, dwarf2out_inline_entry): New.
12903 (dwarf2out_finish): Check that no entries remained in
12904 inline_entry_data_table.
12905 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12906 (final_scan_insn, notice_source_line): Likewise.
12907 (rest_of_clean_state): Skip inline entry markers.
12908 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12910 * gimple.c (gimple_build_debug_inline_entry): New.
12911 * gimple.h (enum gimple_debug_subcode): Add
12912 GIMPLE_DEBUG_INLINE_ENTRY.
12913 (gimple_build_debug_inline_entry): Declare.
12914 (gimple_debug_inline_entry_p): New.
12915 (gimple_debug_nonbind_marker_p): Adjust.
12916 * insn-notes.def (INLINE_ENTRY): New.
12917 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12918 inline entry marker notes.
12919 (print_insn): Likewise.
12920 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12921 (INSN_DEBUG_MARKER_KIND): Likewise.
12922 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12923 * tree-inline.c (expand_call_inline): Build and insert
12924 debug_inline_entry stmt.
12925 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12926 inline entry blocks early, if nonbind markers are enabled.
12927 (dump_scope_block): Dump fragment info.
12928 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12929 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12930 (gimple_build_debug_inline_entry): New.
12931 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12932 Enable/disable inline entry points too.
12933 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12934 (DEBUG_INSN): Describe inline entry markers.
12936 * common.opt (gvariable-location-views): New.
12937 (gvariable-location-views=incompat5): New.
12938 * config.in: Rebuilt.
12939 * configure: Rebuilt.
12940 * configure.ac: Test assembler for view support.
12941 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12942 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12943 * dwarf2out.c (var_loc_view): New typedef.
12944 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12945 (dwarf2out_locviews_in_attribute): New.
12946 (dwarf2out_locviews_in_loclist): New.
12947 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12948 (enum dw_line_info_opcode): Add LI_adv_address.
12949 (struct dw_line_info_table): Add view.
12950 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12951 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12952 (zero_view_p): New variable.
12953 (ZERO_VIEW_P): New macro.
12954 (output_asm_line_debug_info): New.
12955 (struct var_loc_node): Add view.
12956 (add_AT_view_list, AT_loc_list): New.
12957 (add_var_loc_to_decl): Add view param. Test it against last.
12958 (new_loc_list): Add view params. Record them.
12959 (AT_loc_list_ptr): Handle loc and view lists.
12960 (view_list_to_loc_list_val_node): New.
12961 (print_dw_val): Handle dw_val_class_view_list.
12962 (size_of_die): Likewise.
12963 (value_format): Likewise.
12964 (loc_list_has_views): New.
12965 (gen_llsym): Set vl_symbol too.
12966 (maybe_gen_llsym, skip_loc_list_entry): New.
12967 (dwarf2out_maybe_output_loclist_view_pair): New.
12968 (output_loc_list): Output view list or entries too.
12969 (output_view_list_offset): New.
12970 (output_die): Handle dw_val_class_view_list.
12971 (output_dwarf_version): New.
12972 (output_compilation_unit_header): Use it.
12973 (output_skeleton_debug_sections): Likewise.
12974 (output_rnglists, output_line_info): Likewise.
12975 (output_pubnames, output_aranges): Update version comments.
12976 (output_one_line_info_table): Output view numbers in asm comments.
12977 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12978 Call maybe_gen_llsym.
12979 (loc_list_from_tree_1): Adjust.
12980 (add_AT_location_description): Create view list attribute if
12981 needed, check it's absent otherwise.
12982 (convert_cfa_to_fb_loc_list): Adjust.
12983 (maybe_emit_file): Call output_asm_line_debug_info for test.
12984 (dwarf2out_var_location): Reset views as needed. Precompute
12985 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12986 attribute. Set view.
12987 (new_line_info_table): Reset next view.
12988 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12989 (dwarf2out_source_line): Likewise. Output view resets and labels to
12990 the assembler, or select appropriate line info opcodes.
12991 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12992 (optimize_string_length): Catch it. Adjust.
12993 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12994 dw_val_class_view_list, and remove it if no longer needed.
12995 (hash_loc_list): Hash view numbers.
12996 (loc_list_hasher::equal): Compare them.
12997 (optimize_location_lists): Check whether a view list symbol is
12998 needed, and whether the locview attribute is present, and
12999 whether they match. Remove the locview attribute if no longer
13001 (index_location_lists): Call skip_loc_list_entry for test.
13002 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13003 Use output_dwarf_version.
13004 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13005 (struct dw_val_node): Add val_view_list.
13006 * final.c (SEEN_NEXT_VIEW): New.
13007 (set_next_view_needed): New.
13008 (clear_next_view_needed): New.
13009 (maybe_output_next_view): New.
13010 (final_start_function): Rename to...
13011 (final_start_function_1): ... this. Take pointer to FIRST,
13012 add SEEN parameter. Emit param bindings in the initial view.
13013 (final_start_function): Reintroduce SEEN-less interface.
13014 (final): Rename to...
13015 (final_1): ... this. Take SEEN parameter. Output final pending
13016 next view at the end.
13017 (final): Reintroduce seen-less interface.
13018 (final_scan_insn): Output pending next view before switching
13019 sections or ending a block. Mark the next view as needed when
13020 outputting variable locations. Notify debug backend of section
13021 changes, and of location view changes.
13022 (rest_of_handle_final): Adjust.
13023 * toplev.c (process_options): Autodetect value for debug variable
13024 location views option. Warn on incompat5 without -gdwarf-5.
13025 * doc/invoke.texi (gvariable-location-views): New.
13026 (gvariable-location-views=incompat5): New.
13027 (gno-variable-location-views): New.
13029 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13031 PR tree-optimization/84136
13032 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13033 that the result of find_edge is non-NULL.
13035 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13038 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13039 storing integer register in SImode. Fix cost of 256 and 512
13040 byte aligned SSE register store.
13042 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13044 * config/i386/i386.c (ix86_multiplication_cost): Fix
13045 multiplication cost for TARGET_AVX512DQ.
13047 2018-02-08 Marek Polacek <polacek@redhat.com>
13049 PR tree-optimization/84238
13050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13053 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13055 PR tree-optimization/84265
13056 * tree-vect-stmts.c (vectorizable_store): Don't treat
13057 VMAT_CONTIGUOUS accesses as grouped.
13058 (vectorizable_load): Likewise.
13060 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13062 PR tree-optimization/81635
13063 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13064 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13065 (test_round_for_mask): New functions.
13066 (wide_int_cc_tests): Call test_round_for_mask.
13067 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13068 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13069 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13070 range returned by get_range_info.
13072 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13075 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13076 * symtab.c: Include builtins.h
13077 (symtab_node::output_to_lto_symbol_table_p): Move here
13078 from lto-streamer-out.c:output_symbol_p.
13079 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13080 (output_symbol_p): Move all logic to symtab.c
13081 (produce_symtab): Update.
13083 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13085 * config/s390/s390-opts.h (enum indirect_branch): Define.
13086 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13087 (s390_indirect_branch_via_thunk)
13088 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13089 (enum s390_indirect_branch_type): Define.
13090 * config/s390/s390.c (struct s390_frame_layout, struct
13091 machine_function): Remove.
13092 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13093 (indirect_branch_table_label_no, indirect_branch_table_name):
13095 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13096 (enum s390_indirect_branch_option): Define.
13097 (s390_return_addr_from_memory): New function.
13098 (s390_handle_string_attribute): New function.
13099 (s390_attribute_table): Add new attribute handler.
13100 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13101 (s390_indirect_branch_via_thunk): New function.
13102 (s390_indirect_branch_via_inline_thunk): New function.
13103 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13104 sibling call optimization for non z10 compiles.
13105 (s390_emit_call): Force indirect branch target to be a single
13106 register. Add r1 clobber for non-z10 compiles.
13107 (s390_emit_epilogue): Emit return jump via return_use expander.
13108 (s390_reorg): Handle JUMP_INSNs as execute targets.
13109 (s390_option_override_internal): Perform validity checks for the
13110 new command line options.
13111 (s390_indirect_branch_attrvalue): New function.
13112 (s390_indirect_branch_settings): New function.
13113 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13114 (s390_output_indirect_thunk_function): New function.
13115 (s390_code_end): Implement target hook.
13116 (s390_case_values_threshold): Implement target hook.
13117 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13119 * config/s390/s390.h (struct s390_frame_layout)
13120 (struct machine_function): Move here from s390.c.
13121 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13122 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13123 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13124 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13125 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13126 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13127 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13128 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13129 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13130 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13131 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13132 (mnemonic attribute): Add values which aren't recognized
13134 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13135 pattern for branch conversion. Fix mnemonic attribute.
13136 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13137 indirect branch via thunk if requested.
13138 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13139 ("*indirect_jump"): Disable for branch conversion using out of
13141 ("indirect_jump_via_thunk<mode>_z10")
13142 ("indirect_jump_via_thunk<mode>")
13143 ("indirect_jump_via_inlinethunk<mode>_z10")
13144 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13145 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13146 ("casesi_jump_via_inlinethunk<mode>_z10")
13147 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13148 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13149 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13150 ("*indirect2_jump"): Disable for branch conversion.
13151 ("casesi_jump"): Turn into expander and expand patterns for branch
13153 ("return_use"): New expander.
13154 ("*return"): Emit return via thunk and rename it to ...
13155 ("*return<mode>"): ... this one.
13156 * config/s390/s390.opt: Add new options and and enum for the
13159 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13161 * lra-constraints.c (match_reload): Unconditionally use
13162 gen_lowpart_SUBREG, rather than selecting between that
13163 and equivalent gen_rtx_SUBREG code.
13165 2018-02-08 Richard Biener <rguenther@suse.de>
13167 PR tree-optimization/84233
13168 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13169 changed flag instead of boguously re-using phi_inserted.
13171 2018-02-08 Martin Jambor <mjambor@suse.cz>
13173 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13174 static local variables.
13176 2018-02-08 Richard Biener <rguenther@suse.de>
13178 PR tree-optimization/84278
13179 * tree-vect-stmts.c (vectorizable_store): When looking for
13180 smaller vector types to perform grouped strided loads/stores
13181 make sure the mode is supported by the target.
13182 (vectorizable_load): Likewise.
13184 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13186 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13187 Increase LDP/STP opportunities by adding adjacent callee-saves.
13189 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13191 PR rtl-optimization/84068
13192 PR rtl-optimization/83459
13193 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13195 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13197 PR tree-optimization/84224
13198 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13199 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13200 non-zero arguments.
13202 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13205 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13206 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13207 count, remove one USE.
13209 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13211 * doc/install.texi (Configuration): Document the
13212 --with-long-double-format={ibm,ieee} PowerPC configuration
13216 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13217 Convert from define_expand to be define_insn_and_split. Rework
13218 float/double/_Float128 conversions to QI/HI/SImode to work with
13219 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13220 conversions to QI/HImode types did a store and then a load to
13221 truncate the value. For conversions to VSX registers, don't split
13222 the insn, instead emit the code directly. Use the code iterator
13223 any_fix to combine signed and unsigned conversions.
13224 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13225 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13226 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13227 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13228 (fix_<mode>di2_hw): Likewise.
13229 (fixuns_<mode>di2_hw): Likewise.
13230 (fix_<mode>si2_hw): Likewise.
13231 (fixuns_<mode>si2_hw): Likewise.
13232 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13233 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13234 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13235 fix<uns>_trunc<SFDF:mode>si2_p8.
13236 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13237 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13238 (fix<uns>_<mode>_mem): Likewise.
13239 (fctiw<u>z_<mode>_mem): Likewise.
13240 (fix<uns>_<mode>_mem): Likewise.
13241 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13242 the register allocator from doing a direct move to the GPRs to do
13243 a store, and instead use the ISA 3.0 store byte/half-word from
13244 vector register instruction. For IEEE 128-bit floating point,
13245 also optimize stores of 32-bit ints.
13246 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13248 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13250 * genextract.c (push_pathstr_operand): New function to support
13252 (walk_rtx): Call push_pathstr_operand.
13253 (print_path): Support [a-zA-Z].
13255 2018-02-07 Richard Biener <rguenther@suse.de>
13257 PR tree-optimization/84037
13258 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13259 (cse_and_gimplify_to_preheader): Declare.
13260 (vect_get_place_in_interleaving_chain): Likewise.
13261 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13263 (_loop_vec_info::~_loop_vec_info): Delete it.
13264 (cse_and_gimplify_to_preheader): New function.
13265 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13266 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13267 (vectorizable_load): Likewise. For grouped stores always base
13268 the IV on the first element.
13269 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13270 condition before gimplifying.
13272 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13274 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13275 *DIV_EXPR and *MOD_EXPR.
13277 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13280 * config/i386/i386.c (ix86_option_override_internal): Mask out
13281 the CF_SET bit when checking -fcf-protection.
13283 2018-02-07 Tom de Vries <tom@codesourcery.com>
13286 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13289 2018-02-07 Richard Biener <rguenther@suse.de>
13291 PR tree-optimization/84204
13292 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13295 PR tree-optimization/84205
13296 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13297 special-case isl_ast_op_zdiv_r.
13299 PR tree-optimization/84223
13300 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13301 Only add conditions from within the region.
13302 (gather_bbs::after_dom_children): Adjust.
13304 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13307 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13308 * config/avr/avr.md: Only post-reload split REG-REG moves if
13309 either register is GENERAL_REG_P.
13311 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13313 PR tree-optimization/84235
13314 * tree-ssa-scopedtables.c
13315 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13316 if the subtraction is performed in floating point type where NaNs are
13317 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13318 build 1. Formatting fix.
13320 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13323 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13324 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13325 and skip it regardless of bb boundaries. Use CALL_P macro,
13326 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13327 unnecessarily, formatting fix.
13329 2018-02-06 Michael Collison <michael.collison@arm.com>
13331 * config/arm/thumb2.md:
13332 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13333 (*thumb_mov_notscc): Ditto.
13335 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13338 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13339 unsigned_fix, not "s".
13341 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13343 * configure.ac (gcc_fn_eh_frame_ro): New function.
13344 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13345 correct .eh_frame permissions.
13346 * configure: Regenerate.
13348 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13350 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13351 irrelevant options.
13353 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13355 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13356 Display warning message for -mno-speculate-indirect-jumps.
13358 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13360 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13362 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13364 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13366 PR tree-optimization/84225
13367 * tree-eh.c (find_trapping_overflow): Only call
13368 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13370 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13373 * config/i386/i386.c: Reimplement the check of possible options
13374 -mibt/-mshstk conbination. Change error messages.
13375 * doc/invoke.texi: Fix a typo: remove extra '='.
13377 2018-02-06 Marek Polacek <polacek@redhat.com>
13379 PR tree-optimization/84228
13380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13382 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13385 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13386 emitted arch directives.
13387 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13388 __ARM_FEATURE_COPROC before changing architectures.
13390 2018-02-06 Richard Biener <rguenther@suse.de>
13392 * config/i386/i386.c (print_reg): Fix typo.
13393 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13395 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13397 * configure: Regenerate.
13399 2018-02-05 Martin Sebor <msebor@redhat.com>
13401 PR tree-optimization/83369
13402 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13405 2018-02-05 Martin Liska <mliska@suse.cz>
13407 * doc/invoke.texi: Cherry-pick upstream r323995.
13409 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13411 * ira.c (ira_init_register_move_cost): Adjust comment.
13413 2018-02-05 Martin Liska <mliska@suse.cz>
13415 PR gcov-profile/84137
13416 * doc/gcov.texi: Fix typo in documentation.
13418 2018-02-05 Martin Liska <mliska@suse.cz>
13420 PR gcov-profile/83879
13421 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13422 using dlopen functionality.
13424 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13426 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13427 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13428 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13429 _mm_maskz_range_round_ss): New intrinsics.
13430 (__builtin_ia32_rangesd128_round)
13431 (__builtin_ia32_rangess128_round): Remove.
13432 (__builtin_ia32_rangesd128_mask_round,
13433 __builtin_ia32_rangess128_mask_round): New builtins.
13434 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13435 __builtin_ia32_rangess128_round): Remove.
13436 (__builtin_ia32_rangesd128_mask_round,
13437 __builtin_ia32_rangess128_mask_round): New builtins.
13438 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13439 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13440 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13441 "<round_saeonly_constraint>")): Changed to ...
13442 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13443 "<round_saeonly_scalar_constraint>")): ... this.
13444 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13445 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13446 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13447 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13448 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13450 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13452 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13454 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13455 Remove all values except native, 8540 and 8548.
13457 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13459 * config/i386/i386.c (ix86_output_function_return): Pass
13460 INVALID_REGNUM, instead of -1, as invalid register number to
13461 indirect_thunk_name and output_indirect_thunk.
13463 2018-02-02 Julia Koval <julia.koval@intel.com>
13465 * config.gcc: Add -march=icelake.
13466 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13467 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13468 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13469 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13470 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13471 (processor_target_table): Add icelake.
13472 (ix86_option_override_internal): Handle new PTAs.
13473 (get_builtin_code_for_version): Handle icelake.
13474 (M_INTEL_COREI7_ICELAKE): New.
13475 (fold_builtin_cpu): Handle icelake.
13476 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13477 * doc/invoke.texi: Add -march=icelake.
13479 2018-02-02 Julia Koval <julia.koval@intel.com>
13481 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13482 to wide_int_bitmask.
13483 * wide-int-bitmask.h: New.
13485 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13488 * config/i386/i386.md: Replace Pmode with word_mode in
13489 builtin_setjmp_setup and builtin_longjmp to support x32.
13491 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13495 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13497 (rs6000_supported_cpu_names): New static variable.
13498 (linux_cpu_translation_table): Likewise.
13499 (elf_platform) <cpu>: Define new static variable and use it.
13500 Translate kernel AT_PLATFORM name to canonical name if needed.
13501 Error if platform name is unknown.
13503 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13506 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13508 2018-02-01 Jeff Law <law@redhat.com>
13511 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13512 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13513 the scratch if RELEASE_VIA_POP is false.
13514 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13515 If we have to save a temporary register, decrement SIZE appropriately.
13516 Pass new arguments to release_scratch_register_on_entry.
13517 (ix86_adjust_stack_and_probe): Likewise.
13518 (ix86_emit_probe_stack_range): Pass new arguments to
13519 release_scratch_register_on_entry.
13521 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13523 PR rtl-optimization/84157
13524 * combine.c (change_zero_ext): Use REG_P predicate in
13525 front of HARD_REGISTER_P predicate.
13527 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13529 * config/avr/avr.c (avr_option_override): Move disabling of
13530 -fdelete-null-pointer-checks to...
13531 * common/config/avr/avr-common.c (avr_option_optimization_table):
13534 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13536 PR tree-optimization/81635
13537 * tree-data-ref.c (split_constant_offset_1): For types that
13538 wrap on overflow, try to use range info to prove that wrapping
13541 2018-02-01 Renlin Li <renlin.li@arm.com>
13544 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13545 TAILCALL_ADDR_REGS.
13546 (aarch64_register_move_cost): Likewise.
13547 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13548 TAILCALL_ADDR_REGS.
13549 (REG_CLASS_NAMES): Likewise.
13550 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13551 TAILCALL_ADDR_REGS. Remove IP registers.
13552 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13554 2018-02-01 Richard Biener <rguenther@suse.de>
13556 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13557 for specifying RPO order and allow NULL for that.
13558 * domwalk.c (dom_walker::dom_walker): Likewise.
13559 (dom_walker::walk): Handle NULL RPO order.
13560 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13562 (rewrite_update_dom_walker): Likewise.
13563 (mark_def_dom_walker): Likewise.
13565 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13567 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13568 (aarch64_maybe_expand_sve_subreg_move): Declare.
13569 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13570 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13572 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13573 that are semantically a reverse operation.
13574 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13575 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13576 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13578 (aarch64_can_change_mode_class): For big-endian, forbid changes
13579 between two SVE modes if they have different element sizes.
13581 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13583 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13584 the TImode handling for big-endian targets.
13586 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13588 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13589 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13591 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13592 Remove BSWAP handing for big-endian targets and use the form of
13593 LD1RQ appropariate for the mode.
13595 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13597 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13598 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13599 duplicated element.
13601 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13604 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13605 check for operands that need to go through aarch64_sve_reload_be.
13607 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13609 PR tree-optimization/81661
13610 PR tree-optimization/84117
13611 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13612 * tree-eh.c: Include gimplify.h.
13613 (find_trapping_overflow, replace_trapping_overflow,
13614 rewrite_to_non_trapping_overflow): New functions.
13615 * tree-vect-loop.c: Include tree-eh.h.
13616 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13617 * tree-data-ref.c: Include tree-eh.h.
13618 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13620 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13622 PR rtl-optimization/84123
13623 * combine.c (change_zero_ext): Check if hard register satisfies
13624 can_change_dest_mode before calling gen_lowpart_SUBREG.
13626 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13629 * ira.c (ira_init_register_move_cost): Remove assert.
13631 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13633 PR rtl-optimization/84071
13634 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13635 * doc/tm.texi: Regenerate.
13637 2018-01-31 Richard Biener <rguenther@suse.de>
13639 PR tree-optimization/84132
13640 * tree-data-ref.c (analyze_miv_subscript): Properly
13641 check whether evolution_function_is_affine_multivariate_p
13642 before calling gcd_of_steps_may_divide_p.
13644 2018-01-31 Julia Koval <julia.koval@intel.com>
13647 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13648 * config/i386/i386.md (rdpid_rex64) New.
13649 (rdpid): Make 32bit only.
13651 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13654 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13655 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13657 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13660 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13662 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13664 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13666 PR rtl-optimization/84071
13667 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13668 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13670 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13672 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13673 (arc_attribute_table): Add 'aux' attribute.
13674 (arc_in_small_data_p): Consider aux like variables.
13675 (arc_is_aux_reg_p): New function.
13676 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13677 (arc_get_aux_arg): New function.
13678 (prepare_move_operands): Handle aux-register access.
13679 (arc_handle_aux_attribute): New function.
13680 * doc/extend.texi (ARC Variable attributes): Add subsection.
13682 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13684 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13685 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13686 (arc_attribute_table): Add 'uncached' attribute.
13687 (arc_print_operand): Print '.di' flag for uncached memory
13689 (arc_in_small_data_p): Do not consider for small data the uncached
13691 (arc_is_uncached_mem_p): New function.
13692 * config/arc/predicates.md (compact_store_memory_operand): Check
13693 for uncached memory accesses.
13694 (nonvol_nonimm_operand): Likewise.
13695 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13697 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13700 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13701 falign-loops=): Add Optimization flag.
13703 2018-01-30 Jeff Law <law@redhat.com>
13706 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13707 INT_REGISTERS_SAVED. Check it prior to calling
13708 get_scratch_register_on_entry.
13709 (ix86_adjust_stack_and_probe): Similarly.
13710 (ix86_emit_probe_stack_range): Similarly.
13711 (ix86_expand_prologue): Corresponding changes.
13713 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13716 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13717 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13719 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13722 * lra-constraints.c (curr_insn_transform): Process AND in the
13725 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13727 PR rtl-optimization/83986
13728 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13729 dependence against last_pending_memory_flush in addition to
13730 pending_jump_insns.
13732 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13734 PR tree-optimization/81611
13735 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13738 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13741 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13744 2018-01-30 Richard Biener <rguenther@suse.de>
13745 Jakub Jelinek <jakub@redhat.com>
13747 PR tree-optimization/84111
13748 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13749 inner loops added during recursion, as they don't have up-to-date
13752 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13755 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13756 (can_inline_edge_by_limits_p): ... here.
