1 2018-07-12 Richard Biener <rguenther@suse.de>
4 * dwarf2out.c (gen_block_die): Only output blocks when they have
5 at least one !DECL_IGNORED_P variable.
7 2018-07-12 Richard Biener <rguenther@suse.de>
10 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
11 Remove -mieee-fp handling.
13 2018-07-12 Richard Biener <rguenther@suse.de>
15 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
16 left-over from last patch.
18 2018-07-12 Jakub Jelinek <jakub@redhat.com>
20 PR tree-optimization/86492
21 * gimple-ssa-store-merging.c
22 (imm_store_chain_info::coalesce_immediate_stores): Call
23 check_no_overlap even for the merge_overlapping case. Formatting fix.
25 2018-07-12 Richard Biener <rguenther@suse.de>
28 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
29 move possibly trapping operations into the conditional.
31 2018-07-12 Richard Biener <rguenther@suse.de>
33 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
34 (vn_lookup_simplify_result): Remove recursion limit applied
36 (vn_nary_build_or_lookup_1): Adjust.
37 (try_to_simplify): Likewise.
38 * gimple-match-head.c (gimple_resimplify1): Instead apply one
40 (gimple_resimplify2): Likewise.
41 (gimple_resimplify3): Likewise.
42 (gimple_resimplify4): Likewise.
44 2018-07-11 Jakub Jelinek <jakub@redhat.com>
46 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
47 Use __mmask64 type instead of __mmask8 for __M argument.
48 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
49 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
51 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
52 __mmask16 for __M argument.
53 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
54 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
55 to __mmask16 instead of __mmask8.
56 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
57 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
58 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
59 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
60 instead of __mmask16 for __U argument.
61 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
62 __mmask16 instead of __mmask8 for __U argument.
63 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
65 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
67 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
69 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
71 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
73 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
75 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
77 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
79 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
81 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
83 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
84 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
85 return type as well as __M argument type and all casts from __mmask8
87 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
88 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
89 return type as well as __M argument type and all casts from __mmask8
91 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
92 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
93 return type as well as __M argument type and all casts from __mmask8
95 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
96 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
97 return type as well as __M argument type and all casts from __mmask8
99 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
100 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
103 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
105 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
106 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
109 2018-07-11 Paul Koning <ni1d@arrl.net>
111 * doc/md.texi (define_subst): Document how multiple occurrences of
112 the same argument in the replacement pattern are handled.
114 2018-07-11 Paul Koning <ni1d@arrl.net>
116 * doc/extend.texi (Common Variable Attributes): Move "mode" into
118 (Common Type Attributes): Add "mode" attribute.
120 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
122 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
123 stream DECL_ORIGINAL_TYPE.
124 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
125 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
126 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
127 Do not walk original type.
128 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
130 (write_ts_decl_non_common_tree_pointers): Do not stream
132 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
133 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
135 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
137 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
138 thread twice from the same starting edge.
140 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
142 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
143 * gimple.c (gimple_call_nonnull_result_p): ...here...
144 (gimple_call_nonnull_arg): ...and here.
145 * gimple.h (gimple_call_nonnull_result_p): New.
146 (gimple_call_nonnull_arg): New.
148 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
150 * config/arm/arm-cpus.in: Move information from fpu field of each
151 cpu definition to the isa field.
152 * config/arm/parsecpu.awk (fpu): Delete match rule.
153 (gen_comm_data): Don't add bits from the CPU's FPU entry.
155 2018-07-11 Richard Biener <rguenther@suse.de>
158 * dwarf2out.c (init_sections_and_labels): Use
159 output_asm_line_debug_info consistently.
160 (dwarf2out_early_finish): Likewise.
161 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
164 2018-07-11 Richard Biener <rguenther@suse.de>
166 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
167 Rework father_bb setting in a way to avoid propagating constants
168 multiple times on a loop body.
170 2018-07-10 Mark Wielaard <mark@klomp.org>
173 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
175 2018-07-10 Richard Biener <rguenther@suse.de>
177 * hash-map.h (hash_map::iterator::operator*): Return
178 references to key and value.