13757 (can_early_inline_edge_p, check_callers,
13758 update_caller_keys, update_callee_keys, recursive_inlining,
13759 add_new_edges_to_heap, speculation_useful_p,
13760 inline_small_functions,
13761 inline_small_functions, flatten_function,
13762 inline_to_all_callers_1): Update.
13764 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13766 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13769 2018-01-30 Richard Biener <rguenther@suse.de>
13771 PR tree-optimization/83008
13772 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13773 invariant and constant vector uses in stmts when they need
13774 more than one stmt.
13776 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13779 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13780 * configure: Regenerate.
13782 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13784 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13786 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13787 Use gen_rtx_REG rather than gen_lowpart.
13789 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13791 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13792 rather than 0 when creating partial subregs.
13794 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13796 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13799 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13802 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13803 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13804 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13805 flags. This restores the settings used before the 2017-07-24.
13806 Turning off pre increment/decrement/modify allows IVOPTS to
13807 optimize DF/SF loops where the index is an int.
13809 2018-01-29 Richard Biener <rguenther@suse.de>
13810 Kelvin Nilsen <kelvin@gcc.gnu.org>
13813 * tree-vect-stmts.c (vectorizable_call): Don't call
13814 targetm.vectorize_builtin_md_vectorized_function if callee is
13817 2018-01-22 Carl Love <cel@us.ibm.com>
13819 * doc/extend.tex: Fix typo in second arg in
13820 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13822 2018-01-29 Richard Biener <rguenther@suse.de>
13824 PR tree-optimization/84086
13825 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13826 (flush_ssaname_freelist): When SSA names were released reset
13827 the SCEV hash table.
13829 2018-01-29 Richard Biener <rguenther@suse.de>
13831 PR tree-optimization/84057
13832 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13833 removed paths when removing edges.
13835 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13837 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13838 -mfunction-return=@var{choice}.
13840 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13842 PR diagnostic/84034
13843 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13844 Handle CR like TAB.
13845 (layout::print_source_line): Likewise.
13846 (test_get_line_width_without_trailing_whitespace): Add test cases.
13848 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13850 PR middle-end/84040
13851 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13854 2018-01-26 Jim Wilson <jimw@sifive.com>
13856 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13858 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13861 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13863 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13864 and CMP + SUB-immediate -> SUBS.
13866 2018-01-26 Martin Sebor <msebor@redhat.com>
13868 PR tree-optimization/83896
13869 * tree-ssa-strlen.c (get_string_len): Rename...
13870 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13871 Avoid assuming length is constant.
13872 (handle_char_store): Use HOST_WIDE_INT for string length.
13874 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13877 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13878 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13880 2018-01-26 Richard Biener <rguenther@suse.de>
13882 PR rtl-optimization/84003
13883 * dse.c (record_store): Only record redundant stores when
13884 the earlier store aliases at least all accesses the later one does.
13886 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13888 PR rtl-optimization/83985
13889 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13890 REG_CFA_RESTORE insns.
13891 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13892 REG_CFA_RESTORE insns here.
13895 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13896 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13898 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13900 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13901 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13902 (arc_init): Likewise.
13903 (arc_override_options): Likewise.
13904 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13906 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13908 * config/arc/arc.h (TARGET_DBNZ): Define.
13909 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13910 properly set the tune attribute.
13911 (dbnz): Use TARGET_DBNZ guard.
13912 * config/arc/arc.opt (mtune): Add core3 option.
13914 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13916 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13917 recognize new pic like addresses.
13918 (arc_delegitimize_address): Clean up.
13920 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13922 * config/arc/arc-arches.def: Option mrf16 valid for all
13924 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13925 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13926 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13927 * config/arc/arc-tables.opt: Regenerate.
13928 * config/arc/arc.c (arc_conditional_register_usage): Handle
13929 reduced register file case.
13930 (arc_file_start): Set must have build attributes.
13931 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13932 mrf16 option value.
13933 * config/arc/arc.opt (mrf16): Add new option.
13934 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13935 * config/arc/genmultilib.awk: Handle new mrf16 option.
13936 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13937 * config/arc/t-multilib: Regenerate.
13938 * doc/invoke.texi (ARC Options): Document mrf16 option.
13940 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13942 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13943 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13944 (arc_attribute_table): Add 'secure_call' attribute.
13945 (arc_print_operand): Print secure call operand.
13946 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13948 (arc_is_secure_call_p): New function. * config/arc/arc.md
13949 (call_i): Add support for sjli instruction.
13950 (call_value_i): Likewise.
13951 * config/arc/constraints.md (Csc): New constraint.
13953 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13954 John Eric Martin <John.Martin@emmicro-us.com>
13956 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13957 * config/arc/arc.c (_arc_jli_section): New struct.
13958 (arc_jli_section): New type.
13959 (rc_jli_sections): New static variable.
13960 (arc_handle_jli_attribute): New function.
13961 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13962 (arc_file_end): New function.
13963 (TARGET_ASM_FILE_END): Define.
13964 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13965 (arc_add_jli_section): New function.
13966 (jli_call_scan): Likewise.
13967 (arc_reorg): Call jli_call_scan.
13968 (arc_output_addsi): Remove 'S' from printing asm operand.
13969 (arc_is_jli_call_p): New function.
13970 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13972 (movhi_insn): Likewise.
13973 (movsi_insn): Likewise.
13974 (movsi_set_cc_insn): Likewise.
13975 (loadqi_update): Likewise.
13976 (load_zeroextendqisi_update): Likewise.
13977 (load_signextendqisi_update): Likewise.
13978 (loadhi_update): Likewise.
13979 (load_zeroextendhisi_update): Likewise.
13980 (load_signextendhisi_update): Likewise.
13981 (loadsi_update): Likewise.
13982 (loadsf_update): Likewise.
13983 (movsicc_insn): Likewise.
13984 (bset_insn): Likewise.
13985 (bxor_insn): Likewise.
13986 (bclr_insn): Likewise.
13987 (bmsk_insn): Likewise.
13988 (bicsi3_insn): Likewise.
13989 (cmpsi_cc_c_insn): Likewise.
13990 (movsi_ne): Likewise.
13991 (movsi_cond_exec): Likewise.
13992 (clrsbsi2): Likewise.
13993 (norm_f): Likewise.
14003 (call_i): Remove 'S' asm letter, add jli instruction.
14004 (call_value_i): Likewise.
14005 * config/arc/arc.op (mjli-always): New option.
14006 * config/arc/constraints.md (Cji): New constraint.
14007 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14009 (subsf3_fpx): Likewise.
14010 (mulsf3_fpx): Likewise.
14011 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14013 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14014 function attrbutes.
14015 * doc/invoke.texi (ARC): Document mjli-always option.
14017 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14019 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14020 avoid addition with 0 and use incw and decw where possible.
14022 2018-01-26 Richard Biener <rguenther@suse.de>
14024 PR tree-optimization/81082
14025 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14026 association if it requires casting to unsigned.
14027 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14028 from fold_plusminus_mult_expr to catch important cases late when
14029 range info is available.
14031 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14033 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14034 * configure.ac (hidden_linkonce): New test.
14035 * configure: Regenerate.
14036 * config.in: Regenerate.
14038 2018-01-26 Julia Koval <julia.koval@intel.com>
14040 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14041 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14042 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14043 _mm_mask_bitshuffle_epi64_mask): Fix type.
14044 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14045 USI_FTYPE_V4DI_V4DI_USI): Remove.
14046 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14047 __builtin_ia32_vpshufbitqmb256_mask,
14048 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14049 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14050 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14052 2018-01-26 Alan Modra <amodra@gmail.com>
14055 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14056 UNSPEC_VBPERMQ. Sort other unspecs.
14058 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14060 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14062 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14064 PR middle-end/83055
14065 * predict.c (drop_profile): Do not push/pop cfun; update also
14067 (handle_missing_profiles): Fix logic looking for zero profiles.
14069 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14071 PR middle-end/83977
14072 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14073 on functions with #pragma omp declare simd or functions with simd
14075 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14076 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14077 Remove trailing \n from warning_at calls.
14079 2018-01-25 Tom de Vries <tom@codesourcery.com>
14082 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14083 for neutered workers.
14085 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14088 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14089 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14091 2018-01-24 Jeff Law <law@redhat.com>
14094 * i386.c (get_probe_interval): Move to earlier point.
14095 (ix86_compute_frame_layout): If -fstack-clash-protection and
14096 the frame is larger than the probe interval, then use pushes
14097 to save registers rather than reg->mem moves.
14098 (ix86_expand_prologue): Remove conditional for int_registers_saved
14101 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14104 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14105 min/max for never referenced object.
14107 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14109 PR middle-end/83977
14110 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14112 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14113 attributes from DECL_ATTRIBUTES (decl) without affecting
14114 DECL_ATTRIBUTES (current_function_decl).
14115 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14116 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14118 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14120 PR tree-optimization/83979
14121 * fold-const.c (fold_comparison): Use constant_boolean_node
14122 instead of boolean_{true,false}_node.
14124 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14126 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14129 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14131 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14132 Simplify the clause that sets the length attribute.
14133 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14134 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14135 clause that sets the length attribute.
14136 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14138 2018-01-24 Tom de Vries <tom@codesourcery.com>
14141 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14142 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14143 Add strict parameter.
14144 (prevent_branch_around_nothing): Insert dummy insn between branch to
14145 label and label with no ptx insn inbetween.
14146 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14148 2018-01-24 Tom de Vries <tom@codesourcery.com>
14151 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14152 for neutered threads in warp.
14153 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14155 2018-01-24 Richard Biener <rguenther@suse.de>
14157 PR tree-optimization/83176
14158 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14161 2018-01-24 Richard Biener <rguenther@suse.de>
14163 PR tree-optimization/82819
14164 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14165 code generating pluses that are no-ops in the target precision.
14167 2018-01-24 Richard Biener <rguenther@suse.de>
14169 PR middle-end/84000
14170 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14172 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14174 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14175 to merge probabilities.
14176 * predict.c (probably_never_executed): Also mark as cold functions
14177 with global 0 profile and guessed local profile.
14178 * profile-count.c (profile_probability::combine_with_count): New
14180 * profile-count.h (profile_probability::operator*,
14181 profile_probability::operator*=, profile_probability::operator/,
14182 profile_probability::operator/=): Reduce precision to adjusted
14183 and set value to guessed on contradictory divisions.
14184 (profile_probability::combine_with_freq): Remove.
14185 (profile_probability::combine_wiht_count): Declare.
14186 (profile_count::force_nonzero):: Set to adjusted.
14187 (profile_count::probability_in):: Set quality to adjusted.
14188 * tree-ssa-tail-merge.c (replace_block_by): Use
14189 combine_with_count.
14191 2018-01-23 Andrew Waterman <andrew@sifive.com>
14192 Jim Wilson <jimw@sifive.com>
14194 * config/riscv/riscv.c (riscv_stack_boundary): New.
14195 (riscv_option_override): Set riscv_stack_boundary. Handle
14196 riscv_preferred_stack_boundary_arg.
14197 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14198 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14199 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14200 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14201 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14202 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14204 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14207 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14208 of struct ix86_frame.
14209 (ix86_expand_epilogue): Likewise. Add a local variable for
14210 the reg_save_offset field in struct ix86_frame.
14212 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14214 PR tree-optimization/82604
14215 * tree-loop-distribution.c (enum partition_kind): New enum item
14216 PKIND_PARTIAL_MEMSET.
14217 (partition_builtin_p): Support above new enum item.
14218 (generate_code_for_partition): Ditto.
14219 (compute_access_range): Differentiate cases that equality can be
14220 proven at all loops, the innermost loops or no loops.
14221 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14222 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14223 (finalize_partitions, distribute_loop): Don't fuse partition of
14224 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14225 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14226 parloop is enabled.
14228 2018-01-23 Martin Liska <mliska@suse.cz>
14230 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14231 order to ignore the predictor.
14232 (PRED_POLYMORPHIC_CALL): Likewise.
14233 (PRED_RECURSIVE_CALL): Likewise.
14235 2018-01-23 Martin Liska <mliska@suse.cz>
14237 * tree-profile.c (tree_profiling): Print function header to
14238 aware reader which function we are working on.
14239 * value-prof.c (gimple_find_values_to_profile): Do not print
14240 not interesting value histograms.
14242 2018-01-23 Martin Liska <mliska@suse.cz>
14244 * profile-count.h (enum profile_quality): Add
14245 profile_uninitialized as the first value. Do not number values
14246 as they are zero based.
14247 (profile_count::verify): Update sanity check.
14248 (profile_probability::verify): Likewise.
14250 2018-01-23 Nathan Sidwell <nathan@acm.org>
14252 * doc/invoke.texi (ffor-scope): Deprecate.
14254 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14256 PR tree-optimization/83510
14257 * domwalk.c (set_all_edges_as_executable): New function.
14258 (dom_walker::dom_walker): Convert bool param
14259 "skip_unreachable_blocks" to enum reachability. Move setup of
14260 edge flags to set_all_edges_as_executable and only do it when
14261 reachability is REACHABLE_BLOCKS.
14262 * domwalk.h (enum dom_walker::reachability): New enum.
14263 (dom_walker::dom_walker): Convert bool param
14264 "skip_unreachable_blocks" to enum reachability.
14265 (set_all_edges_as_executable): New decl.
14266 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14267 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14269 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14270 but converting true to REACHABLE_BLOCKS.
14271 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14273 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14274 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14275 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14277 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14278 if check_all_array_refs will be called.
14280 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14282 * tree.c (selftest::test_location_wrappers): Add more test
14285 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14287 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14288 (selftest::test_bit_in_range): Likewise.
14290 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14293 * doc/sourcebuild.texi (vect_float): Say that the selector
14294 only describes the situation when -funsafe-math-optimizations is on.
14295 (vect_float_strict): Document.
14297 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14299 PR tree-optimization/83965
14300 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14301 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14302 instead of checking only for a reduction.
14303 (vect_recog_widen_sum_pattern): Likewise.
14305 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14307 * predict.c (probably_never_executed): Only use precise profile info.
14308 (compute_function_frequency): Skip after inlining hack since we now
14309 have quality checking.
14311 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14313 * profile-count.h (profile_probability::very_unlikely,
14314 profile_probability::unlikely, profile_probability::even): Set
14315 precision to guessed.
14317 2018-01-23 Richard Biener <rguenther@suse.de>
14319 PR tree-optimization/83963
14320 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14321 Properly terminate dominator walk when crossing the exit edge not
14322 when visiting its source block.
14324 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14327 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14328 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14330 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14332 PR tree-optimization/83957
14333 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14334 semicolon after for body surrounded by braces.
14336 PR tree-optimization/83081
14337 * profile-count.h (profile_probability::split): New method.
14338 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14339 Use profile_probability::split.
14340 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14341 when splitting a single conditional jump into 2.
14343 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14345 PR tree-optimization/69452
14346 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14349 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14351 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14352 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14353 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14355 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14357 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14359 * config/rl78/rl78.md (movdi): New define_expand.
14360 * config/rl78/rl78.c (rl78_split_movdi): New function.
14362 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14365 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14367 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14368 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14369 128-bit to produce an UNSPEC move to get the double word with the
14370 signbit and then a shift directly to do signbit.
14371 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14372 implementation with a new version that just does either a direct
14373 move or a regular move. Move memory interface to separate insns.
14374 Move insns so they are next to the expander.
14375 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14376 with signbit move. Split big and little endian case.
14377 (signbit<mode>2_dm_mem_le): Likewise.
14378 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14379 (signbit<mode>2_dm2): Likewise.
14381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14383 * config/rl78/rl78.md (anddi3): New define_expand.
14385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14387 * config/rl78/rl78.md (umindi3): New define_expand.
14389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14391 * config/rl78/rl78.md (smindi3): New define_expand.
14393 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14395 * config/rl78/rl78.md (smaxdi3): New define_expand.
14397 2018-01-22 Carl Love <cel@us.ibm.com>
14399 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14400 LVX_V1TI): Add macro expansion.
14401 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14402 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14403 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14404 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14405 Change check to determine if the instruction is a byte reversing
14406 entry. Fix typo in comment.
14407 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14408 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14409 Add def_builtin calls for new builtins.
14410 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14411 Add define_insn expansion.
14413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14415 * config/rl78/rl78.md (umaxdi3): New define_expand.
14417 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14419 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14420 for non-QImode registers.
14422 2018-01-22 Richard Biener <rguenther@suse.de>
14424 PR tree-optimization/83963
14425 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14426 including the loop exit block.
14427 (scop_detection::merge_sese): Likewise.
14428 (scop_detection::add_scop): Do it here instead.
14430 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14432 * doc/sourcebuild.texi (arm_softfloat): Document.
14434 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14437 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14438 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14439 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14441 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14442 David Edelsohn <dje.gcc@gmail.com>
14445 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14446 Change "crset eq" to "crset 2".
14447 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14448 (*call_indirect_aix<mode>_nospec): Likewise.
14449 (*call_value_indirect_aix<mode>_nospec): Likewise.
14450 (*call_indirect_elfv2<mode>_nospec): Likewise.
14451 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14452 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14453 change assembly output from . to $.
14454 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14455 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14456 (*tablejump<mode>_internal1_nospec): Likewise.
14458 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14461 * config/sh/sh_optimize_sett_clrt.cc:
14462 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14464 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14466 PR tree-optimization/83940
14467 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14468 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14469 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14470 use it to pass back the definition type.
14471 (vect_check_store_rhs): Likewise rhs_dt_out.
14472 (vect_build_gather_load_calls): Add a mask_dt argument and use
14473 it instead of a call to vect_is_simple_use.
14474 (vectorizable_store): Update calls to vect_check_load_store_mask
14475 and vect_check_store_rhs. Use the dt returned by the latter instead
14476 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14477 instead of calls to vect_is_simple_use. Pass the scalar rather
14478 than the vector operand to vect_is_simple_use when handling
14479 second and subsequent copies of an rhs value.
14480 (vectorizable_load): Update calls to vect_check_load_store_mask
14481 and vect_build_gather_load_calls. Use the cached mask_dt and
14482 gs_info.offset_dt instead of calls to vect_is_simple_use.
14484 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14486 PR middle-end/83945
14487 * tree-emutls.c: Include gimplify.h.
14488 (lower_emutls_2): New function.
14489 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14490 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14491 it before further processing.
14494 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14495 UINTVAL (trueop1) instead of INTVAL (op1).
14497 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14501 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14502 INCOMING_FRAME_SP_OFFSET if not defined.
14503 (scan_trace): Add ENTRY argument. If true and
14504 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14505 emit a note to adjust the CFA offset.
14506 (create_cfi_notes): Adjust scan_trace callers.
14507 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14508 INCOMING_FRAME_SP_OFFSET in the CIE.
14509 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14510 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14512 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14513 * doc/tm.texi: Regenerated.
14515 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14517 PR rtl-optimization/83147
14518 * lra-constraints.c (remove_inheritance_pseudos): Use
14519 lra_substitute_pseudo_within_insn.
14521 2018-01-19 Tom de Vries <tom@codesourcery.com>
14522 Cesar Philippidis <cesar@codesourcery.com>
14525 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14527 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14530 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14531 spaces for function labels.
14533 2018-01-19 Martin Liska <mliska@suse.cz>
14535 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14536 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14537 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14538 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14539 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14540 (PRED_CONST_RETURN): Change from 69 to 65.
14541 (PRED_NULL_RETURN): Change from 91 to 71.
14542 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14543 (PRED_LOOP_GUARD): Change from 66 to 73.
14545 2018-01-19 Martin Liska <mliska@suse.cz>
14547 * predict.c (predict_insn_def): Add new assert.
14548 (struct branch_predictor): Change type to signed integer.
14549 (test_prediction_value_range): Amend test to cover
14550 PROB_UNINITIALIZED.
14551 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14552 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14553 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14554 (PRED_LOOP_IV_COMPARE): Likewise.
14555 * predict.h (PROB_UNINITIALIZED): Define new constant.
14557 2018-01-19 Martin Liska <mliska@suse.cz>
14559 * predict.c (dump_prediction): Add new format for
14560 analyze_brprob.py script which is enabled with -details
14562 * profile-count.h (precise_p): New function.
14564 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14566 PR tree-optimization/83922
14567 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14568 there are no statements that need masking.
14569 (vect_active_double_reduction_p): New function.
14570 (vect_analyze_loop_operations): Use it when handling phis that
14571 are not in the loop header.
14573 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14575 PR tree-optimization/83914
14576 * tree-vect-loop.c (vectorizable_induction): Don't convert
14577 init_expr or apply the peeling adjustment for inductions
14578 that are nested within the vectorized loop.
14580 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14582 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14585 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14589 * function.h (gimplify_parameters): Add gimple_seq * argument.
14590 * function.c: Include gimple.h and options.h.
14591 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14592 for the added local temporaries if needed.
14593 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14594 if there are any parameter cleanups, wrap whole body into a
14595 try/finally with the cleanups.
14597 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14600 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14601 Use GET_MODE_CLASS for scalar floating point.
14603 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14607 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14608 Fix call of call_cgraph_insertion_hooks.
14610 2018-01-18 Martin Sebor <msebor@redhat.com>
14612 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14614 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14617 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14620 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14623 * common.opt: (-ffile-prefix-map): New option.
14624 * opts.c (common_handle_option): Defer it.
14625 * opts-global.c (handle_common_deferred_options): Handle it.
14626 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14627 * file-prefix-map.h: New file.
14628 (remap_debug_filename, add_debug_prefix_map): ...here.
14629 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14630 * final.c (debug_prefix_map, add_debug_prefix_map
14631 remap_debug_filename): Move to...
14632 * file-prefix-map.c: New file.
14633 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14634 generalize, get rid of alloca(), use strrchr() instead of strchr().
14635 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14636 Implement in terms of add_prefix_map().
14637 (remap_macro_filename, remap_debug_filename): Implement in term of
14639 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14640 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14641 * dbxout.c: Include file-prefix-map.h.
14642 * varasm.c: Likewise.
14643 * vmsdbgout.c: Likewise.
14644 * xcoffout.c: Likewise.
14645 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14646 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14647 * doc/invoke.texi (-ffile-prefix-map): Document.
14648 (-fdebug-prefix-map): Update description.
14650 2018-01-18 Martin Liska <mliska@suse.cz>
14652 * config/i386/i386.c (indirect_thunk_name): Document that also
14654 (output_indirect_thunk): Document why both instructions
14655 (pause and lfence) are generated.
14657 2018-01-18 Richard Biener <rguenther@suse.de>
14659 PR tree-optimization/83887
14660 * graphite-scop-detection.c
14661 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14662 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14663 (scop_detection::merge_sese): Re-implement with a flood-fill
14664 algorithm that properly finds a SESE region if it exists.
14666 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14669 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14670 pointer_diff optimizations use view_convert instead of convert.
14672 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14674 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14675 Generate different code for -mno-speculate-indirect-jumps.
14676 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14677 (*call_indirect_aix<mode>): Disable for
14678 -mno-speculate-indirect-jumps.
14679 (*call_indirect_aix<mode>_nospec): New define_insn.
14680 (*call_value_indirect_aix<mode>): Disable for
14681 -mno-speculate-indirect-jumps.
14682 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14683 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14684 -mno-speculate-indirect-jumps.
14685 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14687 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14689 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14690 long double type, set the flags for noting the default long double
14691 type, even if we don't pass or return a long double type.
14693 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14696 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14699 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14701 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14702 support for merge[hl].
14703 (fold_mergehl_helper): New helper function.
14704 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14705 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14706 (altivec_vmrglw_direct): Add xxmrglw insn.
14708 2018-01-17 Andrew Waterman <andrew@sifive.com>
14710 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14711 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14713 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14716 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14717 call the lto_location_cache before reading the
14718 DECL_SOURCE_LOCATION of the types.
14720 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14721 Richard Sandiford <richard.sandiford@linaro.org>
14723 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14724 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14725 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14726 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14727 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14729 * config/aarch64/constraints.md (aarch64_movti_operand):
14731 * config/aarch64/predicates.md (Uti): Add new constraint.
14733 2018-01-17 Carl Love <cel@us.ibm.com>
14735 * config/rs6000/vsx.md (define_expand xl_len_r,
14736 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14737 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14739 (define_expand, define_insn): Move the shift left from the
14740 define_insn to the define_expand for lxvl and stxvl instructions.
14741 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14742 and XL_LEN_R definitions to PURE.
14744 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14746 * config/i386/i386.c (indirect_thunk_name): Declare regno
14747 as unsigned int. Compare regno with INVALID_REGNUM.
14748 (output_indirect_thunk): Ditto.
14749 (output_indirect_thunk_function): Ditto.
14750 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14751 in the call to output_indirect_thunk_function.
14753 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14755 PR middle-end/83884
14756 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14757 rather than the size of inner_type to determine the stack slot size
14758 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14760 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14763 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14766 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14768 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14769 endian Linux systems to optionally enable multilibs for selecting
14770 the long double type if the user configured an explicit type.
14771 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14772 have no long double multilibs if not defined.
14773 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14774 warn if the user used -mabi={ieee,ibm}longdouble and we built
14775 multilibs for long double.
14776 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14777 appropriate multilib option.
14778 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14780 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14781 for building long double multilibs.
14782 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14784 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14786 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14789 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14791 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14794 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14797 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14800 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14802 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14803 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14805 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14807 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14808 different rtl trees depending on TARGET_64BIT.
14809 (rs6000_gen_lvx): Likewise.
14811 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14813 * config/visium/visium.md (nop): Tweak comment.
14814 (hazard_nop): Likewise.
14816 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14818 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14819 -mspeculate-indirect-jumps.
14820 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14821 for -mno-speculate-indirect-jumps.
14822 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14823 (*call_value_indirect_elfv2<mode>): Disable for
14824 -mno-speculate-indirect-jumps.
14825 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14826 (indirect_jump): Emit different RTL for
14827 -mno-speculate-indirect-jumps.
14828 (*indirect_jump<mode>): Disable for
14829 -mno-speculate-indirect-jumps.
14830 (*indirect_jump<mode>_nospec): New define_insn.
14831 (tablejump): Emit different RTL for
14832 -mno-speculate-indirect-jumps.
14833 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14834 (tablejumpsi_nospec): New define_expand.
14835 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14836 (tablejumpdi_nospec): New define_expand.
14837 (*tablejump<mode>_internal1): Disable for
14838 -mno-speculate-indirect-jumps.
14839 (*tablejump<mode>_internal1_nospec): New define_insn.
14840 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14843 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14845 * caller-save.c (insert_save): Drop unnecessary parameter. All
14848 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14849 Richard Biener <rguenth@suse.de>
14852 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14853 return early, inline manually is_gimple_sizepos. Make sure if we
14854 call gimplify_expr we don't end up with a gimple constant.
14855 * tree.c (variably_modified_type_p): Don't return true for
14856 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14857 * gimplify.h (is_gimple_sizepos): Remove.
14859 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14861 PR tree-optimization/83857
14862 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14863 vectorizable_live_operation for pure SLP statements.
14864 (vectorizable_live_operation): Handle PHIs.
14866 2018-01-16 Richard Biener <rguenther@suse.de>
14868 PR tree-optimization/83867
14869 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14870 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14872 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14875 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14876 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14877 If off is not INTEGER_CST, issue a may not be aligned warning
14878 rather than isn't aligned. Use isn%'t rather than isn't.
14879 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14881 <case MULT_EXPR>: Improve the case when bottom and one of the
14882 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14883 operand, in that case check if the other operand is multiple of
14884 bottom divided by the INTEGER_CST operand.
14886 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14889 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14890 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14891 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14892 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14893 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14894 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14895 (pa_function_arg_size): New function.
14897 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14899 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14900 in a separate statement.
14902 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14904 PR tree-optimization/83847
14905 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14906 group gathers and scatters.
14908 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14910 PR rtl-optimization/86620
14911 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14913 PR rtl-optimization/83213
14914 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14915 to last if both are JUMP_INSNs.
14917 PR tree-optimization/83843
14918 * gimple-ssa-store-merging.c
14919 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14920 store_immediate_info for bswap/nop orig_stores.
14922 2018-01-15 Andrew Waterman <andrew@sifive.com>
14924 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14926 <UDIV>: Increase cost if !TARGET_DIV.
14928 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14930 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14931 (define_attr "cr_logical_3op"): New.
14932 (cceq_ior_compare): Adjust.
14933 (cceq_ior_compare_complement): Adjust.
14934 (*cceq_rev_compare): Adjust.
14935 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14936 (is_cracked_insn): Adjust.
14937 (insn_must_be_first_in_group): Adjust.
14938 * config/rs6000/40x.md: Adjust.
14939 * config/rs6000/440.md: Adjust.
14940 * config/rs6000/476.md: Adjust.
14941 * config/rs6000/601.md: Adjust.
14942 * config/rs6000/603.md: Adjust.
14943 * config/rs6000/6xx.md: Adjust.
14944 * config/rs6000/7450.md: Adjust.
14945 * config/rs6000/7xx.md: Adjust.
14946 * config/rs6000/8540.md: Adjust.
14947 * config/rs6000/cell.md: Adjust.
14948 * config/rs6000/e300c2c3.md: Adjust.
14949 * config/rs6000/e500mc.md: Adjust.
14950 * config/rs6000/e500mc64.md: Adjust.
14951 * config/rs6000/e5500.md: Adjust.
14952 * config/rs6000/e6500.md: Adjust.
14953 * config/rs6000/mpc.md: Adjust.
14954 * config/rs6000/power4.md: Adjust.
14955 * config/rs6000/power5.md: Adjust.
14956 * config/rs6000/power6.md: Adjust.
14957 * config/rs6000/power7.md: Adjust.
14958 * config/rs6000/power8.md: Adjust.
14959 * config/rs6000/power9.md: Adjust.
14960 * config/rs6000/rs64.md: Adjust.
14961 * config/rs6000/titan.md: Adjust.
14963 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14965 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14966 ix86_indirect_branch_register logic.
14968 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14970 * config/i386/constraints.md (Bs): Update
14971 ix86_indirect_branch_register check. Don't check
14972 ix86_indirect_branch_register with GOT_memory_operand.
14974 * config/i386/predicates.md (GOT_memory_operand): Don't check
14975 ix86_indirect_branch_register here.
14976 (GOT32_symbol_operand): Likewise.
14978 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14980 * config/i386/predicates.md (constant_call_address_operand):
14981 Rewrite ix86_indirect_branch_register logic.
14982 (sibcall_insn_operand): Likewise.
14984 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14986 * config/i386/constraints.md (Bs): Replace
14987 ix86_indirect_branch_thunk_register with
14988 ix86_indirect_branch_register.
14990 * config/i386/i386.md (indirect_jump): Likewise.
14991 (tablejump): Likewise.
14992 (*sibcall_memory): Likewise.
14993 (*sibcall_value_memory): Likewise.
14994 Peepholes of indirect call and jump via memory: Likewise.
14995 * config/i386/i386.opt: Likewise.
14996 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14997 (GOT_memory_operand): Likewise.
14998 (call_insn_operand): Likewise.
14999 (sibcall_insn_operand): Likewise.
15000 (GOT32_symbol_operand): Likewise.
15002 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15004 PR middle-end/83837
15005 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15006 type rather than type addr's type points to.
15007 (expand_omp_atomic_mutex): Likewise.
15008 (expand_omp_atomic): Likewise.
15010 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15013 * config/i386/i386.c (output_indirect_thunk_function): Use
15014 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15015 for __x86_return_thunk.
15017 2018-01-15 Richard Biener <rguenther@suse.de>
15019 PR middle-end/83850
15020 * expmed.c (extract_bit_field_1): Fix typo.
15022 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15025 * config/arm/iterators.md (VF): New mode iterator.
15026 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15027 Remove integer-related logic from pattern.
15028 (neon_vabd<mode>_3): Likewise.
15030 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15032 PR middle-end/82694
15033 * common.opt (fstrict-overflow): No longer an alias.
15034 (fwrapv-pointer): New option.
15035 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15036 also for pointer types based on flag_wrapv_pointer.
15037 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15038 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15039 opts->x_flag_wrapv got set.
15040 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15041 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15042 POINTER_TYPE_OVERFLOW_UNDEFINED.
15043 * match.pd: Likewise in address comparison pattern.
15044 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15046 2018-01-15 Richard Biener <rguenther@suse.de>
15049 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15050 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15051 Reset type names to their identifier if their TYPE_DECL doesn't
15052 have linkage (and thus is used for ODR and devirt).
15053 (save_debug_info_for_decl): Remove.
15054 (save_debug_info_for_type): Likewise.
15055 (add_tree_to_fld_list): Adjust.
15056 * tree-pretty-print.c (dump_generic_node): Make dumping of
15057 type names more robust.
15059 2018-01-15 Richard Biener <rguenther@suse.de>
15061 * BASE-VER: Bump to 8.0.1.
15063 2018-01-14 Martin Sebor <msebor@redhat.com>
15066 * builtins.c (check_access): Avoid warning when the no-warning bit
15069 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15071 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15072 * ira-color (allocno_hard_regs_compare): Likewise.
15074 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15077 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15078 Use .pushsection/.popsection.
15080 2018-01-14 Martin Sebor <msebor@redhat.com>
15083 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15085 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15087 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15088 entry from extra_headers.
15089 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15090 extra_headers, make the list bitwise identical to the i?86-*-* one.
15092 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15094 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15095 -mcmodel=large with -mindirect-branch=thunk,
15096 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15097 -mfunction-return=thunk-extern.
15098 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15099 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15100 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15102 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15104 * config/i386/i386.c (print_reg): Print the name of the full
15105 integer register without '%'.
15106 (ix86_print_operand): Handle 'V'.
15107 * doc/extend.texi: Document 'V' modifier.
15109 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15111 * config/i386/constraints.md (Bs): Disallow memory operand for
15112 -mindirect-branch-register.
15114 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15115 (GOT_memory_operand): Likewise.
15116 (call_insn_operand): Likewise.
15117 (sibcall_insn_operand): Likewise.
15118 (GOT32_symbol_operand): Likewise.
15119 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15120 for -mindirect-branch-register.
15121 (tablejump): Likewise.
15122 (*sibcall_memory): Likewise.
15123 (*sibcall_value_memory): Likewise.
15124 Disallow peepholes of indirect call and jump via memory for
15125 -mindirect-branch-register.
15126 (*call_pop): Replace m with Bw.
15127 (*call_value_pop): Likewise.
15128 (*sibcall_pop_memory): Replace m with Bs.
15129 * config/i386/i386.opt (mindirect-branch-register): New option.
15130 * doc/invoke.texi: Document -mindirect-branch-register option.
15132 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15134 * config/i386/i386-protos.h (ix86_output_function_return): New.
15135 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15136 set function_return_type.
15137 (indirect_thunk_name): Add ret_p to indicate thunk for function
15139 (output_indirect_thunk_function): Pass false to
15140 indirect_thunk_name.
15141 (ix86_output_indirect_branch_via_reg): Likewise.
15142 (ix86_output_indirect_branch_via_push): Likewise.
15143 (output_indirect_thunk_function): Create alias for function
15144 return thunk if regno < 0.
15145 (ix86_output_function_return): New function.
15146 (ix86_handle_fndecl_attribute): Handle function_return.
15147 (ix86_attribute_table): Add function_return.
15148 * config/i386/i386.h (machine_function): Add
15149 function_return_type.
15150 * config/i386/i386.md (simple_return_internal): Use
15151 ix86_output_function_return.
15152 (simple_return_internal_long): Likewise.
15153 * config/i386/i386.opt (mfunction-return=): New option.
15154 (indirect_branch): Mention -mfunction-return=.
15155 * doc/extend.texi: Document function_return function attribute.
15156 * doc/invoke.texi: Document -mfunction-return= option.
15158 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15160 * config/i386/i386-opts.h (indirect_branch): New.
15161 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15162 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15163 with local indirect jump when converting indirect call and jump.
15164 (ix86_set_indirect_branch_type): New.
15165 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15166 (indirectlabelno): New.
15167 (indirect_thunk_needed): Likewise.
15168 (indirect_thunk_bnd_needed): Likewise.
15169 (indirect_thunks_used): Likewise.
15170 (indirect_thunks_bnd_used): Likewise.
15171 (INDIRECT_LABEL): Likewise.
15172 (indirect_thunk_name): Likewise.
15173 (output_indirect_thunk): Likewise.
15174 (output_indirect_thunk_function): Likewise.
15175 (ix86_output_indirect_branch_via_reg): Likewise.
15176 (ix86_output_indirect_branch_via_push): Likewise.
15177 (ix86_output_indirect_branch): Likewise.
15178 (ix86_output_indirect_jmp): Likewise.
15179 (ix86_code_end): Call output_indirect_thunk_function if needed.
15180 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15182 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15183 (ix86_attribute_table): Add indirect_branch.
15184 * config/i386/i386.h (machine_function): Add indirect_branch_type
15185 and has_local_indirect_jump.
15186 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15188 (tablejump): Likewise.
15189 (*indirect_jump): Use ix86_output_indirect_jmp.
15190 (*tablejump_1): Likewise.
15191 (simple_return_indirect_internal): Likewise.
15192 * config/i386/i386.opt (mindirect-branch=): New option.
15193 (indirect_branch): New.
15196 (thunk-inline): Likewise.
15197 (thunk-extern): Likewise.
15198 * doc/extend.texi: Document indirect_branch function attribute.
15199 * doc/invoke.texi: Document -mindirect-branch= option.
15201 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15204 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15206 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15208 * ipa-inline.c (want_inline_small_function_p): Return false if
15209 inlining has already failed with CIF_FINAL_ERROR.
15210 (update_caller_keys): Call want_inline_small_function_p before
15212 (update_callee_keys): Likewise.
15214 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15216 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15218 (rs6000_quadword_masked_address_p): Likewise.
15219 (quad_aligned_load_p): Likewise.
15220 (quad_aligned_store_p): Likewise.
15221 (const_load_sequence_p): Add comment to describe the outer-most loop.
15222 (mimic_memory_attributes_and_flags): New function.
15223 (rs6000_gen_stvx): Likewise.
15224 (replace_swapped_aligned_store): Likewise.
15225 (rs6000_gen_lvx): Likewise.
15226 (replace_swapped_aligned_load): Likewise.
15227 (replace_swapped_load_constant): Capitalize argument name in
15228 comment describing this function.
15229 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15230 and stores that access quad-word aligned addresses and replace
15231 with stvx or lvx instructions when appropriate.
15232 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15233 New function prototype.
15234 (rs6000_quadword_masked_address_p): Likewise.
15235 (rs6000_gen_lvx): Likewise.
15236 (rs6000_gen_stvx): Likewise.
15237 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15238 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15239 when memory address is aligned.
15240 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15241 this split to select lvx instruction when memory address is aligned.
15242 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15243 instruction when memory address is aligned.
15244 (*vsx_le_perm_load_v16qi): Likewise.
15245 (four unnamed splitters): Modify to select the stvx instruction
15246 when memory is aligned.
15248 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15250 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15251 which appears in the queue multiple times.
15253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15254 Alan Hayward <alan.hayward@arm.com>
15255 David Sherwood <david.sherwood@arm.com>
15257 * tree-vectorizer.h (vec_lower_bound): New structure.
15258 (_loop_vec_info): Add check_nonzero and lower_bounds.
15259 (LOOP_VINFO_CHECK_NONZERO): New macro.
15260 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15261 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15262 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15263 fields. Make seg_len the distance travelled, not including the
15265 (dr_direction_indicator): Declare.
15266 (dr_zero_step_indicator): Likewise.
15267 (dr_known_forward_stride_p): Likewise.
15268 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15270 (runtime_alias_check_p): Allow runtime alias checks with
15272 (operator ==): Compare access_size and align.
15273 (prune_runtime_alias_test_list): Rework for new distinction between
15274 the access_size and seg_len.
15275 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15277 (get_segment_min_max): New function.
15278 (create_intersect_range_checks): Use it.
15279 (dr_step_indicator): New function.
15280 (dr_direction_indicator): Likewise.
15281 (dr_zero_step_indicator): Likewise.
15282 (dr_known_forward_stride_p): Likewise.
15283 * tree-loop-distribution.c (data_ref_segment_size): Return
15284 DR_STEP * (niters - 1).
15285 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15287 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15288 (vect_preserves_scalar_order_p): New function, split out from...
15289 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15290 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15291 (vect_vfa_access_size): New function.
15292 (vect_vfa_align): Likewise.
15293 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15294 (dump_lower_bound): New function.
15295 (vect_check_lower_bound): Likewise.
15296 (vect_small_gap_p): Likewise.
15297 (vectorizable_with_step_bound_p): Likewise.
15298 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15299 depencies if the vectorization factor is 1. Convert the checks
15300 for nonzero steps into checks on the bounds of DR_STEP. Try using
15301 a bunds check for variable steps if the minimum required step is
15302 relatively small. Update calls to the dr_with_seg_len
15303 constructor and to vect_compile_time_alias.
15304 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15306 (vect_loop_versioning): Call it.
15307 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15309 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15312 Alan Hayward <alan.hayward@arm.com>
15313 David Sherwood <david.sherwood@arm.com>
15315 * doc/sourcebuild.texi (vect_scatter_store): Document.
15316 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15318 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15320 * genopinit.c (main): Add supports_vec_scatter_store and
15321 supports_vec_scatter_store_cached to target_optabs.
15322 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15323 IFN_MASK_SCATTER_STORE.
15324 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15326 * internal-fn.h (internal_store_fn_p): Declare.
15327 (internal_fn_stored_value_index): Likewise.
15328 * internal-fn.c (scatter_store_direct): New macro.
15329 (expand_scatter_store_optab_fn): New function.
15330 (direct_scatter_store_optab_supported_p): New macro.
15331 (internal_store_fn_p): New function.
15332 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15333 IFN_MASK_SCATTER_STORE.
15334 (internal_fn_mask_index): Likewise.
15335 (internal_fn_stored_value_index): New function.
15336 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15337 for scatter stores.
15338 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15339 * optabs-query.c (supports_vec_scatter_store_p): New function.
15340 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15341 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15342 true for scatter stores.
15343 (vect_gather_scatter_fn_p): Handle scatter stores too.
15344 (vect_check_gather_scatter): Consider using scatter stores if
15345 supports_vec_scatter_store_p.
15346 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15347 scatter stores too.
15348 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15349 internal_fn_stored_value_index.
15350 (check_load_store_masking): Handle scatter stores too.
15351 (vect_get_store_rhs): Make public.
15352 (vectorizable_call): Use internal_store_fn_p.
15353 (vectorizable_store): Handle scatter store internal functions.
15354 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15355 when deciding whether the end of the group has been reached.
15356 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15357 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15358 (mask_scatter_store<mode>): New insns.
15360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15361 Alan Hayward <alan.hayward@arm.com>
15362 David Sherwood <david.sherwood@arm.com>
15364 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15365 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15366 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15368 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15369 Use vect_truncate_gather_scatter_offset if we can't treat the
15370 operation as a normal gather load or scatter store.
15371 (get_group_load_store_type): Take the gather_scatter_info
15372 as argument. Try using a gather load or scatter store for
15373 single-element groups.
15374 (get_load_store_type): Update calls to get_group_load_store_type
15375 and vect_use_strided_gather_scatters_p.
15377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15378 Alan Hayward <alan.hayward@arm.com>
15379 David Sherwood <david.sherwood@arm.com>
15381 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15382 optional tree argument.
15383 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15385 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15386 but continue to use the current value as a fallback.
15387 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15388 to compare the updates.
15389 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15390 (get_load_store_type): Use it when handling a strided access.
15391 (vect_get_strided_load_store_ops): New function.
15392 (vect_get_data_ptr_increment): Likewise.
15393 (vectorizable_load): Handle strided gather loads. Always pass
15394 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15397 Alan Hayward <alan.hayward@arm.com>
15398 David Sherwood <david.sherwood@arm.com>
15400 * doc/md.texi (gather_load@var{m}): Document.
15401 (mask_gather_load@var{m}): Likewise.
15402 * genopinit.c (main): Add supports_vec_gather_load and
15403 supports_vec_gather_load_cached to target_optabs.
15404 * optabs-tree.c (init_tree_optimization_optabs): Use
15405 ggc_cleared_alloc to allocate target_optabs.
15406 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15407 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15409 * internal-fn.h (internal_load_fn_p): Declare.
15410 (internal_gather_scatter_fn_p): Likewise.
15411 (internal_fn_mask_index): Likewise.
15412 (internal_gather_scatter_fn_supported_p): Likewise.
15413 * internal-fn.c (gather_load_direct): New macro.
15414 (expand_gather_load_optab_fn): New function.
15415 (direct_gather_load_optab_supported_p): New macro.
15416 (direct_internal_fn_optab): New function.
15417 (internal_load_fn_p): Likewise.
15418 (internal_gather_scatter_fn_p): Likewise.
15419 (internal_fn_mask_index): Likewise.
15420 (internal_gather_scatter_fn_supported_p): Likewise.
15421 * optabs-query.c (supports_at_least_one_mode_p): New function.
15422 (supports_vec_gather_load_p): Likewise.
15423 * optabs-query.h (supports_vec_gather_load_p): Declare.
15424 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15425 and memory_type field.
15426 (NUM_PATTERNS): Bump to 15.
15427 * tree-vect-data-refs.c: Include internal-fn.h.
15428 (vect_gather_scatter_fn_p): New function.
15429 (vect_describe_gather_scatter_call): Likewise.
15430 (vect_check_gather_scatter): Try using internal functions for
15431 gather loads. Recognize existing calls to a gather load function.
15432 (vect_analyze_data_refs): Consider using gather loads if
15433 supports_vec_gather_load_p.
15434 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15435 (vect_get_gather_scatter_offset_type): Likewise.
15436 (vect_convert_mask_for_vectype): Likewise.
15437 (vect_add_conversion_to_patterm): Likewise.
15438 (vect_try_gather_scatter_pattern): Likewise.
15439 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15440 (vect_vect_recog_func_ptrs): Add it.
15441 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15442 internal_fn_mask_index and internal_gather_scatter_fn_p.
15443 (check_load_store_masking): Take the gather_scatter_info as an
15444 argument and handle gather loads.
15445 (vect_get_gather_scatter_ops): New function.
15446 (vectorizable_call): Check internal_load_fn_p.
15447 (vectorizable_load): Likewise. Handle gather load internal
15449 (vectorizable_store): Update call to check_load_store_masking.
15450 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15451 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15452 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15453 (aarch64_gather_scale_operand_d): New predicates.
15454 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15455 (mask_gather_load<mode>): New insns.