180 2018-07-10 Jakub Jelinek <jakub@redhat.com>
183 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
184 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
185 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
186 (gimplify_omp_for): For composite loops, move outer
187 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
188 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
189 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
190 TREE_LIST for both the original class iterator and the "last" helper
191 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
192 loop, remember has_decl_expr from outer composite loops for the
193 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
195 2018-07-09 Martin Sebor <msebor@redhat.com>
199 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
200 accessing implicitly initialized array elements.
201 * expr.c (string_constant): Handle string initializers of
202 character arrays within aggregates.
203 * gimple-fold.c (fold_array_ctor_reference): Add argument.
204 Store element offset. As a special case, handle zero size.
205 (fold_nonarray_ctor_reference): Same.
206 (fold_ctor_reference): Add argument. Store subobject offset.
207 * gimple-fold.h (fold_ctor_reference): Add argument.
209 2018-07-09 Paul Koning <ni1d@arrl.net>
211 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
212 (pdp11_insn_cost): New function.
213 (pdp11_md_asm_adjust): New function.
214 (TARGET_INVALID_WITHIN_DOLOOP): Define.
215 (pdp11_rtx_costs): Update to match machine better.
216 (output_addr_const_pdp11): Correct format mismatch warnings.
217 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
218 * config/pdp11/pdp11.md: General change to add base_cost and/or
219 length attributes for use by new pdp11_insn_cost function.
220 (MIN_BRANCH): Correct definition.
222 (doloop_end): Use standard pattern name for looping pattern.
223 (doloop_end_nocc): New.
224 (movsf): Add another constraint alternative.
225 (zero_extendqihi2): Add constraint alternatives for not in place
227 (zero_extendhisi2): Remove.
228 (shift patterns): Add CC handling variants.
232 (define_peephole2): New peephole to recognize mov that sets CC for
235 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
237 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
238 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
241 2018-07-09 Jakub Jelinek <jakub@redhat.com>
244 * real.c (real_nextafter): Return true if result is denormal.
246 2018-07-09 Martin Liska <mliska@suse.cz>
248 * common.opt: Add back wrongly removed attribute.
250 2018-07-09 Richard Biener <rguenther@suse.de>
253 * dwarf2out.c (gen_block_die): For an early generated DIE
254 always output high/low PC attributes.
256 2018-07-09 Tom de Vries <tdevries@suse.de>
258 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
259 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
260 onto VAR_DECL with abstract origin.
262 2018-07-07 Jim Wilson <jimw@sifive.com>
264 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
266 2018-07-07 Tom de Vries <tdevries@suse.de>
268 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
270 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
272 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
274 (combine_bound): Use wide-int overflow calculation instead of
276 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
278 * fold-const.c (int_const_binop_2): Same.
279 (extract_muldiv_1): Same.
280 (fold_div_compare): Same.
281 (fold_abs_const): Same.
283 * poly-int.h (add): Same.
287 * predict.c (predict_iv_comparison): Same.
288 * profile-count.c (slow_safe_scale_64bit): Same.
289 * simplify-rtx.c (simplify_const_binary_operation): Same.
290 * tree-chrec.c (tree_fold_binomial): Same.
291 * tree-data-ref.c (split_constant_offset_1): Same.
292 * tree-if-conv.c (idx_within_array_bound): Same.
293 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
294 * tree-ssa-phiopt.c (minmax_replacement): Same.
295 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
296 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
297 * vr-values.c (vr_values::adjust_range_with_scev): Same.
298 * wide-int.cc (wi::add_large): Same.
299 (wi::mul_internal): Same.
300 (wi::sub_large): Same.
301 (wi::divmod_internal): Same.
302 * wide-int.h: Change overflow type to overflow_type for neg, add,
303 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
304 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
305 mul_internal, divmod_internal.
306 (overflow_type): New enum.
307 (accumulate_overflow): New.
309 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
311 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
312 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
314 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
316 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
317 argument is checked for zero before entering loop, avoid checking again.
319 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
321 * gimplify.h (generic_expr_could_trap_p): Set as global function.
322 * gimplify.h (generic_expr_could_trap_p): Likwise.