15457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15458 Alan Hayward <alan.hayward@arm.com>
15459 David Sherwood <david.sherwood@arm.com>
15461 * optabs.def (fold_left_plus_optab): New optab.
15462 * doc/md.texi (fold_left_plus_@var{m}): Document.
15463 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15464 * internal-fn.c (fold_left_direct): Define.
15465 (expand_fold_left_optab_fn): Likewise.
15466 (direct_fold_left_optab_supported_p): Likewise.
15467 * fold-const-call.c (fold_const_fold_left): New function.
15468 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15469 * tree-parloops.c (valid_reduction_p): New function.
15470 (gather_scalar_reductions): Use it.
15471 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15472 (vect_finish_replace_stmt): Declare.
15473 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15474 (needs_fold_left_reduction_p): New function, split out from...
15475 (vect_is_simple_reduction): ...here. Accept reductions that
15476 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15477 (vect_force_simple_reduction): Also store the reduction type in
15478 the assignment's STMT_VINFO_REDUC_TYPE.
15479 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15480 (merge_with_identity): New function.
15481 (vect_expand_fold_left): Likewise.
15482 (vectorize_fold_left_reduction): Likewise.
15483 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15484 scalar phi in place for it. Check for target support and reject
15485 cases that would reassociate the operation. Defer the transform
15486 phase to vectorize_fold_left_reduction.
15487 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15488 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15489 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15493 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15494 call to ifc_temp_var.
15496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15497 Alan Hayward <alan.hayward@arm.com>
15498 David Sherwood <david.sherwood@arm.com>
15500 * target.def (legitimize_address_displacement): Take the original
15501 offset as a poly_int.
15502 * targhooks.h (default_legitimize_address_displacement): Update
15504 * targhooks.c (default_legitimize_address_displacement): Likewise.
15505 * doc/tm.texi: Regenerate.
15506 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15507 as an argument, moving assert of ad->disp == ad->disp_term to...
15508 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15509 Try calling targetm.legitimize_address_displacement before expanding
15510 the address rather than afterwards, and adjust for the new interface.
15511 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15512 Match the new hook interface. Handle SVE addresses.
15513 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15514 new hook interface.
15516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15518 * Makefile.in (OBJS): Add early-remat.o.
15519 * target.def (select_early_remat_modes): New hook.
15520 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15521 * doc/tm.texi: Regenerate.
15522 * targhooks.h (default_select_early_remat_modes): Declare.
15523 * targhooks.c (default_select_early_remat_modes): New function.
15524 * timevar.def (TV_EARLY_REMAT): New timevar.
15525 * passes.def (pass_early_remat): New pass.
15526 * tree-pass.h (make_pass_early_remat): Declare.
15527 * early-remat.c: New file.
15528 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15530 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15533 Alan Hayward <alan.hayward@arm.com>
15534 David Sherwood <david.sherwood@arm.com>
15536 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15537 vfm1 with a bound_epilog parameter.
15538 (vect_do_peeling): Update calls accordingly, and move the prologue
15539 call earlier in the function. Treat the base bound_epilog as 0 for
15540 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15541 this base when peeling for gaps.
15542 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15543 with fully-masked loops.
15544 (vect_estimate_min_profitable_iters): Handle the single peeled
15545 iteration in that case.
15547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15548 Alan Hayward <alan.hayward@arm.com>
15549 David Sherwood <david.sherwood@arm.com>
15551 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15552 single-element interleaving even if the size is not a power of 2.
15553 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15554 accesses for single-element interleaving if the group size is
15557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15558 Alan Hayward <alan.hayward@arm.com>
15559 David Sherwood <david.sherwood@arm.com>
15561 * doc/md.texi (fold_extract_last_@var{m}): Document.
15562 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15563 * optabs.def (fold_extract_last_optab): New optab.
15564 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15565 * internal-fn.c (fold_extract_direct): New macro.
15566 (expand_fold_extract_optab_fn): Likewise.
15567 (direct_fold_extract_optab_supported_p): Likewise.
15568 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15569 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15570 EXTRACT_LAST_REDUCTION.
15571 (get_initial_def_for_reduction): Do not create an initial vector
15572 for EXTRACT_LAST_REDUCTION reductions.
15573 (vectorizable_reduction): Leave the scalar phi in place for
15574 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15575 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15576 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15577 transform phase to vectorizable_condition.
15578 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15580 (vect_finish_stmt_generation): ...here.
15581 (vect_finish_replace_stmt): New function.
15582 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15583 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15585 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15588 Alan Hayward <alan.hayward@arm.com>
15589 David Sherwood <david.sherwood@arm.com>
15591 * doc/md.texi (extract_last_@var{m}): Document.
15592 * optabs.def (extract_last_optab): New optab.
15593 * internal-fn.def (EXTRACT_LAST): New internal function.
15594 * internal-fn.c (cond_unary_direct): New macro.
15595 (expand_cond_unary_optab_fn): Likewise.
15596 (direct_cond_unary_optab_supported_p): Likewise.
15597 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15598 loops using EXTRACT_LAST.
15599 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15600 (extract_last_<mode>): ...this optab.
15601 (vec_extract<mode><Vel>): Update accordingly.
15603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15604 Alan Hayward <alan.hayward@arm.com>
15605 David Sherwood <david.sherwood@arm.com>
15607 * target.def (empty_mask_is_expensive): New hook.
15608 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15609 * doc/tm.texi: Regenerate.
15610 * targhooks.h (default_empty_mask_is_expensive): Declare.
15611 * targhooks.c (default_empty_mask_is_expensive): New function.
15612 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15613 if the target says that empty masks are expensive.
15614 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15616 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15619 Alan Hayward <alan.hayward@arm.com>
15620 David Sherwood <david.sherwood@arm.com>
15622 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15623 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15624 (vect_use_loop_mask_for_alignment_p): New function.
15625 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15626 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15627 niters_skip argument. Make sure that the first niters_skip elements
15628 of the first iteration are inactive.
15629 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15630 Update call to vect_set_loop_masks_directly.
15631 (get_misalign_in_elems): New function, split out from...
15632 (vect_gen_prolog_loop_niters): ...here.
15633 (vect_update_init_of_dr): Take a code argument that specifies whether
15634 the adjustment should be added or subtracted.
15635 (vect_update_init_of_drs): Likewise.
15636 (vect_prepare_for_masked_peels): New function.
15637 (vect_do_peeling): Skip prologue peeling if we're using a mask
15638 instead. Update call to vect_update_inits_of_drs.
15639 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15641 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15642 alignment. Do not include the number of peeled iterations in
15643 the minimum threshold in that case.
15644 (vectorizable_induction): Adjust the start value down by
15645 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15646 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15647 Take the number of skipped iterations into account when calculating
15649 * tree-vect-stmts.c (vect_gen_while_not): New function.
15651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15652 Alan Hayward <alan.hayward@arm.com>
15653 David Sherwood <david.sherwood@arm.com>
15655 * doc/sourcebuild.texi (vect_fully_masked): Document.
15656 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15657 default value to 0.
15658 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15660 (vect_analyze_loop_2): ...here. Don't check the vectorization
15661 factor against the number of loop iterations if the loop is
15664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15665 Alan Hayward <alan.hayward@arm.com>
15666 David Sherwood <david.sherwood@arm.com>
15668 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15669 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15670 (dump_groups): Update accordingly.
15671 (iv_use::mem_type): New member variable.
15672 (address_p): New function.
15673 (record_use): Add a mem_type argument and initialize the new
15675 (record_group_use): Add a mem_type argument. Use address_p.
15676 Remove obsolete null checks of base_object. Update call to record_use.
15677 (find_interesting_uses_op): Update call to record_group_use.
15678 (find_interesting_uses_cond): Likewise.
15679 (find_interesting_uses_address): Likewise.
15680 (get_mem_type_for_internal_fn): New function.
15681 (find_address_like_use): Likewise.
15682 (find_interesting_uses_stmt): Try find_address_like_use before
15683 calling find_interesting_uses_op.
15684 (addr_offset_valid_p): Use the iv mem_type field as the type
15685 of the addressed memory.
15686 (add_autoinc_candidates): Likewise.
15687 (get_address_cost): Likewise.
15688 (split_small_address_groups_p): Use address_p.
15689 (split_address_groups): Likewise.
15690 (add_iv_candidate_for_use): Likewise.
15691 (autoinc_possible_for_pair): Likewise.
15692 (rewrite_groups): Likewise.
15693 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15694 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15695 (get_alias_ptr_type_for_ptr_address): New function.
15696 (rewrite_use_address): Rewrite address uses in calls that were
15697 identified by find_address_like_use.
15699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15700 Alan Hayward <alan.hayward@arm.com>
15701 David Sherwood <david.sherwood@arm.com>
15703 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15705 * gimple-expr.h (is_gimple_addressable: Likewise.
15706 * gimple-expr.c (is_gimple_address): Likewise.
15707 * internal-fn.c (expand_call_mem_ref): New function.
15708 (expand_mask_load_optab_fn): Use it.
15709 (expand_mask_store_optab_fn): Likewise.
15711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15712 Alan Hayward <alan.hayward@arm.com>
15713 David Sherwood <david.sherwood@arm.com>
15715 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15716 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15717 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15718 (cond_umax@var{mode}): Document.
15719 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15720 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15721 (cond_umin_optab, cond_umax_optab): New optabs.
15722 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15723 (COND_IOR, COND_XOR): New internal functions.
15724 * internal-fn.h (get_conditional_internal_fn): Declare.
15725 * internal-fn.c (cond_binary_direct): New macro.
15726 (expand_cond_binary_optab_fn): Likewise.
15727 (direct_cond_binary_optab_supported_p): Likewise.
15728 (get_conditional_internal_fn): New function.
15729 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15730 Cope with reduction statements that are vectorized as calls rather
15732 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15733 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15734 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15735 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15736 (UNSPEC_COND_EOR): New unspecs.
15737 (optab): Add mappings for them.
15738 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15739 (sve_int_op, sve_fp_op): New int attributes.
15741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15742 Alan Hayward <alan.hayward@arm.com>
15743 David Sherwood <david.sherwood@arm.com>
15745 * optabs.def (while_ult_optab): New optab.
15746 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15747 * internal-fn.def (WHILE_ULT): New internal function.
15748 * internal-fn.h (direct_internal_fn_supported_p): New override
15749 that takes two types as argument.
15750 * internal-fn.c (while_direct): New macro.
15751 (expand_while_optab_fn): New function.
15752 (convert_optab_supported_p): Likewise.
15753 (direct_while_optab_supported_p): New macro.
15754 * wide-int.h (wi::udiv_ceil): New function.
15755 * tree-vectorizer.h (rgroup_masks): New structure.
15756 (vec_loop_masks): New typedef.
15757 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15758 and fully_masked_p.
15759 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15760 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15761 (vect_max_vf): New function.
15762 (slpeel_make_loop_iterate_ntimes): Delete.
15763 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15764 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15765 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15766 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15767 internal-fn.h, stor-layout.h and optabs-query.h.
15768 (vect_set_loop_mask): New function.
15769 (add_preheader_seq): Likewise.
15770 (add_header_seq): Likewise.
15771 (interleave_supported_p): Likewise.
15772 (vect_maybe_permute_loop_masks): Likewise.
15773 (vect_set_loop_masks_directly): Likewise.
15774 (vect_set_loop_condition_masked): Likewise.
15775 (vect_set_loop_condition_unmasked): New function, split out from
15776 slpeel_make_loop_iterate_ntimes.
15777 (slpeel_make_loop_iterate_ntimes): Rename to..
15778 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15779 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15780 (vect_do_peeling): Update call accordingly.
15781 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15783 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15784 mask_compare_type, can_fully_mask_p and fully_masked_p.
15785 (release_vec_loop_masks): New function.
15786 (_loop_vec_info): Use it to free the loop masks.
15787 (can_produce_all_loop_masks_p): New function.
15788 (vect_get_max_nscalars_per_iter): Likewise.
15789 (vect_verify_full_masking): Likewise.
15790 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15791 retries, and free the mask rgroups before retrying. Check loop-wide
15792 reasons for disallowing fully-masked loops. Make the final decision
15793 about whether use a fully-masked loop or not.
15794 (vect_estimate_min_profitable_iters): Do not assume that peeling
15795 for the number of iterations will be needed for fully-masked loops.
15796 (vectorizable_reduction): Disable fully-masked loops.
15797 (vectorizable_live_operation): Likewise.
15798 (vect_halve_mask_nunits): New function.
15799 (vect_double_mask_nunits): Likewise.
15800 (vect_record_loop_mask): Likewise.
15801 (vect_get_loop_mask): Likewise.
15802 (vect_transform_loop): Handle the case in which the final loop
15803 iteration might handle a partial vector. Call vect_set_loop_condition
15804 instead of slpeel_make_loop_iterate_ntimes.
15805 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15806 (check_load_store_masking): New function.
15807 (prepare_load_store_mask): Likewise.
15808 (vectorizable_store): Handle fully-masked loops.
15809 (vectorizable_load): Likewise.
15810 (supportable_widening_operation): Use vect_halve_mask_nunits for
15812 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15813 (vect_gen_while): New function.
15814 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15815 (aarch64_uqdec<mode>): New insn.
15817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15818 Alan Hayward <alan.hayward@arm.com>
15819 David Sherwood <david.sherwood@arm.com>
15821 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15822 (reduc_xor_scal_optab): New optabs.
15823 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15824 (reduc_xor_scal_@var{m}): Document.
15825 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15826 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15827 internal functions.
15828 * fold-const-call.c (fold_const_call): Handle them.
15829 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15830 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15831 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15832 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15833 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15834 (UNSPEC_XORV): New unspecs.
15835 (optab): Add entries for them.
15836 (BITWISEV): New int iterator.
15837 (bit_reduc_op): New int attributes.
15839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15840 Alan Hayward <alan.hayward@arm.com>
15841 David Sherwood <david.sherwood@arm.com>
15843 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15844 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15845 * optabs.def (vec_shl_insert_optab): New optab.
15846 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15847 (duplicate_and_interleave): Likewise.
15848 * tree-vect-loop.c: Include internal-fn.h.
15849 (neutral_op_for_slp_reduction): New function, split out from
15850 get_initial_defs_for_reduction.
15851 (get_initial_def_for_reduction): Handle option 2 for variable-length
15852 vectors by loading the neutral value into a vector and then shifting
15853 the initial value into element 0.
15854 (get_initial_defs_for_reduction): Replace the code argument with
15855 the neutral value calculated by neutral_op_for_slp_reduction.
15856 Use gimple_build_vector for constant-length vectors.
15857 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15858 but the first group_size elements have a neutral value.
15859 Use duplicate_and_interleave otherwise.
15860 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15861 Update call to get_initial_defs_for_reduction. Handle SLP
15862 reductions for variable-length vectors by creating one vector
15863 result for each scalar result, with the elements associated
15864 with other scalar results stubbed out with the neutral value.
15865 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15866 Require IFN_VEC_SHL_INSERT for double reductions on
15867 variable-length vectors, or SLP reductions that have
15868 a neutral value. Require can_duplicate_and_interleave_p
15869 support for variable-length unchained SLP reductions if there
15870 is no neutral value, such as for MIN/MAX reductions. Also require
15871 the number of vector elements to be a multiple of the number of
15872 SLP statements when doing variable-length unchained SLP reductions.
15873 Update call to vect_create_epilog_for_reduction.
15874 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15875 and remove initial values.
15876 (duplicate_and_interleave): Make public.
15877 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15878 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15881 Alan Hayward <alan.hayward@arm.com>
15882 David Sherwood <david.sherwood@arm.com>
15884 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15885 (can_duplicate_and_interleave_p): New function.
15886 (vect_get_and_check_slp_defs): Take the vector of statements
15887 rather than just the current one. Remove excess parentheses.
15888 Restriction rejectinon of vect_constant_def and vect_external_def
15889 for variable-length vectors to boolean types, or types for which
15890 can_duplicate_and_interleave_p is false.
15891 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15892 (duplicate_and_interleave): New function.
15893 (vect_get_constant_vectors): Use gimple_build_vector for
15894 constant-length vectors and suitable variable-length constant
15895 vectors. Use duplicate_and_interleave for other variable-length
15896 vectors. Don't defer the update when inserting new statements.
15898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15899 Alan Hayward <alan.hayward@arm.com>
15900 David Sherwood <david.sherwood@arm.com>
15902 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15903 min_profitable_iters doesn't go negative.
15905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15906 Alan Hayward <alan.hayward@arm.com>
15907 David Sherwood <david.sherwood@arm.com>
15909 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15910 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15911 * optabs.def (vec_mask_load_lanes_optab): New optab.
15912 (vec_mask_store_lanes_optab): Likewise.
15913 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15914 (MASK_STORE_LANES): Likewise.
15915 * internal-fn.c (mask_load_lanes_direct): New macro.
15916 (mask_store_lanes_direct): Likewise.
15917 (expand_mask_load_optab_fn): Handle masked operations.
15918 (expand_mask_load_lanes_optab_fn): New macro.
15919 (expand_mask_store_optab_fn): Handle masked operations.
15920 (expand_mask_store_lanes_optab_fn): New macro.
15921 (direct_mask_load_lanes_optab_supported_p): Likewise.
15922 (direct_mask_store_lanes_optab_supported_p): Likewise.
15923 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15925 (vect_load_lanes_supported): Likewise.
15926 * tree-vect-data-refs.c (strip_conversion): New function.
15927 (can_group_stmts_p): Likewise.
15928 (vect_analyze_data_ref_accesses): Use it instead of checking
15929 for a pair of assignments.
15930 (vect_store_lanes_supported): Take a masked_p parameter.
15931 (vect_load_lanes_supported): Likewise.
15932 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15933 vect_store_lanes_supported and vect_load_lanes_supported.
15934 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15935 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15936 parameter. Don't allow gaps for masked accesses.
15937 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15938 and vect_load_lanes_supported.
15939 (get_load_store_type): Take a masked_p parameter and update
15940 call to get_group_load_store_type.
15941 (vectorizable_store): Update call to get_load_store_type.
15942 Handle IFN_MASK_STORE_LANES.
15943 (vectorizable_load): Update call to get_load_store_type.
15944 Handle IFN_MASK_LOAD_LANES.
15946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15947 Alan Hayward <alan.hayward@arm.com>
15948 David Sherwood <david.sherwood@arm.com>
15950 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15952 * config/aarch64/aarch64-protos.h
15953 (aarch64_sve_struct_memory_operand_p): Declare.
15954 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15955 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15956 (VPRED, vpred): Handle SVE structure modes.
15957 * config/aarch64/constraints.md (Utx): New constraint.
15958 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15959 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15960 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15961 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15962 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15963 structure modes. Split into pieces after RA.
15964 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15965 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15967 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15968 SVE structure modes.
15969 (aarch64_classify_address): Likewise.
15970 (sizetochar): Move earlier in file.
15971 (aarch64_print_operand): Handle SVE register lists.
15972 (aarch64_array_mode): New function.
15973 (aarch64_sve_struct_memory_operand_p): Likewise.
15974 (TARGET_ARRAY_MODE): Redefine.
15976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15977 Alan Hayward <alan.hayward@arm.com>
15978 David Sherwood <david.sherwood@arm.com>
15980 * target.def (array_mode): New target hook.
15981 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15982 * doc/tm.texi: Regenerate.
15983 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15984 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15985 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15986 targetm.array_mode.
15987 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15991 Alan Hayward <alan.hayward@arm.com>
15992 David Sherwood <david.sherwood@arm.com>
15994 * fold-const.c (fold_binary_loc): Check the argument types
15995 rather than the result type when testing for a vector operation.
15997 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15999 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16000 * doc/tm.texi: Regenerate.
16002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16003 Alan Hayward <alan.hayward@arm.com>
16004 David Sherwood <david.sherwood@arm.com>
16006 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16007 (sve): Document new AArch64 extension.
16008 * doc/md.texi (w): Extend the description of the AArch64
16009 constraint to include SVE vectors.
16010 (Upl, Upa): Document new AArch64 predicate constraints.
16011 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16013 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16014 (msve-vector-bits=): New option.
16015 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16016 SVE when these are disabled.
16017 (sve): New extension.
16018 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16019 modes. Adjust their number of units based on aarch64_sve_vg.
16020 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16021 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16022 aarch64_addr_query_type.
16023 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16024 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16025 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16026 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16027 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16028 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16029 (aarch64_simd_imm_zero_p): Delete.
16030 (aarch64_check_zero_based_sve_index_immediate): Declare.
16031 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16032 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16033 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16034 (aarch64_sve_float_mul_immediate_p): Likewise.
16035 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16036 rather than an rtx.
16037 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16038 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16039 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16040 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16041 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16042 (aarch64_regmode_natural_size): Likewise.
16043 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16044 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16046 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16047 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16048 for VG and the SVE predicate registers.
16049 (V_ALIASES): Add a "z"-prefixed alias.
16050 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16051 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16052 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16053 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16054 (REG_CLASS_NAMES): Add entries for them.
16055 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16056 and the predicate registers.
16057 (aarch64_sve_vg): Declare.
16058 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16059 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16060 (REGMODE_NATURAL_SIZE): Define.
16061 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16063 * config/aarch64/aarch64.c: Include cfgrtl.h.
16064 (simd_immediate_info): Add a constructor for series vectors,
16065 and an associated step field.
16066 (aarch64_sve_vg): New variable.
16067 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16068 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16069 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16070 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16071 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16072 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16073 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16074 (aarch64_get_mask_mode): New functions.
16075 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16076 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16077 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16078 predicate modes and predicate registers. Explicitly restrict
16079 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16080 to store a vector mode if it is recognized by
16081 aarch64_classify_vector_mode.
16082 (aarch64_regmode_natural_size): New function.
16083 (aarch64_hard_regno_caller_save_mode): Return the original mode
16085 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16086 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16087 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16088 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16090 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16091 does not overlap dest if the function is frame-related. Handle
16093 (aarch64_split_add_offset): New function.
16094 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16095 them aarch64_add_offset.
16096 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16097 and update call to aarch64_sub_sp.
16098 (aarch64_add_cfa_expression): New function.
16099 (aarch64_expand_prologue): Pass extra temporary registers to the
16100 functions above. Handle the case in which we need to emit new
16101 DW_CFA_expressions for registers that were originally saved
16102 relative to the stack pointer, but now have to be expressed
16103 relative to the frame pointer.
16104 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16106 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16107 IP0 and IP1 values for SVE frames.
16108 (aarch64_expand_vec_series): New function.
16109 (aarch64_expand_sve_widened_duplicate): Likewise.
16110 (aarch64_expand_sve_const_vector): Likewise.
16111 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16112 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16113 into the register, rather than emitting a SET directly.
16114 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16115 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16116 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16117 (offset_9bit_signed_scaled_p): New functions.
16118 (aarch64_replicate_bitmask_imm): New function.
16119 (aarch64_bitmask_imm): Use it.
16120 (aarch64_cannot_force_const_mem): Reject expressions involving
16121 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16122 (aarch64_classify_index): Handle SVE indices, by requiring
16123 a plain register index with a scale that matches the element size.
16124 (aarch64_classify_address): Handle SVE addresses. Assert that
16125 the mode of the address is VOIDmode or an integer mode.
16126 Update call to aarch64_classify_symbol.
16127 (aarch64_classify_symbolic_expression): Update call to
16128 aarch64_classify_symbol.
16129 (aarch64_const_vec_all_in_range_p): New function.
16130 (aarch64_print_vector_float_operand): Likewise.
16131 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16132 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16133 and the FP immediates 1.0 and 0.5.