323 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
325 2018-07-06 Jakub Jelinek <jakub@redhat.com>
327 PR tree-optimization/86401
328 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
329 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
330 (fold_bit_and_mask): ... here. New helper function for match.pd.
331 * fold-const.h (fold_bit_and_mask): Declare.
332 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
334 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
337 * target.def (translate_mode_attribute): New hook.
338 * targhooks.h (default_translate_mode_attribute): Declare.
339 * targhooks.c (default_translate_mode_attribute): New function.
340 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
341 * doc/tm.texi: Regenerate.
342 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
343 (rs6000_translate_mode_attribute): New function.
345 2018-07-06 Paul Koning <ni1d@arrl.net>
347 * doc/md.texi (define_split): Document DONE and FAIL.
348 (define_peephole2): Ditto.
350 2018-07-05 Jeff Law <law@redhat.com>
352 PR tree-optimization/86010
353 * tree-ssa-dse.c (compute_trims): More aggressively trim at
354 both the head and tail of mem* and str* calls.
356 2018-07-05 Jim Wilson <jimw@sifive.com>
358 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
360 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
362 * config/aarch64/aarch64-simd.md: correct flags text for
363 MIN_EXPR replacement.
365 2018-07-05 James Clarke <jrtc27@jrtc27.com>
367 * configure: Regenerated.
369 2018-07-05 Carl Love <cel@us.ibm.com>
371 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
372 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
373 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
374 VSX_BUILTIN_DOUBLEL_V4SF.
376 2018-07-05 Martin Sebor <msebor@redhat.com>
379 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
380 than its domain to compute its the upper bound of a char array.
382 2018-07-05 Nathan Sidwell <nathan@acm.org>
384 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
385 * doc/cpp.texi: Update comment.
386 * doc/tm.texi: Rebuilt.
387 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
388 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
389 * doc/extend.texi (Backwards Compatibility): Clarify it is system
390 headers affected by extern "C".
391 * system.h: Poison NO_IMPLICIT_EXTERN_C.
392 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
393 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
394 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
395 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
396 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
397 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
398 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
399 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
400 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
401 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
402 config/sparc/sp64-elf.h, config/spu/spu.h,
403 config/stormy16/stormy16.h, config/v850/v850.h,
404 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
405 define NO_IMPLICIT_EXTERN_C.
406 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
408 2018-07-05 Tamar Christina <tamar.christina@arm.com>
411 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
412 instead of GET_MODE_SIZE when comparing Units.
414 2018-07-05 Tamar Christina <tamar.christina@arm.com>
417 * rtlanal.c (set_noop_p): Constrain on mode change,
418 include hard-reg-set.h
420 2018-07-05 Tamar Christina <tamar.christina@arm.com>
422 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
424 2018-07-05 Jakub Jelinek <jakub@redhat.com>
427 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
430 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
432 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
434 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
437 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
439 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
441 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
444 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
446 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
448 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
451 2018-07-04 Martin Liska <mliska@suse.cz>
453 * tree-switch-conversion.c: Define
454 max_ratio_for_speed and max_ratio_for_size constants.
456 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
457 Martin Liska <mliska@suse.cz>
462 * common.opt: Rename align options with 'str_' prefix.
463 * common/config/i386/i386-common.c (set_malign_value): New
465 (ix86_handle_option): Use it to set -falign-* options/
466 * config/aarch64/aarch64-protos.h (struct tune_params): Change
467 type from int to string.
468 * config/aarch64/aarch64.c: Update default values from int
470 * config/alpha/alpha.c (alpha_override_options_after_change):
472 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
473 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
474 max skip conditionally.
475 * config/i386/freebsd.h (SUBALIGN_LOG): New.
476 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
477 max skip conditionally.
478 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
479 max skip conditionally.
480 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
481 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
482 max skip conditionally.
483 * config/i386/i386.c (struct ptt): Change type from int to
485 (ix86_default_align): Set default values.
486 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
487 max skip conditionally.
488 * config/i386/iamcu.h (SUBALIGN_LOG): New.
489 (ASM_OUTPUT_MAX_SKIP_ALIGN):
490 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
491 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
492 max skip conditionally.
493 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
494 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
495 * config/i386/x86-64.h (SUBALIGN_LOG): New.