16134 (aarch64_print_address_internal): Handle SVE addresses.
16135 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16136 (aarch64_regno_regclass): Handle predicate registers.
16137 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16139 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16140 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16141 (aarch64_convert_sve_vector_bits): New function.
16142 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16143 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16144 rather than an rtx.
16145 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16146 Handle SVE vector and predicate modes. Accept VL-based constants
16147 that need only one temporary register, and VL offsets that require
16148 no temporary registers.
16149 (aarch64_conditional_register_usage): Mark the predicate registers
16150 as fixed if SVE isn't available.
16151 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16152 Return true for SVE vector and predicate modes.
16153 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16154 rather than an unsigned int. Handle SVE modes.
16155 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16157 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16159 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16160 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16161 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16162 (aarch64_sve_float_mul_immediate_p): New functions.
16163 (aarch64_sve_valid_immediate): New function.
16164 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16165 Explicitly reject structure modes. Check for INDEX constants.
16166 Handle PTRUE and PFALSE constants.
16167 (aarch64_check_zero_based_sve_index_immediate): New function.
16168 (aarch64_simd_imm_zero_p): Delete.
16169 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16170 vector modes. Accept constants in the range of CNT[BHWD].
16171 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16172 ask for an Advanced SIMD mode.
16173 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16174 (aarch64_simd_vector_alignment): Handle SVE predicates.
16175 (aarch64_vectorize_preferred_vector_alignment): New function.
16176 (aarch64_simd_vector_alignment_reachable): Use it instead of
16178 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16179 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16181 (MAX_VECT_LEN): Delete.
16182 (expand_vec_perm_d): Add a vec_flags field.
16183 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16184 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16185 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16187 (aarch64_evpc_rev): Rename to...
16188 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16189 (aarch64_evpc_rev_global): New function.
16190 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16191 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16193 (aarch64_evpc_sve_tbl): New function.
16194 (aarch64_expand_vec_perm_const_1): Update after rename of
16195 aarch64_evpc_rev. Handle SVE permutes too, trying
16196 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16197 than aarch64_evpc_tbl.
16198 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16199 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16200 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16201 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16202 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16203 (aarch64_expand_sve_vcond): New functions.
16204 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16205 of aarch64_vector_mode_p.
16206 (aarch64_dwarf_poly_indeterminate_value): New function.
16207 (aarch64_compute_pressure_classes): Likewise.
16208 (aarch64_can_change_mode_class): Likewise.
16209 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16210 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16211 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16212 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16213 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16214 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16215 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16216 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16218 (Dn, Dl, Dr): Accept const as well as const_vector.
16219 (Dz): Likewise. Compare against CONST0_RTX.
16220 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16221 of "vector" where appropriate.
16222 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16223 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16224 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16225 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16226 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16227 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16228 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16229 (v_int_equiv): Extend to SVE modes.
16230 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16232 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16233 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16234 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16235 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16236 (SVE_COND_FP_CMP): New int iterators.
16237 (perm_hilo): Handle the new unpack unspecs.
16238 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16240 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16241 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16242 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16243 (aarch64_equality_operator, aarch64_constant_vector_operand)
16244 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16245 (aarch64_sve_nonimmediate_operand): Likewise.
16246 (aarch64_sve_general_operand): Likewise.
16247 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16248 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16249 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16250 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16251 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16252 (aarch64_sve_float_arith_immediate): Likewise.
16253 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16254 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16255 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16256 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16257 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16258 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16259 (aarch64_sve_float_arith_operand): Likewise.
16260 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16261 (aarch64_sve_float_mul_operand): Likewise.
16262 (aarch64_sve_vec_perm_operand): Likewise.
16263 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16264 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16265 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16266 as well as const_vector.
16267 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16268 in file. Use CONST0_RTX and CONSTM1_RTX.
16269 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16270 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16271 Use aarch64_simd_imm_zero.
16272 * config/aarch64/aarch64-sve.md: New file.
16273 * config/aarch64/aarch64.md: Include it.
16274 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16275 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16276 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16277 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16278 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16279 (sve): New attribute.
16280 (enabled): Disable instructions with the sve attribute unless
16282 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16283 aarch64_expand_mov_immediate.
16284 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16285 CNT[BHSD] immediates.
16286 (movti): Split CONST_POLY_INT moves into two halves.
16287 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16288 Split additions that need a temporary here if the destination
16289 is the stack pointer.
16290 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16291 (*add<mode>3_poly_1): New instruction.
16292 (set_clobber_cc): New expander.
16294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16296 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16297 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16298 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16299 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16300 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16301 Change innermode from fixed_mode_size to machine_mode.
16302 (simplify_subreg): Update call accordingly. Handle a constant-sized
16303 subreg of a variable-length CONST_VECTOR.
16305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16306 Alan Hayward <alan.hayward@arm.com>
16307 David Sherwood <david.sherwood@arm.com>
16309 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16310 (add_offset_to_base): New function, split out from...
16311 (create_mem_ref): ...here. When handling a scale other than 1,
16312 check first whether the address is valid without the offset.
16313 Add it into the base if so, leaving the index and scale as-is.
16315 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16318 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16319 fold_for_warn before checking if arg2 is INTEGER_CST.
16321 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16323 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16324 (store_multiple_operation): Delete.
16325 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16326 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16327 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16328 guarded by TARGET_STRING.
16329 (rs6000_output_load_multiple): Delete.
16330 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16331 OPTION_MASK_STRING / TARGET_STRING handling.
16332 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16333 (const rs6000_opt_masks) <"string">: Change mask to 0.
16334 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16335 (MASK_STRING): Delete.
16336 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16338 (load_multiple): Delete.
16345 (store_multiple): Delete.
16352 (movmemsi_8reg): Delete.
16353 (corresponding unnamed define_insn): Delete.
16354 (movmemsi_6reg): Delete.
16355 (corresponding unnamed define_insn): Delete.
16356 (movmemsi_4reg): Delete.
16357 (corresponding unnamed define_insn): Delete.
16358 (movmemsi_2reg): Delete.
16359 (corresponding unnamed define_insn): Delete.
16360 (movmemsi_1reg): Delete.
16361 (corresponding unnamed define_insn): Delete.
16362 * config/rs6000/rs6000.opt (mno-string): New.
16363 (mstring): Replace by deprecation warning stub.
16364 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16366 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16368 * regrename.c (regrename_do_replace): If replacing the same
16369 reg multiple times, try to reuse last created gen_raw_REG.
16372 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16373 main to workaround a bug in GDB.
16375 2018-01-12 Tom de Vries <tom@codesourcery.com>
16378 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16380 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16382 PR rtl-optimization/80481
16383 * ira-color.c (get_cap_member): New function.
16384 (allocnos_conflict_by_live_ranges_p): Use it.
16385 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16386 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16388 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16391 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16392 (*saddl_se_1): Ditto.
16393 (*ssubsi_1): Ditto.
16394 (*ssubl_se_1): Ditto.
16396 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16398 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16399 rather than wi::to_widest for DR_INITs.
16400 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16401 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16402 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16404 (vect_analyze_group_access_1): Note that here.
16406 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16408 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16409 polynomial type sizes.
16411 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16413 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16414 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16415 (gimple_add_tmp_var): Likewise.
16417 2018-01-12 Martin Liska <mliska@suse.cz>
16419 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16420 (gimple_alloc_sizes): Likewise.
16421 (dump_gimple_statistics): Use PRIu64 in printf format.
16422 * gimple.h: Change uint64_t to int.
16424 2018-01-12 Martin Liska <mliska@suse.cz>
16426 * tree-core.h: Use uint64_t instead of int.
16427 * tree.c (tree_node_counts): Likewise.
16428 (tree_node_sizes): Likewise.
16429 (dump_tree_statistics): Use PRIu64 in printf format.
16431 2018-01-12 Martin Liska <mliska@suse.cz>
16433 * Makefile.in: As qsort_chk is implemented in vec.c, add
16434 vec.o to linkage of gencfn-macros.
16435 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16436 passing the info to record_node_allocation_statistics.
16437 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16439 * ggc-common.c (struct ggc_usage): Add operator== and use
16440 it in operator< and compare function.
16441 * mem-stats.h (struct mem_usage): Likewise.
16442 * vec.c (struct vec_usage): Remove operator< and compare
16443 function. Can be simply inherited.
16445 2018-01-12 Martin Jambor <mjambor@suse.cz>
16448 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16449 * tree-ssa-math-opts.c: Include domwalk.h.
16450 (convert_mult_to_fma_1): New function.
16451 (fma_transformation_info): New type.
16452 (fma_deferring_state): Likewise.
16453 (cancel_fma_deferring): New function.
16454 (result_of_phi): Likewise.
16455 (last_fma_candidate_feeds_initial_phi): Likewise.
16456 (convert_mult_to_fma): Added deferring logic, split actual
16457 transformation to convert_mult_to_fma_1.
16458 (math_opts_dom_walker): New type.
16459 (math_opts_dom_walker::after_dom_children): New method, body moved
16460 here from pass_optimize_widening_mul::execute, added deferring logic
16462 (pass_optimize_widening_mul::execute): Moved most of code to
16463 math_opts_dom_walker::after_dom_children.
16464 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16465 * config/i386/i386.c (ix86_option_override_internal): Added
16466 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16468 2018-01-12 Richard Biener <rguenther@suse.de>
16471 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16472 inline instance vars.
16474 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16477 * config/rx/rx.c (rx_is_restricted_memory_address):
16478 Handle SUBREG case.
16480 2018-01-12 Richard Biener <rguenther@suse.de>
16482 PR tree-optimization/80846
16483 * target.def (split_reduction): New target hook.
16484 * targhooks.c (default_split_reduction): New function.
16485 * targhooks.h (default_split_reduction): Declare.
16486 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16487 target requests first reduce vectors by combining low and high
16489 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16490 (get_vectype_for_scalar_type_and_size): Export.
16491 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16492 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16493 * doc/tm.texi: Regenerate.
16494 * config/i386/i386.c (ix86_split_reduction): Implement
16495 TARGET_VECTORIZE_SPLIT_REDUCTION.
16497 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16500 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16501 in PIC mode except for TARGET_VXWORKS_RTP.
16502 * config/sparc/sparc.c: Include cfgrtl.h.
16503 (TARGET_INIT_PIC_REG): Define.
16504 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16505 (sparc_pic_register_p): New predicate.
16506 (sparc_legitimate_address_p): Use it.
16507 (sparc_legitimize_pic_address): Likewise.
16508 (sparc_delegitimize_address): Likewise.
16509 (sparc_mode_dependent_address_p): Likewise.
16510 (gen_load_pcrel_sym): Remove 4th parameter.
16511 (load_got_register): Adjust call to above. Remove obsolete stuff.
16512 (sparc_expand_prologue): Do not call load_got_register here.
16513 (sparc_flat_expand_prologue): Likewise.
16514 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16515 (sparc_use_pseudo_pic_reg): New function.
16516 (sparc_init_pic_reg): Likewise.
16517 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16518 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16520 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16522 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16523 Add item for branch_cost.
16525 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16527 PR rtl-optimization/83565
16528 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16529 not extend the result to a larger mode for rotate operations.
16530 (num_sign_bit_copies1): Likewise.
16532 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16535 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16537 Use values-Xc.o for -pedantic.
16538 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16540 2018-01-12 Martin Liska <mliska@suse.cz>
16543 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16545 (possible_polymorphic_call_targets): Use it.
16546 (ipa_devirt): Likewise.
16548 2018-01-12 Martin Liska <mliska@suse.cz>
16550 * profile-count.h (enum profile_quality): Use 0 as invalid
16551 enum value of profile_quality.
16553 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16555 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16556 -mext-string options.
16558 2018-01-12 Richard Biener <rguenther@suse.de>
16560 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16561 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16562 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16564 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16566 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16568 * configure.ac (--with-long-double-format): Add support for the
16569 configuration option to change the default long double format on
16571 * config.gcc (powerpc*-linux*-*): Likewise.
16572 * configure: Regenerate.
16573 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16574 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16575 used without modification.
16577 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16579 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16580 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16581 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16582 MISC_BUILTIN_SPEC_BARRIER.
16583 (rs6000_init_builtins): Likewise.
16584 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16586 (speculation_barrier): New define_insn.
16587 * doc/extend.texi: Document __builtin_speculation_barrier.
16589 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16592 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16593 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16594 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16596 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16597 integral modes instead of "ss" and "sd".
16598 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16599 vectors with 32-bit and 64-bit elements.
16600 (vecdupssescalarmodesuffix): New mode attribute.
16601 (vec_dup<mode>): Use it.
16603 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16606 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16607 frame if argument is passed on stack.
16609 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16612 * ree.c (combine_reaching_defs): Optimize also
16613 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16614 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16616 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16618 PR middle-end/83189
16619 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16621 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16623 PR middle-end/83718
16624 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16625 after they are computed.
16627 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16629 PR tree-optimization/83695
16630 * gimple-loop-linterchange.cc
16631 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16632 reset cached scev information after interchange.
16633 (pass_linterchange::execute): Remove call to scev_reset_htab.
16635 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16637 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16638 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16639 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16640 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16641 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16642 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16643 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16644 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16645 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16646 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16647 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16648 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16649 (V_lane_reg): Likewise.
16650 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16652 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16653 (vfmal_lane_low<mode>_intrinsic,
16654 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16655 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16656 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16657 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16658 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16659 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16661 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16663 * config/arm/arm-cpus.in (fp16fml): New feature.
16664 (ALL_SIMD): Add fp16fml.
16665 (armv8.2-a): Add fp16fml as an option.
16666 (armv8.3-a): Likewise.
16667 (armv8.4-a): Add fp16fml as part of fp16.
16668 * config/arm/arm.h (TARGET_FP16FML): Define.
16669 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16671 * config/arm/arm-modes.def (V2HF): Define.
16672 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16673 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16674 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16675 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16676 vfmsl_low, vfmsl_high): New set of builtins.
16677 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16678 (vfml_op): New code attribute.
16679 (VFMLHALVES): New int iterator.
16680 (VFML, VFMLSEL): New mode attributes.
16681 (V_reg): Define mapping for V2HF.
16682 (V_hi, V_lo): New mode attributes.
16683 (VF_constraint): Likewise.
16684 (vfml_half, vfml_half_selector): New int attributes.
16685 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16687 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16688 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16690 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16691 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16692 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16693 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16695 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16696 Document new effective target and option set.
16698 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16700 * config/arm/arm-cpus.in (armv8_4): New feature.
16701 (ARMv8_4a): New fgroup.
16702 (armv8.4-a): New arch.
16703 * config/arm/arm-tables.opt: Regenerate.
16704 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16705 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16706 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16707 Add matching rules for -march=armv8.4-a and extensions.
16708 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16710 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16713 * config/rx/rx.md (BW): New mode attribute.
16714 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16716 2018-01-11 Richard Biener <rguenther@suse.de>
16718 PR tree-optimization/83435
16719 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16720 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16721 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16723 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16724 Alan Hayward <alan.hayward@arm.com>
16725 David Sherwood <david.sherwood@arm.com>
16727 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16729 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16730 (aarch64_print_address_internal): Use it to check for a zero offset.
16732 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16733 Alan Hayward <alan.hayward@arm.com>
16734 David Sherwood <david.sherwood@arm.com>
16736 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16737 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16738 Return a poly_int64 rather than a HOST_WIDE_INT.
16739 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16740 rather than a HOST_WIDE_INT.
16741 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16742 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16743 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16744 final_offset from HOST_WIDE_INT to poly_int64.
16745 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16746 to_constant when getting the number of units in an Advanced SIMD
16748 (aarch64_builtin_vectorized_function): Check for a constant number
16750 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16752 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16753 attribute instead of GET_MODE_NUNITS.
16754 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16755 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16756 GET_MODE_SIZE for fixed-size registers.
16757 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16758 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16759 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16760 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16761 (aarch64_print_operand, aarch64_print_address_internal)
16762 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16763 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16764 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16765 Handle polynomial GET_MODE_SIZE.
16766 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16767 wider than SImode without modification.
16768 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16769 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16770 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16771 passing and returning SVE modes.
16772 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16773 rather than GEN_INT.
16774 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16775 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16776 (aarch64_allocate_and_probe_stack_space): Likewise.
16777 (aarch64_layout_frame): Cope with polynomial offsets.
16778 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16779 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16780 polynomial offsets.
16781 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16782 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16783 poly_int64 rather than a HOST_WIDE_INT.
16784 (aarch64_get_separate_components, aarch64_process_components)
16785 (aarch64_expand_prologue, aarch64_expand_epilogue)
16786 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16787 (aarch64_anchor_offset): New function, split out from...
16788 (aarch64_legitimize_address): ...here.
16789 (aarch64_builtin_vectorization_cost): Handle polynomial
16790 TYPE_VECTOR_SUBPARTS.
16791 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16793 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16794 number of elements from the PARALLEL rather than the mode.
16795 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16796 rather than GET_MODE_BITSIZE.
16797 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16798 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16799 (aarch64_expand_vec_perm_const_1): Handle polynomial
16800 d->perm.length () and d->perm elements.
16801 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16802 Apply to_constant to d->perm elements.
16803 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16804 polynomial CONST_VECTOR_NUNITS.
16805 (aarch64_move_pointer): Take amount as a poly_int64 rather
16807 (aarch64_progress_pointer): Avoid temporary variable.
16808 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16809 the mode attribute instead of GET_MODE.
16811 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16812 Alan Hayward <alan.hayward@arm.com>
16813 David Sherwood <david.sherwood@arm.com>
16815 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16816 x exists before using it.
16817 (aarch64_add_constant_internal): Rename to...
16818 (aarch64_add_offset_1): ...this. Replace regnum with separate
16819 src and dest rtxes. Handle the case in which they're different,
16820 including when the offset is zero. Replace scratchreg with an rtx.
16821 Use 2 additions if there is no spare register into which we can
16822 move a 16-bit constant.
16823 (aarch64_add_constant): Delete.
16824 (aarch64_add_offset): Replace reg with separate src and dest
16825 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16826 Use aarch64_add_offset_1.
16827 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16828 an rtx rather than an int. Take the delta as a poly_int64
16829 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16830 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16831 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16832 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16833 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16834 and aarch64_add_sp.
16835 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16836 aarch64_add_constant.
16838 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16840 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16841 Use scalar_float_mode.
16843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16845 * config/aarch64/aarch64-simd.md
16846 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16847 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16848 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16849 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16850 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16851 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16852 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16853 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16854 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16855 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16857 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16860 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16861 targ_options->x_arm_arch_string is non NULL.
16863 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16865 * config/aarch64/aarch64.h
16866 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16868 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16871 * expmed.c (emit_store_flag_force): Swap if const op0
16872 and change VOIDmode to mode of op0.
16874 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16876 PR rtl-optimization/83761
16877 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16878 than bytes to mode_for_size.
16880 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16882 PR middle-end/83189
16883 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16884 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16887 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16889 PR middle-end/83575
16890 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16891 when in layout mode.
16892 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16893 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16896 2018-01-10 Michael Collison <michael.collison@arm.com>
16898 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16899 * config/aarch64/aarch64-option-extension.def: Add
16900 AARCH64_OPT_EXTENSION of 'fp16fml'.
16901 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16902 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16903 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16904 * config/aarch64/constraints.md (Ui7): New constraint.
16905 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16906 (VFMLA_SEL_W): Ditto.
16909 (VFMLA16_LOW): New int iterator.
16910 (VFMLA16_HIGH): Ditto.
16911 (UNSPEC_FMLAL): New unspec.
16912 (UNSPEC_FMLSL): Ditto.
16913 (UNSPEC_FMLAL2): Ditto.
16914 (UNSPEC_FMLSL2): Ditto.
16915 (f16mac): New code attribute.
16916 * config/aarch64/aarch64-simd-builtins.def
16917 (aarch64_fmlal_lowv2sf): Ditto.
16918 (aarch64_fmlsl_lowv2sf): Ditto.
16919 (aarch64_fmlalq_lowv4sf): Ditto.
16920 (aarch64_fmlslq_lowv4sf): Ditto.
16921 (aarch64_fmlal_highv2sf): Ditto.
16922 (aarch64_fmlsl_highv2sf): Ditto.
16923 (aarch64_fmlalq_highv4sf): Ditto.
16924 (aarch64_fmlslq_highv4sf): Ditto.
16925 (aarch64_fmlal_lane_lowv2sf): Ditto.
16926 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16927 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16928 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16929 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16930 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16931 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16932 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16933 (aarch64_fmlal_lane_highv2sf): Ditto.
16934 (aarch64_fmlsl_lane_highv2sf): Ditto.
16935 (aarch64_fmlal_laneq_highv2sf): Ditto.
16936 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16937 (aarch64_fmlalq_lane_highv4sf): Ditto.
16938 (aarch64_fmlsl_lane_highv4sf): Ditto.
16939 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16940 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16941 * config/aarch64/aarch64-simd.md:
16942 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16943 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16944 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16945 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16946 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16947 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16948 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16949 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16950 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16951 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16952 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16953 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16954 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16955 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16956 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16957 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16958 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16959 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16960 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16961 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16962 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16963 (vfmlsl_low_u32): Ditto.
16964 (vfmlalq_low_u32): Ditto.
16965 (vfmlslq_low_u32): Ditto.
16966 (vfmlal_high_u32): Ditto.
16967 (vfmlsl_high_u32): Ditto.
16968 (vfmlalq_high_u32): Ditto.
16969 (vfmlslq_high_u32): Ditto.
16970 (vfmlal_lane_low_u32): Ditto.
16971 (vfmlsl_lane_low_u32): Ditto.
16972 (vfmlal_laneq_low_u32): Ditto.
16973 (vfmlsl_laneq_low_u32): Ditto.
16974 (vfmlalq_lane_low_u32): Ditto.
16975 (vfmlslq_lane_low_u32): Ditto.
16976 (vfmlalq_laneq_low_u32): Ditto.
16977 (vfmlslq_laneq_low_u32): Ditto.
16978 (vfmlal_lane_high_u32): Ditto.
16979 (vfmlsl_lane_high_u32): Ditto.
16980 (vfmlal_laneq_high_u32): Ditto.
16981 (vfmlsl_laneq_high_u32): Ditto.
16982 (vfmlalq_lane_high_u32): Ditto.
16983 (vfmlslq_lane_high_u32): Ditto.
16984 (vfmlalq_laneq_high_u32): Ditto.
16985 (vfmlslq_laneq_high_u32): Ditto.
16986 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16987 (AARCH64_FL_FOR_ARCH8_4): New.
16988 (AARCH64_ISA_F16FML): New ISA flag.
16989 (TARGET_F16FML): New feature flag for fp16fml.
16990 (doc/invoke.texi): Document new fp16fml option.
16992 2018-01-10 Michael Collison <michael.collison@arm.com>
16994 * config/aarch64/aarch64-builtins.c:
16995 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16996 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16997 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16998 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16999 (AARCH64_ISA_SHA3): New ISA flag.
17000 (TARGET_SHA3): New feature flag for sha3.
17001 * config/aarch64/iterators.md (sha512_op): New int attribute.
17002 (CRYPTO_SHA512): New int iterator.
17003 (UNSPEC_SHA512H): New unspec.
17004 (UNSPEC_SHA512H2): Ditto.
17005 (UNSPEC_SHA512SU0): Ditto.
17006 (UNSPEC_SHA512SU1): Ditto.
17007 * config/aarch64/aarch64-simd-builtins.def
17008 (aarch64_crypto_sha512hqv2di): New builtin.
17009 (aarch64_crypto_sha512h2qv2di): Ditto.
17010 (aarch64_crypto_sha512su0qv2di): Ditto.
17011 (aarch64_crypto_sha512su1qv2di): Ditto.
17012 (aarch64_eor3qv8hi): Ditto.
17013 (aarch64_rax1qv2di): Ditto.
17014 (aarch64_xarqv2di): Ditto.
17015 (aarch64_bcaxqv8hi): Ditto.
17016 * config/aarch64/aarch64-simd.md:
17017 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17018 (aarch64_crypto_sha512su0qv2di): Ditto.
17019 (aarch64_crypto_sha512su1qv2di): Ditto.
17020 (aarch64_eor3qv8hi): Ditto.
17021 (aarch64_rax1qv2di): Ditto.
17022 (aarch64_xarqv2di): Ditto.
17023 (aarch64_bcaxqv8hi): Ditto.
17024 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17025 (vsha512h2q_u64): Ditto.
17026 (vsha512su0q_u64): Ditto.
17027 (vsha512su1q_u64): Ditto.
17028 (veor3q_u16): Ditto.
17029 (vrax1q_u64): Ditto.
17030 (vxarq_u64): Ditto.
17031 (vbcaxq_u16): Ditto.
17032 * config/arm/types.md (crypto_sha512): New type attribute.
17033 (crypto_sha3): Ditto.
17034 (doc/invoke.texi): Document new sha3 option.
17036 2018-01-10 Michael Collison <michael.collison@arm.com>
17038 * config/aarch64/aarch64-builtins.c:
17039 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17040 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17041 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17042 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17043 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17044 (AARCH64_ISA_SM4): New ISA flag.
17045 (TARGET_SM4): New feature flag for sm4.
17046 * config/aarch64/aarch64-simd-builtins.def
17047 (aarch64_sm3ss1qv4si): Ditto.
17048 (aarch64_sm3tt1aq4si): Ditto.
17049 (aarch64_sm3tt1bq4si): Ditto.
17050 (aarch64_sm3tt2aq4si): Ditto.
17051 (aarch64_sm3tt2bq4si): Ditto.
17052 (aarch64_sm3partw1qv4si): Ditto.
17053 (aarch64_sm3partw2qv4si): Ditto.
17054 (aarch64_sm4eqv4si): Ditto.
17055 (aarch64_sm4ekeyqv4si): Ditto.
17056 * config/aarch64/aarch64-simd.md:
17057 (aarch64_sm3ss1qv4si): Ditto.
17058 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17059 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17060 (aarch64_sm4eqv4si): Ditto.
17061 (aarch64_sm4ekeyqv4si): Ditto.
17062 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17063 (sm3part_op): Ditto.
17064 (CRYPTO_SM3TT): Ditto.
17065 (CRYPTO_SM3PART): Ditto.
17066 (UNSPEC_SM3SS1): New unspec.
17067 (UNSPEC_SM3TT1A): Ditto.
17068 (UNSPEC_SM3TT1B): Ditto.
17069 (UNSPEC_SM3TT2A): Ditto.
17070 (UNSPEC_SM3TT2B): Ditto.
17071 (UNSPEC_SM3PARTW1): Ditto.
17072 (UNSPEC_SM3PARTW2): Ditto.
17073 (UNSPEC_SM4E): Ditto.
17074 (UNSPEC_SM4EKEY): Ditto.
17075 * config/aarch64/constraints.md (Ui2): New constraint.
17076 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17077 * config/arm/types.md (crypto_sm3): New type attribute.
17078 (crypto_sm4): Ditto.
17079 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17080 (vsm3tt1aq_u32): Ditto.
17081 (vsm3tt1bq_u32): Ditto.
17082 (vsm3tt2aq_u32): Ditto.
17083 (vsm3tt2bq_u32): Ditto.
17084 (vsm3partw1q_u32): Ditto.
17085 (vsm3partw2q_u32): Ditto.
17086 (vsm4eq_u32): Ditto.
17087 (vsm4ekeyq_u32): Ditto.
17088 (doc/invoke.texi): Document new sm4 option.
17090 2018-01-10 Michael Collison <michael.collison@arm.com>
17092 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17093 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17094 (AARCH64_FL_FOR_ARCH8_4): New.
17095 (AARCH64_FL_V8_4): New flag.
17096 (doc/invoke.texi): Document new armv8.4-a option.
17098 2018-01-10 Michael Collison <michael.collison@arm.com>
17100 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17101 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17102 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17103 * config/aarch64/aarch64-option-extension.def: Add
17104 AARCH64_OPT_EXTENSION of 'sha2'.
17105 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17106 (crypto): Disable sha2 and aes if crypto disabled.
17107 (crypto): Enable aes and sha2 if enabled.
17108 (simd): Disable sha2 and aes if simd disabled.
17109 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17111 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17112 (TARGET_SHA2): New feature flag for sha2.
17113 (TARGET_AES): New feature flag for aes.
17114 * config/aarch64/aarch64-simd.md:
17115 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17116 conditional on TARGET_AES.
17117 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17118 (aarch64_crypto_sha1hsi): Make pattern conditional
17120 (aarch64_crypto_sha1hv4si): Ditto.
17121 (aarch64_be_crypto_sha1hv4si): Ditto.
17122 (aarch64_crypto_sha1su1v4si): Ditto.
17123 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17124 (aarch64_crypto_sha1su0v4si): Ditto.
17125 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17126 (aarch64_crypto_sha256su0v4si): Ditto.
17127 (aarch64_crypto_sha256su1v4si): Ditto.
17128 (doc/invoke.texi): Document new aes and sha2 options.
17130 2018-01-10 Martin Sebor <msebor@redhat.com>
17132 PR tree-optimization/83781
17133 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17136 2018-01-11 Martin Sebor <msebor@gmail.com>
17137 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17139 PR tree-optimization/83501
17140 PR tree-optimization/81703
17142 * tree-ssa-strlen.c (get_string_cst): Rename...
17143 (get_string_len): ...to this. Handle global constants.
17144 (handle_char_store): Adjust.
17146 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17147 Jim Wilson <jimw@sifive.com>
17149 * config/riscv/riscv-protos.h (riscv_output_return): New.
17150 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17151 (riscv_attribute_table, riscv_output_return),
17152 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17153 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17154 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17156 (riscv_expand_prologue): Add early return for naked function.
17157 (riscv_expand_epilogue): Likewise.
17158 (riscv_function_ok_for_sibcall): Return false for naked function.
17159 (riscv_set_current_function): New.
17160 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17161 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17162 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17163 * doc/extend.texi (RISC-V Function Attributes): New.
17165 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17167 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17168 check for 128-bit long double before checking TCmode.
17169 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17170 128-bit long doubles before checking TFmode or TCmode.
17171 (FLOAT128_IBM_P): Likewise.
17173 2018-01-10 Martin Sebor <msebor@redhat.com>
17175 PR tree-optimization/83671
17176 * builtins.c (c_strlen): Unconditionally return zero for the empty
17178 Use -Warray-bounds for warnings.
17179 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17180 for non-constant array indices with COMPONENT_REF, arrays of
17181 arrays, and pointers to arrays.
17182 (gimple_fold_builtin_strlen): Determine and set length range for
17183 non-constant character arrays.
17185 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17187 PR middle-end/81897
17188 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17191 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17193 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17195 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17198 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17199 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17200 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17201 indexed_or_indirect_operand predicate.
17202 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17203 (*vsx_le_perm_load_v8hi): Likewise.
17204 (*vsx_le_perm_load_v16qi): Likewise.
17205 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17206 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17207 (*vsx_le_perm_store_v8hi): Likewise.
17208 (*vsx_le_perm_store_v16qi): Likewise.
17209 (eight unnamed splitters): Likewise.
17211 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17213 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17214 * config/rs6000/emmintrin.h: Likewise.
17215 * config/rs6000/mmintrin.h: Likewise.
17216 * config/rs6000/xmmintrin.h: Likewise.
17218 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17221 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17223 * tree.c (tree_nop_conversion): Return true for location wrapper
17225 (maybe_wrap_with_location): New function.
17226 (selftest::check_strip_nops): New function.
17227 (selftest::test_location_wrappers): New function.
17228 (selftest::tree_c_tests): Call it.
17229 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17230 (maybe_wrap_with_location): New decl.
17231 (EXPR_LOCATION_WRAPPER_P): New macro.
17232 (location_wrapper_p): New inline function.
17233 (tree_strip_any_location_wrapper): New inline function.
17235 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17238 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17239 stack_realign_offset for the largest alignment of stack slot
17241 (ix86_find_max_used_stack_alignment): New function.
17242 (ix86_finalize_stack_frame_flags): Use it. Set
17243 max_used_stack_alignment if we don't realign stack.
17244 * config/i386/i386.h (machine_function): Add
17245 max_used_stack_alignment.
17247 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17249 * config/arm/arm.opt (-mbranch-cost): New option.
17250 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17253 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17256 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17257 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17259 2018-01-10 Richard Biener <rguenther@suse.de>
17262 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17263 early out so it also covers the case where we have a non-NULL
17266 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17268 PR tree-optimization/83753
17269 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17270 for non-strided grouped accesses if the number of elements is 1.
17272 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17275 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17276 * i386.h (TARGET_USE_GATHER): Define.
17277 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17279 2018-01-10 Martin Liska <mliska@suse.cz>
17282 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17283 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17285 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17286 CLEANUP_NO_PARTITIONING is not set.
17288 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17290 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17291 for vectors, as a partial revert of r254296.
17292 * rtl.h (const_vec_p): Delete.
17293 (const_vec_duplicate_p): Don't test for vector CONSTs.
17294 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17295 * expmed.c (make_tree): Likewise.
17298 * common.md (E, F): Use CONSTANT_P instead of checking for
17300 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17301 checking for CONST_VECTOR.
17303 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17305 PR middle-end/83575
17306 * predict.c (force_edge_cold): Handle in more sane way edges
17307 with no prediction.
17309 2018-01-09 Carl Love <cel@us.ibm.com>
17311 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17313 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17314 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17315 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17316 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17317 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17318 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17319 * config/rs6000/rs6000-protos.h: Add extern defition for
17320 rs6000_generate_float2_double_code.
17321 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17323 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17324 (float2_v2df): Add define_expand.
17326 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17329 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17330 op_mode in the force_to_mode call.
17332 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17334 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17335 instead of checking each element individually.
17336 (aarch64_evpc_uzp): Likewise.
17337 (aarch64_evpc_zip): Likewise.
17338 (aarch64_evpc_ext): Likewise.
17339 (aarch64_evpc_rev): Likewise.
17340 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17341 instead of checking each element individually. Return true without
17343 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17344 whether all selected elements come from the same input, instead of
17345 checking each element individually. Remove calls to gen_rtx_REG,
17346 start_sequence and end_sequence and instead assert that no rtl is
17349 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17351 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17352 order of HIGH and CONST checks.
17354 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17356 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17357 if the destination isn't an SSA_NAME.
17359 2018-01-09 Richard Biener <rguenther@suse.de>
17361 PR tree-optimization/83668
17362 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17364 (canonicalize_loop_form): ... here, renamed from ...
17365 (canonicalize_loop_closed_ssa_form): ... this and amended to
17366 swap successor edges for loop exit blocks to make us use
17367 the RPO order we need for initial schedule generation.
17369 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17371 PR tree-optimization/64811
17372 * match.pd: When optimizing comparisons with Inf, avoid
17373 introducing or losing exceptions from comparisons with NaN.
17375 2018-01-09 Martin Liska <mliska@suse.cz>
17378 * asan.c (shadow_mem_size): Add gcc_assert.
17380 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17382 Don't save registers in main().
17385 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17386 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17387 * config/avr/avr.c (avr_set_current_function): Don't error if
17388 naked, OS_task or OS_main are specified at the same time.
17389 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17391 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17393 * common/config/avr/avr-common.c (avr_option_optimization_table):
17394 Switch on -mmain-is-OS_task for optimizing compilations.
17396 2018-01-09 Richard Biener <rguenther@suse.de>
17398 PR tree-optimization/83572
17399 * graphite.c: Include cfganal.h.
17400 (graphite_transform_loops): Connect infinite loops to exit
17401 and remove fake edges at the end.
17403 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17405 * ipa-inline.c (edge_badness): Revert accidental checkin.
17407 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17410 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17411 symbols; not inline clones.
17413 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17416 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17417 hard registers. Formatting fixes.
17419 PR preprocessor/83722
17420 * gcc.c (try_generate_repro): Pass
17421 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17422 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17425 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17426 Kito Cheng <kito.cheng@gmail.com>
17428 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17429 (riscv_leaf_function_p): Delete.
17430 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17432 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17434 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17436 (do_ifelse): New function.
17437 (do_isel): New function.
17438 (do_sub3): New function.
17439 (do_add3): New function.
17440 (do_load_mask_compare): New function.
17441 (do_overlap_load_compare): New function.
17442 (expand_compare_loop): New function.
17443 (expand_block_compare): Call expand_compare_loop() when appropriate.
17444 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17445 option description.
17446 (-mblock-compare-inline-loop-limit): New option.
17448 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17451 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17452 Reverse order of second and third operands in first alternative.
17453 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17454 of first and second elements in UNSPEC_VPERMR vector.
17455 (altivec_expand_vec_perm_le): Likewise.
17457 2018-01-08 Jeff Law <law@redhat.com>
17459 PR rtl-optimizatin/81308
17460 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17461 (process_switch): If group_case_labels makes a change, then set
17463 (pass_convert_switch::execute): If a switch is converted, then
17464 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17466 PR rtl-optimization/81308
17467 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17470 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17472 PR target/83663 - Revert r255946
17473 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17474 generation for cases where splatting a value is not useful.
17475 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17476 across a vec_duplicate and a paradoxical subreg forming a vector
17477 mode to a vec_concat.
17479 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17481 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17482 -march=armv8.3-a variants.
17483 * config/arm/t-multilib: Likewise.
17484 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17486 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17488 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17490 (cceq_ior_compare_complement): Give it a name so I can use it, and
17491 change boolean_or_operator predicate to boolean_operator so it can
17492 be used to generate a crand.
17493 (eqne): New code iterator.
17494 (bd/bd_neg): New code_attrs.
17495 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17496 a single define_insn.
17497 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17498 decrement (bdnzt/bdnzf/bdzt/bdzf).
17499 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17500 with the new names of the branch decrement patterns, and added the
17501 names of the branch decrement conditional patterns.
17503 2018-01-08 Richard Biener <rguenther@suse.de>
17505 PR tree-optimization/83563
17506 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17509 2018-01-08 Richard Biener <rguenther@suse.de>
17511 PR middle-end/83713
17512 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17514 2018-01-08 Richard Biener <rguenther@suse.de>
17516 PR tree-optimization/83685
17517 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17518 references to abnormals.
17520 2018-01-08 Richard Biener <rguenther@suse.de>
17523 * dwarf2out.c (output_indirect_strings): Handle empty
17524 skeleton_debug_str_hash.
17525 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17527 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17529 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17530 (emit_store_direct): Likewise.
17531 (arc_trampoline_adjust_address): Likewise.
17532 (arc_asm_trampoline_template): New function.
17533 (arc_initialize_trampoline): Use asm_trampoline_template.
17534 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17535 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17536 * config/arc/arc.md (flush_icache): Delete pattern.
17538 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17540 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17541 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17544 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17547 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17548 by not USED_FOR_TARGET.
17549 (make_pass_resolve_sw_modes): Likewise.
17551 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17553 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17556 2018-01-08 Richard Biener <rguenther@suse.de>
17558 PR middle-end/83580
17559 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17561 2018-01-08 Richard Biener <rguenther@suse.de>
17563 PR middle-end/83517
17564 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17566 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17568 PR middle-end/81897
17569 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17570 basic blocks with a small number of successors.
17571 (convert_control_dep_chain_into_preds): Improve handling of
17573 (dump_predicates): Split apart into...
17574 (dump_pred_chain): ...here...
17575 (dump_pred_info): ...and here.
17576 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17577 (can_chain_union_be_invalidated_p): Improve check for invalidation
17579 (uninit_uses_cannot_happen): Avoid unnecessary if
17580 convert_control_dep_chain_into_preds yielded nothing.
17582 2018-01-06 Martin Sebor <msebor@redhat.com>
17584 PR tree-optimization/83640
17585 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17586 subtracting negative offset from size.
17587 (builtin_access::overlap): Adjust offset bounds of the access to fall
17588 within the size of the object if possible.
17590 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17592 PR rtl-optimization/83699
17593 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17594 extract_bit_field_as_subreg to cases in which the extracted
17595 value is also a vector.
17597 * lra-constraints.c (process_alt_operands): Test for the equivalence
17598 substitutions when detecting a possible reload cycle.
17600 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17603 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17604 by default if flag_selective_schedling{,2}. Formatting fixes.
17606 PR rtl-optimization/83682
17607 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17608 if it has non-VECTOR_MODE element mode.
17609 (vec_duplicate_p): Likewise.
17611 PR middle-end/83694
17612 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17613 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17615 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17618 * config/i386/i386-builtin.def
17619 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17620 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17621 Require also OPTION_MASK_ISA_AVX512F in addition to
17622 OPTION_MASK_ISA_GFNI.
17623 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17624 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17625 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17626 to OPTION_MASK_ISA_GFNI.
17627 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17628 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17629 OPTION_MASK_ISA_AVX512BW.
17630 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17631 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17632 addition to OPTION_MASK_ISA_GFNI.
17633 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17634 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17635 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17636 to OPTION_MASK_ISA_GFNI.
17637 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17638 a requirement for all ISAs rather than any of them with a few
17640 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17642 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17643 bitmasks to be enabled with 3 exceptions, instead of requiring any
17644 enabled ISA with lots of exceptions.
17645 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17646 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17647 Change avx512bw in isa attribute to avx512f.
17648 * config/i386/sgxintrin.h: Add license boilerplate.
17649 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17650 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17651 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17652 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17654 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17655 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17656 temporarily sse2 rather than sse if not enabled already.
17659 * config/i386/sse.md (VI248_VLBW): Rename to ...
17660 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17661 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17662 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17663 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17664 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17665 mode iterator instead of VI248_VLBW.
17667 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17669 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17670 (record_modified): Skip clobbers; add debug output.
17671 (param_change_prob): Use sreal frequencies.
17673 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17675 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17676 punt for user-aligned variables.
17678 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17680 * tree-chrec.c (chrec_contains_symbols): Return true for
17683 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17686 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17687 of (x|y) == x for BICS pattern.
17689 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17691 PR tree-optimization/83605
17692 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17693 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17696 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17698 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17699 * config/epiphany/rtems.h: New file.
17701 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17702 Uros Bizjak <ubizjak@gmail.com>
17705 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17706 QIreg_operand instead of register_operand predicate.
17707 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17708 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17709 comments instead of -fmitigate[-_]rop.
17711 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17714 * cgraphunit.c (symbol_table::compile): Switch to text_section
17715 before calling assembly_start debug hook.
17716 * run-rtl-passes.c (run_rtl_passes): Likewise.
17719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17721 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17722 range_int_cst_p rather than !symbolic_range_p before calling
17723 extract_range_from_multiplicative_op_1.
17725 2018-01-04 Jeff Law <law@redhat.com>
17727 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17728 redundant test in assertion.
17730 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17732 * doc/rtl.texi: Document machine_mode wrapper classes.
17734 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17736 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17737 using tree_to_uhwi.
17739 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17741 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17742 the VEC_PERM_EXPR fold to fail.
17744 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17747 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17748 to switched_sections.
17750 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17753 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17754 test for d.testing.
17756 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17759 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17760 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17762 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17765 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17766 is BLKmode and bitpos not zero or mode change is needed.
17768 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17771 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17774 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17777 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17778 instead of MULT rtx. Update all corresponding splitters.
17779 (*saddl_se): Ditto.
17780 (*ssub<modesuffix>): Ditto.
17781 (*ssubl_se): Ditto.
17782 (*cmp_sadd_di): Update split patterns.
17783 (*cmp_sadd_si): Ditto.
17784 (*cmp_sadd_sidi): Ditto.
17785 (*cmp_ssub_di): Ditto.
17786 (*cmp_ssub_si): Ditto.
17787 (*cmp_ssub_sidi): Ditto.
17788 * config/alpha/predicates.md (const23_operand): New predicate.
17789 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17790 Look for ASHIFT, not MULT inner operand.
17791 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17793 2018-01-04 Martin Liska <mliska@suse.cz>
17795 PR gcov-profile/83669
17796 * gcov.c (output_intermediate_file): Add version to intermediate
17798 * doc/gcov.texi: Document new field 'version' in intermediate
17799 file format. Fix location of '-k' option of gcov command.
17801 2018-01-04 Martin Liska <mliska@suse.cz>
17804 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17806 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17808 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17810 2018-01-03 Martin Sebor <msebor@redhat.com>
17812 PR tree-optimization/83655
17813 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17814 checking calls with invalid arguments.
17816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17818 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17819 (vectorizable_mask_load_store): Delete.
17820 (vectorizable_call): Return false for masked loads and stores.
17821 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17822 instead of gimple_assign_rhs1.
17823 (vectorizable_load): Handle IFN_MASK_LOAD.
17824 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17828 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17830 (vectorizable_mask_load_store): ...here.
17831 (vectorizable_load): ...and here.
17833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17835 * tree-vect-stmts.c (vect_build_all_ones_mask)
17836 (vect_build_zero_merge_argument): New functions, split out from...
17837 (vectorizable_load): ...here.
17839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17841 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17843 (vectorizable_mask_load_store): ...here.
17844 (vectorizable_store): ...and here.
17846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17848 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17850 (vectorizable_mask_load_store): ...here.
17852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17854 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17855 (vect_model_store_cost): Take a vec_load_store_type instead of a
17857 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17858 (vect_model_store_cost): Take a vec_load_store_type instead of a
17860 (vectorizable_mask_load_store): Update accordingly.
17861 (vectorizable_store): Likewise.
17862 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17866 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17867 IFN_MASK_LOAD calls here rather than...
17868 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17871 Alan Hayward <alan.hayward@arm.com>
17872 David Sherwood <david.sherwood@arm.com>
17874 * expmed.c (extract_bit_field_1): For vector extracts,
17875 fall back to extract_bit_field_as_subreg if vec_extract
17878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17879 Alan Hayward <alan.hayward@arm.com>
17880 David Sherwood <david.sherwood@arm.com>
17882 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17883 they are variable or constant sized.
17884 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17885 slots for constant-sized data.
17887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17888 Alan Hayward <alan.hayward@arm.com>
17889 David Sherwood <david.sherwood@arm.com>
17891 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17892 handling COND_EXPRs with boolean comparisons, try to find a better
17893 basis for the mask type than the boolean itself.
17895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17897 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17898 is calculated and how it can be overridden.
17899 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17900 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17902 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17906 Alan Hayward <alan.hayward@arm.com>
17907 David Sherwood <david.sherwood@arm.com>
17909 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17910 Remove the mode argument.
17911 (aarch64_simd_valid_immediate): Remove the mode and inverse
17913 * config/aarch64/iterators.md (bitsize): New iterator.
17914 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17915 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17916 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17917 aarch64_simd_valid_immediate.
17918 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17919 (aarch64_reg_or_bic_imm): Likewise.
17920 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17921 with an insn_type enum and msl with a modifier_type enum.
17922 Replace element_width with a scalar_mode. Change the shift
17923 to unsigned int. Add constructors for scalar_float_mode and
17924 scalar_int_mode elements.