496 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
497 max skip conditionally.
498 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
499 * config/ia64/ia64.c (ia64_option_override): Set default values
500 for alignment options.
501 * config/m68k/m68k.c: Handle new str_align_* options.
502 * config/mips/mips.c (mips_set_compression_mode): Change
504 (mips_option_override): Set default values for options.
505 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
507 * config/rs6000/rs6000.c (rs6000_option_override_internal):
509 * config/rx/rx.c (rx_option_override): Likewise.
510 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
511 (LABEL_ALIGN): Use align_labels_log.
512 (LOOP_ALIGN): Use align_loops_align.
513 * config/s390/s390.c (s390_asm_output_function_label): Use new
515 * config/sh/sh.c (sh_override_options_after_change):
516 Change type of constants.
517 * config/spu/spu.c (spu_sched_init): Likewise.
518 * config/sparc/sparc.c (sparc_option_override): Set default
520 * config/visium/visium.c (visium_option_override): Likewise.
521 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
522 emit p2align format with last argument if it's not needed.
523 * doc/invoke.texi: Document extended format of -falign-*.
524 * final.c: Use align_labels alignment.
525 * flags.h (struct target_flag_state): Change type to use
527 (struct align_flags_tuple): New.
528 (struct align_flags): Likewise.
529 (align_loops_log): Redefine macro to use new types.
530 (align_loops_max_skip): Redefine macro to use new types.
531 (align_jumps_log): Redefine macro to use new types.
532 (align_jumps_max_skip): Redefine macro to use new types.
533 (align_labels_log): Redefine macro to use new types.
534 (align_labels_max_skip): Redefine macro to use new types.
535 (align_functions_log): Redefine macro to use new types.
536 (align_loops): Redefine macro to use new types.
537 (align_jumps): Redefine macro to use new types.
538 (align_labels): Redefine macro to use new types.
539 (align_functions): Redefine macro to use new types.
540 (align_functions_max_skip): Redefine macro to use new types.
541 (align_loops_value): New macro.
542 (align_jumps_value): New macro.
543 (align_labels_value): New macro.
544 (align_functions_value): New macro.
545 * function.c (invoke_set_current_function_hook): Propagate
546 alignment values from flags to global variables default in
548 * ipa-icf.c (sem_function::equals_wpa): Use
549 cl_optimization_option_eq instead of memcmp.
550 * lto-streamer.h (cl_optimization_stream_out): Support streaming
552 (cl_optimization_stream_in): Likewise.
553 * optc-save-gen.awk: Support strings in cl_optimization.
554 * opth-gen.awk: Likewise.
555 * opts.c (finish_options): Remove error checking of invalid
557 (MAX_CODE_ALIGN): Remove.
558 (MAX_CODE_ALIGN_VALUE): Likewise.
559 (parse_and_check_align_values): New function.
560 (check_alignment_argument): Likewise.
561 (common_handle_option): Use check_alignment_argument.
562 * opts.h (parse_and_check_align_values): Declare.
563 * toplev.c (init_alignments): Remove.
564 (read_log_maxskip): New.
565 (parse_N_M): Likewise.
566 (parse_alignment_opts): Likewise.
567 (backend_init_target): Remove usage of init_alignments.
568 * toplev.h (parse_alignment_opts): Declare.
569 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
571 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
572 * tree.c (cl_option_hasher::equal): New.
573 * varasm.c: Use new global macros.
575 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
577 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
578 Use a simpler align directive also if MAXSKIP = ALIGN-1.
579 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
580 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
581 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
582 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
583 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
584 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
585 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
586 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
587 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
588 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
589 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
591 2018-07-04 Martin Liska <mliska@suse.cz>
592 Jonathan Wakely <jwakely@redhat.com>
594 * coverage.c: Use correct type.
595 * doc/invoke.texi: Language correction.
597 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
600 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
601 ENDBRANCH for non-tail call which may return via indirect branch.
602 * doc/extend.texi: Document indirect_return attribute.
604 2018-07-03 Martin Sebor <msebor@redhat.com>
606 PR tree-optimization/86274
607 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
609 (format_floating): Correct handling of infinities and NaNs.