17925 (aarch64_vect_float_const_representable_p): Delete.
17926 (aarch64_can_const_movi_rtx_p)
17927 (aarch64_simd_scalar_immediate_valid_for_move)
17928 (aarch64_simd_make_constant): Update call to
17929 aarch64_simd_valid_immediate.
17930 (aarch64_advsimd_valid_immediate_hs): New function.
17931 (aarch64_advsimd_valid_immediate): Likewise.
17932 (aarch64_simd_valid_immediate): Remove mode and inverse
17933 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17934 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17935 and aarch64_float_const_representable_p on the result.
17936 (aarch64_output_simd_mov_immediate): Remove mode argument.
17937 Update call to aarch64_simd_valid_immediate and use of
17938 simd_immediate_info.
17939 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17947 (mode_nunits): Likewise CONST_MODE_NUNITS.
17948 * machmode.def (ADJUST_NUNITS): Document.
17949 * genmodes.c (mode_data::need_nunits_adj): New field.
17950 (blank_mode): Update accordingly.
17951 (adj_nunits): New variable.
17952 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17954 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17955 listed in adj_nunits.
17956 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17957 listed in adj_nunits. Don't emit case statements for such modes.
17958 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17959 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17960 nothing if adj_nunits is nonnull.
17961 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17962 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17963 (emit_mode_fbit): Update use of print_maybe_const_decl.
17964 (emit_move_size): Likewise. Treat the array as non-const
17966 (emit_mode_adjustments): Handle adj_nunits.
17968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17970 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17971 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17972 (VECTOR_MODES): Use it.
17973 (make_vector_modes): Take the prefix as an argument.
17975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17976 Alan Hayward <alan.hayward@arm.com>
17977 David Sherwood <david.sherwood@arm.com>
17979 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17980 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17981 for MODE_VECTOR_BOOL.
17982 * machmode.def (VECTOR_BOOL_MODE): Document.
17983 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17984 (make_vector_bool_mode): New function.
17985 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17987 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17988 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17990 * stor-layout.c (int_mode_for_mode): Likewise.
17991 * tree.c (build_vector_type_for_mode): Likewise.
17992 * varasm.c (output_constant_pool_2): Likewise.
17993 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17994 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17995 for MODE_VECTOR_BOOL.
17996 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17997 of mode class checks.
17998 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17999 instead of a list of mode class checks.
18000 (expand_vector_scalar_condition): Likewise.
18001 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18004 Alan Hayward <alan.hayward@arm.com>
18005 David Sherwood <david.sherwood@arm.com>
18007 * machmode.h (mode_size): Change from unsigned short to
18009 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18010 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18011 or if measurement_type is not polynomial.
18012 (fixed_size_mode::includes_p): Check for constant-sized modes.
18013 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18014 return a poly_uint16 rather than an unsigned short.
18015 (emit_mode_size): Change the type of mode_size from unsigned short
18016 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18017 (emit_mode_adjustments): Cope with polynomial vector sizes.
18018 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18020 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18022 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18023 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18024 * caller-save.c (setup_save_areas): Likewise.
18025 (replace_reg_with_saved_mem): Likewise.
18026 * calls.c (emit_library_call_value_1): Likewise.
18027 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18028 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18029 (gen_lowpart_for_combine): Likewise.
18030 * convert.c (convert_to_integer_1): Likewise.
18031 * cse.c (equiv_constant, cse_insn): Likewise.
18032 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18033 (cselib_subst_to_values): Likewise.
18034 * dce.c (word_dce_process_block): Likewise.
18035 * df-problems.c (df_word_lr_mark_ref): Likewise.
18036 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18037 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18038 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18039 (rtl_for_decl_location): Likewise.
18040 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18041 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18042 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18043 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18044 (expand_expr_real_1): Likewise.
18045 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18046 (pad_below): Likewise.
18047 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18048 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18049 * ira.c (get_subreg_tracking_sizes): Likewise.
18050 * ira-build.c (ira_create_allocno_objects): Likewise.
18051 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18052 (ira_sort_regnos_for_alter_reg): Likewise.
18053 * ira-costs.c (record_operand_costs): Likewise.
18054 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18055 (resolve_simple_move): Likewise.
18056 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18057 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18058 (lra_constraints): Likewise.
18059 (CONST_POOL_OK_P): Reject variable-sized modes.
18060 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18061 (add_pseudo_to_slot, lra_spill): Likewise.
18062 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18063 * optabs-query.c (get_best_extraction_insn): Likewise.
18064 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18065 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18066 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18067 * recog.c (offsettable_address_addr_space_p): Likewise.
18068 * regcprop.c (maybe_mode_change): Likewise.
18069 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18070 * regrename.c (build_def_use): Likewise.
18071 * regstat.c (dump_reg_info): Likewise.
18072 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18073 (find_reloads, find_reloads_subreg_address): Likewise.
18074 * reload1.c (eliminate_regs_1): Likewise.
18075 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18076 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18077 (simplify_binary_operation_1, simplify_subreg): Likewise.
18078 * targhooks.c (default_function_arg_padding): Likewise.
18079 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18080 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18081 (verify_gimple_assign_ternary): Likewise.
18082 * tree-inline.c (estimate_move_cost): Likewise.
18083 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18084 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18085 (get_address_cost_ainc): Likewise.
18086 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18087 (vect_supportable_dr_alignment): Likewise.
18088 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18089 (vectorizable_reduction): Likewise.
18090 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18091 (vectorizable_operation, vectorizable_load): Likewise.
18092 * tree.c (build_same_sized_truth_vector_type): Likewise.
18093 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18094 * var-tracking.c (emit_note_insn_var_location): Likewise.
18095 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18096 (ADDR_VEC_ALIGN): Likewise.
18098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18099 Alan Hayward <alan.hayward@arm.com>
18100 David Sherwood <david.sherwood@arm.com>
18102 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18104 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18105 or if measurement_type is polynomial.
18106 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18107 * combine.c (make_extraction): Likewise.
18108 * dse.c (find_shift_sequence): Likewise.
18109 * dwarf2out.c (mem_loc_descriptor): Likewise.
18110 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18111 (extract_bit_field, extract_low_bits): Likewise.
18112 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18113 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18114 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18115 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18116 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18117 * reload.c (find_reloads): Likewise.
18118 * reload1.c (alter_reg): Likewise.
18119 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18120 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18121 * tree-if-conv.c (predicate_mem_writes): Likewise.
18122 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18123 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18124 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18125 * valtrack.c (dead_debug_insert_temp): Likewise.
18126 * varasm.c (mergeable_constant_section): Likewise.
18127 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18130 Alan Hayward <alan.hayward@arm.com>
18131 David Sherwood <david.sherwood@arm.com>
18133 * expr.c (expand_assignment): Cope with polynomial mode sizes
18134 when assigning to a CONCAT.
18136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18137 Alan Hayward <alan.hayward@arm.com>
18138 David Sherwood <david.sherwood@arm.com>
18140 * machmode.h (mode_precision): Change from unsigned short to
18142 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18144 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18145 or if measurement_type is not polynomial.
18146 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18147 in which the mode is already known to be a scalar_int_mode.
18148 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18149 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18151 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18152 for GET_MODE_PRECISION.
18153 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18154 for GET_MODE_PRECISION.
18155 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18157 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18158 (expand_field_assignment, make_extraction): Likewise.
18159 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18160 (get_last_value): Likewise.
18161 * convert.c (convert_to_integer_1): Likewise.
18162 * cse.c (cse_insn): Likewise.
18163 * expr.c (expand_expr_real_1): Likewise.
18164 * lra-constraints.c (simplify_operand_subreg): Likewise.
18165 * optabs-query.c (can_atomic_load_p): Likewise.
18166 * optabs.c (expand_atomic_load): Likewise.
18167 (expand_atomic_store): Likewise.
18168 * ree.c (combine_reaching_defs): Likewise.
18169 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18170 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18171 * tree.h (type_has_mode_precision_p): Likewise.
18172 * ubsan.c (instrument_si_overflow): Likewise.
18174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18175 Alan Hayward <alan.hayward@arm.com>
18176 David Sherwood <david.sherwood@arm.com>
18178 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18179 polynomial numbers of units.
18180 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18181 (valid_vector_subparts_p): New function.
18182 (build_vector_type): Remove temporary shim and take the number
18183 of units as a poly_uint64 rather than an int.
18184 (build_opaque_vector_type): Take the number of units as a
18185 poly_uint64 rather than an int.
18186 * tree.c (build_vector_from_ctor): Handle polynomial
18187 TYPE_VECTOR_SUBPARTS.
18188 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18189 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18190 (build_vector_from_val): If the number of units is variable,
18191 use build_vec_duplicate_cst for constant operands and
18192 VEC_DUPLICATE_EXPR otherwise.
18193 (make_vector_type): Remove temporary is_constant ().
18194 (build_vector_type, build_opaque_vector_type): Take the number of
18195 units as a poly_uint64 rather than an int.
18196 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18198 * cfgexpand.c (expand_debug_expr): Likewise.
18199 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18200 (store_constructor, expand_expr_real_1): Likewise.
18201 (const_scalar_mask_from_tree): Likewise.
18202 * fold-const-call.c (fold_const_reduction): Likewise.
18203 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18204 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18205 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18206 (fold_relational_const): Likewise.
18207 (native_interpret_vector): Likewise. Change the size from an
18208 int to an unsigned int.
18209 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18210 TYPE_VECTOR_SUBPARTS.
18211 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18212 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18213 duplicating a non-constant operand into a variable-length vector.
18214 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18215 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18216 * ipa-icf.c (sem_variable::equals): Likewise.
18217 * match.pd: Likewise.
18218 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18219 * print-tree.c (print_node): Likewise.
18220 * stor-layout.c (layout_type): Likewise.
18221 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18222 * tree-cfg.c (verify_gimple_comparison): Likewise.
18223 (verify_gimple_assign_binary): Likewise.
18224 (verify_gimple_assign_ternary): Likewise.
18225 (verify_gimple_assign_single): Likewise.
18226 * tree-pretty-print.c (dump_generic_node): Likewise.
18227 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18228 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18229 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18230 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18231 (vect_shift_permute_load_chain): Likewise.
18232 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18233 (expand_vector_condition, optimize_vector_constructor): Likewise.
18234 (lower_vec_perm, get_compute_type): Likewise.
18235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18236 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18237 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18238 (vect_recog_mask_conversion_pattern): Likewise.
18239 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18240 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18241 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18242 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18243 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18244 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18245 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18246 (supportable_widening_operation): Likewise.
18247 (supportable_narrowing_operation): Likewise.
18248 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18250 * varasm.c (output_constant): Likewise.
18252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18253 Alan Hayward <alan.hayward@arm.com>
18254 David Sherwood <david.sherwood@arm.com>
18256 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18257 so that both the length == 3 and length != 3 cases set up their
18258 own permute vectors. Add comments explaining why we know the
18259 number of elements is constant.
18260 (vect_permute_load_chain): Likewise.
18262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18263 Alan Hayward <alan.hayward@arm.com>
18264 David Sherwood <david.sherwood@arm.com>
18266 * machmode.h (mode_nunits): Change from unsigned char to
18268 (ONLY_FIXED_SIZE_MODES): New macro.
18269 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18270 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18271 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18273 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18274 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18275 or if measurement_type is not polynomial.
18276 * genmodes.c (ZERO_COEFFS): New macro.
18277 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18279 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18280 Use ZERO_COEFFS when emitting initializers.
18281 * data-streamer.h (bp_pack_poly_value): New function.
18282 (bp_unpack_poly_value): Likewise.
18283 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18284 for GET_MODE_NUNITS.
18285 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18286 for GET_MODE_NUNITS.
18287 * tree.c (make_vector_type): Remove temporary shim and make
18288 the real function take the number of units as a poly_uint64
18289 rather than an int.
18290 (build_vector_type_for_mode): Handle polynomial nunits.
18291 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18292 * emit-rtl.c (const_vec_series_p_1): Likewise.
18293 (gen_rtx_CONST_VECTOR): Likewise.
18294 * fold-const.c (test_vec_duplicate_folding): Likewise.
18295 * genrecog.c (validate_pattern): Likewise.
18296 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18297 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18298 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18299 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18300 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18301 * rtlanal.c (subreg_get_info): Likewise.
18302 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18303 (vect_grouped_load_supported): Likewise.
18304 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18305 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18306 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18307 (simplify_const_unary_operation, simplify_binary_operation_1)
18308 (simplify_const_binary_operation, simplify_ternary_operation)
18309 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18310 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18311 instead of CONST_VECTOR_NUNITS.
18312 * varasm.c (output_constant_pool_2): Likewise.
18313 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18314 explicit-encoded elements in the XVEC for variable-length vectors.
18316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18318 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18321 Alan Hayward <alan.hayward@arm.com>
18322 David Sherwood <david.sherwood@arm.com>
18324 * coretypes.h (fixed_size_mode): Declare.
18325 (fixed_size_mode_pod): New typedef.
18326 * builtins.h (target_builtins::x_apply_args_mode)
18327 (target_builtins::x_apply_result_mode): Change type to
18328 fixed_size_mode_pod.
18329 * builtins.c (apply_args_size, apply_result_size, result_vector)
18330 (expand_builtin_apply_args_1, expand_builtin_apply)
18331 (expand_builtin_return): Update accordingly.
18333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18335 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18336 * cselib.c (cselib_hash_rtx): Likewise.
18337 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18338 CONST_VECTOR encoding.
18340 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18341 Jeff Law <law@redhat.com>
18344 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18345 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18346 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18347 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18350 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18351 explicitly probe *sp in a noreturn function if there were any callee
18352 register saves or frame pointer is needed.
18354 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18357 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18358 BLKmode for ternary, binary or unary expressions.
18361 * var-tracking.c (delete_vta_debug_insn): New inline function.
18362 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18363 insns from get_insns () to NULL instead of each bb separately.
18364 Use delete_vta_debug_insn. No longer static.
18365 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18366 delete_vta_debug_insns callers.
18367 * rtl.h (delete_vta_debug_insns): Declare.
18368 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18369 instead of variable_tracking_main.
18371 2018-01-03 Martin Sebor <msebor@redhat.com>
18373 PR tree-optimization/83603
18374 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18375 arguments past the endof the argument list in functions declared
18376 without a prototype.
18377 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18378 Avoid checking when arguments are null.
18380 2018-01-03 Martin Sebor <msebor@redhat.com>
18383 * doc/extend.texi (attribute const): Fix a typo.
18384 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18385 issuing -Wsuggest-attribute for void functions.
18387 2018-01-03 Martin Sebor <msebor@redhat.com>
18389 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18390 offset_int::from instead of wide_int::to_shwi.
18391 (maybe_diag_overlap): Remove assertion.
18392 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18393 * gimple-ssa-sprintf.c (format_directive): Same.
18394 (parse_directive): Same.
18395 (sprintf_dom_walker::compute_format_length): Same.
18396 (try_substitute_return_value): Same.
18398 2018-01-03 Jeff Law <law@redhat.com>
18400 PR middle-end/83654
18401 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18402 non-constant residual for zero at runtime and avoid probing in
18403 that case. Reorganize code for trailing problem to mirror handling
18406 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18408 PR tree-optimization/83501
18409 * tree-ssa-strlen.c (get_string_cst): New.
18410 (handle_char_store): Call get_string_cst.
18412 2018-01-03 Martin Liska <mliska@suse.cz>
18414 PR tree-optimization/83593
18415 * tree-ssa-strlen.c: Include tree-cfg.h.
18416 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18417 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18418 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18420 (strlen_dom_walker::before_dom_children): Call
18421 gimple_purge_dead_eh_edges. Dump tranformation with details
18423 (strlen_dom_walker::before_dom_children): Update call by adding
18424 new argument cleanup_eh.
18425 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18427 2018-01-03 Martin Liska <mliska@suse.cz>
18430 * cif-code.def (VARIADIC_THUNK): New enum value.
18431 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18434 2018-01-03 Jan Beulich <jbeulich@suse.com>
18436 * sse.md (mov<mode>_internal): Tighten condition for when to use
18437 vmovdqu<ssescalarsize> for TI and OI modes.
18439 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18441 Update copyright years.
18443 2018-01-03 Martin Liska <mliska@suse.cz>
18446 * ipa-visibility.c (function_and_variable_visibility): Skip
18447 functions with noipa attribure.
18449 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18451 * gcc.c (process_command): Update copyright notice dates.
18452 * gcov-dump.c (print_version): Ditto.
18453 * gcov.c (print_version): Ditto.
18454 * gcov-tool.c (print_version): Ditto.
18455 * gengtype.c (create_file): Ditto.
18456 * doc/cpp.texi: Bump @copying's copyright year.
18457 * doc/cppinternals.texi: Ditto.
18458 * doc/gcc.texi: Ditto.
18459 * doc/gccint.texi: Ditto.
18460 * doc/gcov.texi: Ditto.
18461 * doc/install.texi: Ditto.
18462 * doc/invoke.texi: Ditto.
18464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18466 * vector-builder.h (vector_builder::m_full_nelts): Change from
18467 unsigned int to poly_uint64.
18468 (vector_builder::full_nelts): Update prototype accordingly.
18469 (vector_builder::new_vector): Likewise.
18470 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18471 (vector_builder::operator ==): Likewise.
18472 (vector_builder::finalize): Likewise.
18473 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18474 Take the number of elements as a poly_uint64 rather than an
18476 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18477 from unsigned int to poly_uint64.
18478 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18479 (vec_perm_indices::new_vector): Likewise.
18480 (vec_perm_indices::length): Likewise.
18481 (vec_perm_indices::nelts_per_input): Likewise.
18482 (vec_perm_indices::input_nelts): Likewise.
18483 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18484 number of elements per input as a poly_uint64 rather than an
18485 unsigned int. Use the original encoding for variable-length
18486 vectors, rather than clamping each individual element.
18487 For the second and subsequent elements in each pattern,
18488 clamp the step and base before clamping their sum.
18489 (vec_perm_indices::series_p): Handle polynomial element counts.
18490 (vec_perm_indices::all_in_range_p): Likewise.
18491 (vec_perm_indices_to_tree): Likewise.
18492 (vec_perm_indices_to_rtx): Likewise.
18493 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18494 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18495 (tree_vector_builder::new_binary_operation): Handle polynomial
18496 element counts. Return false if we need to know the number
18497 of elements at compile time.
18498 * fold-const.c (fold_vec_perm): Punt if the number of elements
18499 isn't known at compile time.
18501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18503 * vec-perm-indices.h (vec_perm_builder): Change element type
18504 from HOST_WIDE_INT to poly_int64.
18505 (vec_perm_indices::element_type): Update accordingly.
18506 (vec_perm_indices::clamp): Handle polynomial element_types.
18507 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18508 (vec_perm_indices::all_in_range_p): Likewise.
18509 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18511 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18512 polynomial vec_perm_indices element types.
18513 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18514 * fold-const.c (fold_vec_perm): Likewise.
18515 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18516 * tree-vect-generic.c (lower_vec_perm): Likewise.
18517 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18518 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18519 element type to HOST_WIDE_INT.
18521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18522 Alan Hayward <alan.hayward@arm.com>
18523 David Sherwood <david.sherwood@arm.com>
18525 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18526 rather than an int. Use plus_constant.
18527 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18528 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18531 Alan Hayward <alan.hayward@arm.com>
18532 David Sherwood <david.sherwood@arm.com>
18534 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18535 a HOST_WIDE_INT to a poly_int64.
18537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18538 Alan Hayward <alan.hayward@arm.com>
18539 David Sherwood <david.sherwood@arm.com>
18541 * calls.c (load_register_parameters): Cope with polynomial
18542 mode sizes. Require a constant size for BLKmode parameters
18543 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18544 forces a parameter to be padded at the lsb end in order to
18545 fill a complete number of words, require the parameter size
18546 to be ordered wrt UNITS_PER_WORD.
18548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18549 Alan Hayward <alan.hayward@arm.com>
18550 David Sherwood <david.sherwood@arm.com>
18552 * reload1.c (spill_stack_slot_width): Change element type
18553 from unsigned int to poly_uint64_pod.
18554 (alter_reg): Treat mode sizes as polynomial.
18556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18557 Alan Hayward <alan.hayward@arm.com>
18558 David Sherwood <david.sherwood@arm.com>
18560 * reload.c (complex_word_subreg_p): New function.
18561 (reload_inner_reg_of_subreg, push_reload): Use it.
18563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18564 Alan Hayward <alan.hayward@arm.com>
18565 David Sherwood <david.sherwood@arm.com>
18567 * lra-constraints.c (process_alt_operands): Reject matched
18568 operands whose sizes aren't ordered.
18569 (match_reload): Refer to this check here.
18571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18572 Alan Hayward <alan.hayward@arm.com>
18573 David Sherwood <david.sherwood@arm.com>
18575 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18576 that the mode size is in the set {1, 2, 4, 8, 16}.
18578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18579 Alan Hayward <alan.hayward@arm.com>
18580 David Sherwood <david.sherwood@arm.com>
18582 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18583 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18590 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18591 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18592 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18593 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18594 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18595 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18596 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18597 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18598 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18600 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18601 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18602 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18603 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18604 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18605 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18606 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18607 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18608 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18609 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18611 * expr.c (emit_move_resolve_push): Treat the input and result
18612 of PUSH_ROUNDING as a poly_int64.
18613 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18614 (emit_push_insn): Likewise.
18615 * lra-eliminations.c (mark_not_eliminable): Likewise.
18616 * recog.c (push_operand): Likewise.
18617 * reload1.c (elimination_effects): Likewise.
18618 * rtlanal.c (nonzero_bits1): Likewise.
18619 * calls.c (store_one_arg): Likewise. Require the padding to be
18620 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18627 Use plus_constant instead of gen_rtx_PLUS.
18629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18630 Alan Hayward <alan.hayward@arm.com>
18631 David Sherwood <david.sherwood@arm.com>
18633 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18634 rather than an int.
18636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18637 Alan Hayward <alan.hayward@arm.com>
18638 David Sherwood <david.sherwood@arm.com>
18640 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18641 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18642 via stack temporaries. Treat the mode size as polynomial too.
18644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18645 Alan Hayward <alan.hayward@arm.com>
18646 David Sherwood <david.sherwood@arm.com>
18648 * expr.c (expand_expr_real_2): When handling conversions involving
18649 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18650 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18651 as a poly_uint64 too.
18653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18654 Alan Hayward <alan.hayward@arm.com>
18655 David Sherwood <david.sherwood@arm.com>
18657 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
18664 REGMODE_NATURAL_SIZE.
18665 * expmed.c (store_bit_field_1): Likewise.
18666 * expr.c (store_constructor): Likewise.
18667 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18668 and polynomial REGMODE_NATURAL_SIZE.
18669 (gen_lowpart_common): Likewise.
18670 * reginfo.c (record_subregs_of_mode): Likewise.
18671 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18678 numbers of elements.
18680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18681 Alan Hayward <alan.hayward@arm.com>
18682 David Sherwood <david.sherwood@arm.com>
18684 * match.pd: Cope with polynomial numbers of vector elements.
18686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18687 Alan Hayward <alan.hayward@arm.com>
18688 David Sherwood <david.sherwood@arm.com>
18690 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18691 in a POINTER_PLUS_EXPR.
18693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18694 Alan Hayward <alan.hayward@arm.com>
18695 David Sherwood <david.sherwood@arm.com>
18697 * omp-simd-clone.c (simd_clone_subparts): New function.
18698 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18699 (ipa_simd_modify_function_body): Likewise.
18701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18702 Alan Hayward <alan.hayward@arm.com>
18703 David Sherwood <david.sherwood@arm.com>
18705 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18706 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18707 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18708 (expand_vector_condition, vector_element): Likewise.
18709 (subparts_gt): New function.
18710 (get_compute_type): Use subparts_gt.
18711 (count_type_subparts): Delete.
18712 (expand_vector_operations_1): Use subparts_gt instead of
18713 count_type_subparts.
18715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18716 Alan Hayward <alan.hayward@arm.com>
18717 David Sherwood <david.sherwood@arm.com>
18719 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18720 (vect_compile_time_alias): ...this new function. Do the calculation
18721 on poly_ints rather than trees.
18722 (vect_prune_runtime_alias_test_list): Update call accordingly.
18724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18725 Alan Hayward <alan.hayward@arm.com>
18726 David Sherwood <david.sherwood@arm.com>
18728 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18730 (vect_schedule_slp_instance): Likewise.
18732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18733 Alan Hayward <alan.hayward@arm.com>
18734 David Sherwood <david.sherwood@arm.com>
18736 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18737 constant and extern definitions for variable-length vectors.
18738 (vect_get_constant_vectors): Note that the number of units
18739 is known to be constant.
18741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18742 Alan Hayward <alan.hayward@arm.com>
18743 David Sherwood <david.sherwood@arm.com>
18745 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18746 of units as polynomial. Choose between WIDE and NARROW based
18749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18750 Alan Hayward <alan.hayward@arm.com>
18751 David Sherwood <david.sherwood@arm.com>
18753 * tree-vect-stmts.c (simd_clone_subparts): New function.
18754 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18757 Alan Hayward <alan.hayward@arm.com>
18758 David Sherwood <david.sherwood@arm.com>
18760 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18761 vectors as polynomial. Use build_index_vector for
18762 IFN_GOMP_SIMD_LANE.
18764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18765 Alan Hayward <alan.hayward@arm.com>
18766 David Sherwood <david.sherwood@arm.com>
18768 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18769 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18770 for variable-length vectors.
18771 (vectorizable_mask_load_store): Treat the number of units as
18772 polynomial, asserting that it is constant if the condition has
18773 already been enforced.
18774 (vectorizable_store, vectorizable_load): Likewise.
18776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18777 Alan Hayward <alan.hayward@arm.com>
18778 David Sherwood <david.sherwood@arm.com>
18780 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18781 of units as polynomial. Punt if we can't tell at compile time
18782 which vector contains the final result.
18784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18785 Alan Hayward <alan.hayward@arm.com>
18786 David Sherwood <david.sherwood@arm.com>
18788 * tree-vect-loop.c (vectorizable_induction): Treat the number
18789 of units as polynomial. Punt on SLP inductions. Use an integer
18790 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18791 cast of such a series for variable-length floating-point
18794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18795 Alan Hayward <alan.hayward@arm.com>
18796 David Sherwood <david.sherwood@arm.com>
18798 * tree.h (build_index_vector): Declare.
18799 * tree.c (build_index_vector): New function.
18800 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18801 of units as polynomial, forcibly converting it to a constant if
18802 vectorizable_reduction has already enforced the condition.
18803 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18804 to create a {1,2,3,...} vector.
18805 (vectorizable_reduction): Treat the number of units as polynomial.
18806 Choose vectype_in based on the largest scalar element size rather
18807 than the smallest number of units. Enforce the restrictions
18810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18811 Alan Hayward <alan.hayward@arm.com>
18812 David Sherwood <david.sherwood@arm.com>
18814 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18815 number of units as polynomial.
18817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18818 Alan Hayward <alan.hayward@arm.com>
18819 David Sherwood <david.sherwood@arm.com>
18821 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18822 * target.def (autovectorize_vector_sizes): Return the vector sizes
18823 by pointer, using vector_sizes rather than a bitmask.
18824 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18825 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18826 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18828 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18829 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18830 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18831 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18832 * omp-general.c (omp_max_vf): Likewise.
18833 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18834 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18835 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18836 * tree-vect-slp.c (vect_slp_bb): Likewise.
18837 * doc/tm.texi: Regenerate.
18838 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18840 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18841 the vector size as a poly_uint64 rather than an unsigned int.
18842 (current_vector_size): Change from an unsigned int to a poly_uint64.
18843 (get_vectype_for_scalar_type): Update accordingly.
18844 * tree.h (build_truth_vector_type): Take the size and number of
18845 units as a poly_uint64 rather than an unsigned int.
18846 (build_vector_type): Add a temporary overload that takes
18847 the number of units as a poly_uint64 rather than an unsigned int.
18848 * tree.c (make_vector_type): Likewise.
18849 (build_truth_vector_type): Take the number of units as a poly_uint64
18850 rather than an unsigned int.
18852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18853 Alan Hayward <alan.hayward@arm.com>
18854 David Sherwood <david.sherwood@arm.com>
18856 * target.def (get_mask_mode): Take the number of units and length
18857 as poly_uint64s rather than unsigned ints.
18858 * targhooks.h (default_get_mask_mode): Update accordingly.
18859 * targhooks.c (default_get_mask_mode): Likewise.
18860 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18861 * doc/tm.texi: Regenerate.
18863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18864 Alan Hayward <alan.hayward@arm.com>
18865 David Sherwood <david.sherwood@arm.com>
18867 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18868 * omp-general.c (omp_max_vf): Likewise.
18869 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18870 (expand_omp_simd): Handle polynomial safelen.
18871 * omp-low.c (omplow_simd_context): Add a default constructor.
18872 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18873 (lower_rec_simd_input_clauses): Update accordingly.
18874 (lower_rec_input_clauses): Likewise.
18876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18877 Alan Hayward <alan.hayward@arm.com>
18878 David Sherwood <david.sherwood@arm.com>
18880 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18881 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18882 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18883 (vect_analyze_slp_cost): Likewise.
18884 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18885 (vect_model_load_cost): Likewise.
18887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18888 Alan Hayward <alan.hayward@arm.com>
18889 David Sherwood <david.sherwood@arm.com>
18891 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18892 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18893 from an unsigned int * to a poly_uint64_pod *.
18894 (calculate_unrolling_factor): New function.
18895 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18898 Alan Hayward <alan.hayward@arm.com>
18899 David Sherwood <david.sherwood@arm.com>
18901 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18902 from an unsigned int to a poly_uint64.
18903 (_loop_vec_info::slp_unrolling_factor): Likewise.
18904 (_loop_vec_info::vectorization_factor): Change from an int
18906 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18907 (vect_get_num_vectors): New function.
18908 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18909 (vect_get_num_copies): Use vect_get_num_vectors.
18910 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18911 to an unsigned int *.
18912 (vect_analyze_data_refs): Change min_vf from an int * to a
18914 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18915 than an unsigned HOST_WIDE_INT.
18916 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18917 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18918 to an unsigned int *.
18919 (vect_analyze_data_ref_dependences): Likewise.
18920 (vect_compute_data_ref_alignment): Handle polynomial vf.
18921 (vect_enhance_data_refs_alignment): Likewise.
18922 (vect_prune_runtime_alias_test_list): Likewise.
18923 (vect_shift_permute_load_chain): Likewise.
18924 (vect_supportable_dr_alignment): Likewise.
18925 (dependence_distance_ge_vf): Take the vectorization factor as a
18926 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18927 (vect_analyze_data_refs): Change min_vf from an int * to a
18929 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18930 vfm1 as a poly_uint64 rather than an int. Make the same change
18931 for the returned bound_scalar.
18932 (vect_gen_vector_loop_niters): Handle polynomial vf.
18933 (vect_do_peeling): Likewise. Update call to
18934 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18935 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18937 * tree-vect-loop.c (vect_determine_vectorization_factor)
18938 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18939 (vect_get_known_peeling_cost): Likewise.
18940 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18941 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18942 (vect_transform_loop): Likewise. Use the lowest possible VF when
18943 updating the upper bounds of the loop.
18944 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18945 rather than an int.
18946 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18947 polynomial unroll factors.
18948 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18949 (vect_make_slp_decision): Likewise.
18950 (vect_supported_load_permutation_p): Likewise, and polynomial
18952 (vect_analyze_slp_cost): Handle polynomial vf.
18953 (vect_slp_analyze_node_operations): Likewise.
18954 (vect_slp_analyze_bb_1): Likewise.
18955 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18956 than an unsigned HOST_WIDE_INT.
18957 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18958 (vectorizable_load): Handle polynomial vf.
18959 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18961 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18964 Alan Hayward <alan.hayward@arm.com>
18965 David Sherwood <david.sherwood@arm.com>
18967 * match.pd: Handle bit operations involving three constants
18968 and try to fold one pair.
18970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18972 * tree-vect-loop-manip.c: Include gimple-fold.h.
18973 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18974 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18975 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18976 Add a path that uses a step of VF instead of 1, but disable it
18978 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18979 and niters_no_overflow parameters. Update calls to
18980 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18981 Create a new SSA name if the latter choses to use a ste other
18982 than zero, and return it via niters_vector_mult_vf_var.
18983 * tree-vect-loop.c (vect_transform_loop): Update calls to
18984 vect_do_peeling, vect_gen_vector_loop_niters and
18985 slpeel_make_loop_iterate_ntimes.
18986 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18987 (vect_gen_vector_loop_niters): Update declarations after above changes.
18989 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18991 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18992 128-bit round to integer instructions.
18993 (ceil<mode>2): Likewise.
18994 (btrunc<mode>2): Likewise.
18995 (round<mode>2): Likewise.
18997 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18999 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19000 unaligned VSX load/store on P8/P9.
19001 (expand_block_clear): Allow the use of unaligned VSX
19002 load/store on P8/P9.
19004 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19006 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19008 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19009 swap associated with both a load and a store.
19011 2018-01-02 Andrew Waterman <andrew@sifive.com>
19013 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19014 * config/riscv/riscv.md (clear_cache): Use it.
19016 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19018 * web.c: Remove out-of-date comment.
19020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19022 * expr.c (fixup_args_size_notes): Check that any existing
19023 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19024 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19025 (emit_single_push_insn): ...here.
19027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19029 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19030 (const_vector_encoded_nelts): New function.
19031 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19032 (const_vector_int_elt, const_vector_elt): Declare.
19033 * emit-rtl.c (const_vector_int_elt_1): New function.
19034 (const_vector_elt): Likewise.
19035 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19036 of CONST_VECTOR_ELT.
19038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19040 * expr.c: Include rtx-vector-builder.h.
19041 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19042 directly on the tree encoding.
19043 (const_vector_from_tree): Likewise.
19044 * optabs.c: Include rtx-vector-builder.h.
19045 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19046 sequence of "u" values.
19047 * vec-perm-indices.c: Include rtx-vector-builder.h.
19048 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19049 directly on the vec_perm_indices encoding.
19051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19053 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19054 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19055 * rtx-vector-builder.h: New file.
19056 * rtx-vector-builder.c: Likewise.
19057 * rtl.h (rtx_def::u2): Add a const_vector field.
19058 (CONST_VECTOR_NPATTERNS): New macro.
19059 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19060 (CONST_VECTOR_DUPLICATE_P): Likewise.
19061 (CONST_VECTOR_STEPPED_P): Likewise.
19062 (CONST_VECTOR_ENCODED_ELT): Likewise.
19063 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19064 (unwrap_const_vec_duplicate): Likewise.
19065 (const_vec_series_p): Check for a non-duplicated vector encoding.
19066 Say that the function only returns true for integer vectors.
19067 * emit-rtl.c: Include rtx-vector-builder.h.
19068 (gen_const_vec_duplicate_1): Delete.
19069 (gen_const_vector): Call gen_const_vec_duplicate instead of
19070 gen_const_vec_duplicate_1.
19071 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19072 (gen_const_vec_duplicate): Use rtx_vector_builder.
19073 (gen_const_vec_series): Likewise.
19074 (gen_rtx_CONST_VECTOR): Likewise.
19075 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19076 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19077 Build a new vector rather than modifying a CONST_VECTOR in-place.
19078 (handle_special_swappables): Update call accordingly.
19079 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19080 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19081 Build a new vector rather than modifying a CONST_VECTOR in-place.
19082 (handle_special_swappables): Update call accordingly.
19084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19086 * simplify-rtx.c (simplify_const_binary_operation): Use
19087 CONST_VECTOR_ELT instead of XVECEXP.
19089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19091 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19092 the selector elements to be different from the data elements
19093 if the selector is a VECTOR_CST.
19094 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19095 ssizetype for the selector.
19097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19099 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19100 before testing each element individually.
19101 * tree-vect-generic.c (lower_vec_perm): Likewise.
19103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19105 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19106 * selftest-run-tests.c (selftest::run_tests): Call it.
19107 * vector-builder.h (vector_builder::operator ==): New function.
19108 (vector_builder::operator !=): Likewise.
19109 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19110 (vec_perm_indices::all_from_input_p): New function.
19111 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19112 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19113 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19114 instead of reading the VECTOR_CST directly. Detect whether both
19115 vector inputs are the same before constructing the vec_perm_indices,
19116 and update the number of inputs argument accordingly. Use the
19117 utility functions added above. Only construct sel2 if we need to.
19119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19121 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19122 the broadcast of the low byte.
19123 (expand_mult_highpart): Use an explicit encoding for the permutes.
19124 * optabs-query.c (can_mult_highpart_p): Likewise.
19125 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19126 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19127 (vectorizable_bswap): Likewise.
19128 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19129 explicit encoding for the power-of-2 permutes.
19130 (vect_permute_store_chain): Likewise.
19131 (vect_grouped_load_supported): Likewise.
19132 (vect_permute_load_chain): Likewise.
19134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19136 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19137 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19138 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19139 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19140 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19141 (vect_gen_perm_mask_any): Likewise.
19143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19145 * int-vector-builder.h: New file.
19146 * vec-perm-indices.h: Include int-vector-builder.h.
19147 (vec_perm_indices): Redefine as an int_vector_builder.
19148 (auto_vec_perm_indices): Delete.
19149 (vec_perm_builder): Redefine as a stand-alone class.
19150 (vec_perm_indices::vec_perm_indices): New function.
19151 (vec_perm_indices::clamp): Likewise.
19152 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19153 (vec_perm_indices::new_vector): New function.
19154 (vec_perm_indices::new_expanded_vector): Update for new
19155 vec_perm_indices class.
19156 (vec_perm_indices::rotate_inputs): New function.
19157 (vec_perm_indices::all_in_range_p): Operate directly on the
19158 encoded form, without computing elided elements.
19159 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19160 encoding. Update for new vec_perm_indices class.
19161 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19162 the given vec_perm_builder.
19163 (expand_vec_perm_var): Update vec_perm_builder constructor.
19164 (expand_mult_highpart): Use vec_perm_builder instead of
19165 auto_vec_perm_indices.
19166 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19167 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19168 or double series encoding as appropriate.
19169 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19170 vec_perm_indices instead of auto_vec_perm_indices.
19171 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19172 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19173 (vect_permute_store_chain): Likewise.
19174 (vect_grouped_load_supported): Likewise.
19175 (vect_permute_load_chain): Likewise.
19176 (vect_shift_permute_load_chain): Likewise.
19177 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19178 (vect_transform_slp_perm_load): Likewise.
19179 (vect_schedule_slp_instance): Likewise.
19180 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19181 (vectorizable_mask_load_store): Likewise.
19182 (vectorizable_bswap): Likewise.
19183 (vectorizable_store): Likewise.
19184 (vectorizable_load): Likewise.
19185 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19186 vec_perm_indices instead of auto_vec_perm_indices. Use
19187 tree_to_vec_perm_builder to read the vector from a tree.
19188 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19189 vec_perm_builder instead of a vec_perm_indices.
19190 (have_whole_vector_shift): Use vec_perm_builder and
19191 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19192 truncation to calc_vec_perm_mask_for_shift.
19193 (vect_create_epilog_for_reduction): Likewise.
19194 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19195 from auto_vec_perm_indices to vec_perm_indices.
19196 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19197 instead of changing individual elements.
19198 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19199 the vector in d.perm.
19200 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19201 from auto_vec_perm_indices to vec_perm_indices.
19202 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19203 instead of changing individual elements.
19204 (arm_vectorize_vec_perm_const): Use new_vector to install
19205 the vector in d.perm.
19206 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19207 Update vec_perm_builder constructor.
19208 (rs6000_expand_interleave): Likewise.
19209 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19210 (rs6000_expand_interleave): Likewise.
19212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19214 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19215 to qimode could truncate the indices.
19216 * optabs.c (expand_vec_perm_var): Likewise.
19218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19220 * Makefile.in (OBJS): Add vec-perm-indices.o.
19221 * vec-perm-indices.h: New file.
19222 * vec-perm-indices.c: Likewise.
19223 * target.h (vec_perm_indices): Replace with a forward class
19225 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19226 * optabs.h: Include vec-perm-indices.h.
19227 (expand_vec_perm): Delete.
19228 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19229 (expand_vec_perm_const): Declare.
19230 * target.def (vec_perm_const_ok): Replace with...
19231 (vec_perm_const): ...this new hook.
19232 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19233 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19234 * doc/tm.texi: Regenerate.
19235 * optabs.def (vec_perm_const): Delete.
19236 * doc/md.texi (vec_perm_const): Likewise.
19237 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19238 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19239 expand_vec_perm for constant permutation vectors. Assert that
19240 the mode of variable permutation vectors is the integer equivalent
19241 of the mode that is being permuted.
19242 * optabs-query.h (selector_fits_mode_p): Declare.
19243 * optabs-query.c: Include vec-perm-indices.h.
19244 (selector_fits_mode_p): New function.
19245 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19246 is defined, instead of checking whether the vec_perm_const_optab
19247 exists. Use targetm.vectorize.vec_perm_const instead of
19248 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19249 fit in the vector mode before using a variable permute.
19250 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19251 vec_perm_indices instead of an rtx.
19252 (expand_vec_perm): Replace with...
19253 (expand_vec_perm_const): ...this new function. Take the selector
19254 as a vec_perm_indices rather than an rtx. Also take the mode of
19255 the selector. Update call to shift_amt_for_vec_perm_mask.
19256 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19257 Use vec_perm_indices::new_expanded_vector to expand the original
19258 selector into bytes. Check whether the indices fit in the vector
19259 mode before using a variable permute.
19260 (expand_vec_perm_var): Make global.
19261 (expand_mult_highpart): Use expand_vec_perm_const.
19262 * fold-const.c: Includes vec-perm-indices.h.
19263 * tree-ssa-forwprop.c: Likewise.
19264 * tree-vect-data-refs.c: Likewise.
19265 * tree-vect-generic.c: Likewise.
19266 * tree-vect-loop.c: Likewise.
19267 * tree-vect-slp.c: Likewise.
19268 * tree-vect-stmts.c: Likewise.
19269 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19271 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19272 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19273 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19274 (aarch64_vectorize_vec_perm_const): ...this new function.
19275 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19276 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19277 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19278 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19279 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19280 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19281 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19283 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19284 check for NEON modes.
19285 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19286 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19287 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19288 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19290 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19291 the old VEC_PERM_CONST conditions.
19292 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19293 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19294 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19295 (ia64_vectorize_vec_perm_const_ok): Merge into...
19296 (ia64_vectorize_vec_perm_const): ...this new function.
19297 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19298 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19299 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19300 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19301 * config/mips/mips.c (mips_expand_vec_perm_const)
19302 (mips_vectorize_vec_perm_const_ok): Merge into...
19303 (mips_vectorize_vec_perm_const): ...this new function.
19304 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19305 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19306 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19307 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19308 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19309 (rs6000_expand_vec_perm_const): Delete.
19310 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19312 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19313 (altivec_expand_vec_perm_const_le): Take each operand individually.
19314 Operate on constant selectors rather than rtxes.
19315 (altivec_expand_vec_perm_const): Likewise. Update call to
19316 altivec_expand_vec_perm_const_le.
19317 (rs6000_expand_vec_perm_const): Delete.
19318 (rs6000_vectorize_vec_perm_const_ok): Delete.
19319 (rs6000_vectorize_vec_perm_const): New function.
19320 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19321 an element count and rtx array.
19322 (rs6000_expand_extract_even): Update call accordingly.
19323 (rs6000_expand_interleave): Likewise.
19324 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19325 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19326 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19327 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19328 (rs6000_expand_vec_perm_const): Delete.
19329 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19330 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19331 (altivec_expand_vec_perm_const_le): Take each operand individually.
19332 Operate on constant selectors rather than rtxes.
19333 (altivec_expand_vec_perm_const): Likewise. Update call to
19334 altivec_expand_vec_perm_const_le.
19335 (rs6000_expand_vec_perm_const): Delete.
19336 (rs6000_vectorize_vec_perm_const_ok): Delete.
19337 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19338 reference to the SPE evmerge intructions.
19339 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19340 an element count and rtx array.
19341 (rs6000_expand_extract_even): Update call accordingly.
19342 (rs6000_expand_interleave): Likewise.
19343 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19344 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19346 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19350 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19351 vector mode and that that mode matches the mode of the data
19353 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19354 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19355 directly using expand_vec_perm_1 when forcing selectors into
19357 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19361 * optabs-query.h (can_vec_perm_p): Delete.
19362 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19363 * optabs-query.c (can_vec_perm_p): Split into...
19364 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19365 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19366 particular selector is valid.
19367 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19368 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19369 (vect_grouped_load_supported): Likewise.
19370 (vect_shift_permute_load_chain): Likewise.
19371 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19372 (vect_transform_slp_perm_load): Likewise.
19373 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19374 (vectorizable_bswap): Likewise.
19375 (vect_gen_perm_mask_checked): Likewise.
19376 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19377 implementations of variable permutation vectors into account
19378 when deciding which selector to use.
19379 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19380 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19381 with a false third argument.
19382 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19383 to test whether the constant selector is valid and can_vec_perm_var_p
19384 to test whether a variable selector is valid.
19386 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19388 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19389 * optabs-query.c (can_vec_perm_p): Likewise.
19390 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19391 instead of vec_perm_indices.
19392 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19393 (vect_gen_perm_mask_checked): Likewise,
19394 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19395 (vect_gen_perm_mask_checked): Likewise,
19397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19399 * optabs-query.h (qimode_for_vec_perm): Declare.
19400 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19401 (qimode_for_vec_perm): ...this new function.
19402 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19404 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19406 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19407 does not have a conditional at the top.
19409 2018-01-02 Richard Biener <rguenther@suse.de>
19411 * ipa-inline.c (big_speedup_p): Fix expression.
19413 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19416 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19419 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19423 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19424 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19425 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19426 cond_taken_branch_cost 3->4.
19428 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19430 PR tree-optimization/83581
19431 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19432 TODO_cleanup_cfg if any changes have been made.
19434 PR middle-end/83608
19435 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19436 convert_modes if target mode has the right side, but different mode
19439 PR middle-end/83609
19440 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19441 last argument when extracting from CONCAT. If either from_real or
19442 from_imag is NULL, use expansion through memory. If result is not
19443 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19444 the parts directly to inner mode, if even that fails, use expansion
19447 PR middle-end/83623
19448 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19449 check for bswap in mode rather than HImode and use that in expand_unop
19452 Copyright (C) 2018 Free Software Foundation, Inc.
19454 Copying and distribution of this file, with or without modification,
19455 are permitted in any medium without royalty provided the copyright
19456 notice and this notice are preserved.