611 2018-07-03 Martin Sebor <msebor@redhat.com>
613 * print-tree.c (print_real_cst): New function.
614 (print_node_brief): Call it.
617 2018-07-03 Jeff Law <law@redhat.com>
619 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
620 into a single pattern.
622 * config/h8300/h8300.md (ors code_iterator): New.
623 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
624 a single pattern and single splitter.
625 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
626 (iorqi3_1, xorqi3_1): Likewise.
627 (iorqi3, xorqi3 expanders): Similarly.
629 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
630 (movmd_internal) into a single pattern using the P mode iterator.
631 (movmd splitters): Similarly.
632 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
633 (movsd splitters): Similarly.
635 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
636 ADDB, ADDW and ADDL into a single ADD attribute which selects the
637 right table based on the size of the operand.
638 * config/h8300/h8300.md (length_table): Corresponding changes. All
639 references to "addb", "addw" and "addl" changed to "add".
640 (btst patterns): Merge two variants into a single pattern.
641 (tstqi, tsthi): Likewise.
642 (addhi3_incdec, addsi3_incdec): Likewise.
643 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
644 (mulhi3, mulsi3): Likewise.
645 (udivhi3, udivsi3): Likewise.
646 (divhi3, divsi3): Likewise.
647 (andorqi3, andorhi3, andorsi3): Likewise.
649 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
652 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
653 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
655 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
657 PR tree-optimization/85694
658 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
659 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
661 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
662 (<u>avg<mode>3_ceil): New patterns.
664 2018-07-03 David Malcolm <dmalcolm@redhat.com>
666 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
667 scan-tree-dump directive.
668 * gcc.dg/vect/slp-perm-2.c: Likewise.
669 * gcc.dg/vect/slp-perm-3.c: Likewise.
670 * gcc.dg/vect/slp-perm-5.c: Likewise.
671 * gcc.dg/vect/slp-perm-6.c: Likewise.
672 * gcc.dg/vect/slp-perm-7.c: Likewise.
673 * gcc.dg/vect/slp-perm-8.c: Likewise.
675 2018-07-03 Marek Polacek <polacek@redhat.com>
678 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
680 2018-07-03 Richard Biener <rguenther@suse.de>
683 * tree-ssa-structalias.c (find_func_clobbers): Properly
684 handle indirect calls.
686 2018-07-03 Jeff Law <law@redhat.com>
688 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
689 (shifts): New code iterator.
690 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
691 expander. Fix HImode handling on H8/SX.
692 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
693 (subqi3, subhi3, subsi3 expanders): Likewise.
694 (andqi3, andhi3, andsi3 expanders): Likewise.
695 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
696 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
697 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
698 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
699 (zero_extendqihi2, zero_extendqisi2): Likewise.
700 (extendqihi2, extendqisi2): Likewise.
701 (rotlqi3, rotlhi3, rotlsi3): Likewise.
702 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
703 (rotlqi3_1, rotlhi3_1): Likewise.
704 (logicalhi3_sn, logicalsi3_sn): Likewise.
705 (logicalhi3, logicalsi3): Likewise.
707 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
709 * tree-vect-patterns.c (vect_recog_rotate_pattern)
710 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
711 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
712 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
713 type to append_pattern_def_seq instead of creating a stmt_vec_info
715 (build_mask_conversion): Likewise. Remove vinfo argument.
716 (vect_add_conversion_to_patterm): Likewise, renaming to...
717 (vect_add_conversion_to_pattern): ...this.
718 (vect_recog_mask_conversion_pattern): Update call to
719 build_mask_conversion. Pass the vector type to
720 append_pattern_def_seq here too.
721 (vect_recog_gather_scatter_pattern): Update call to
722 vect_add_conversion_to_pattern.
724 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
726 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
727 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
728 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
729 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
730 STMT_VINFO_PATTERN_DEF_SEQ to null here.
731 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
732 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
733 append_pattern_def_seq instead of new_pattern_def_seq.
734 (vect_recog_divmod_pattern): Do both of the above.
735 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
738 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
740 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
741 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
742 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
743 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
744 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
745 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
746 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
747 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
748 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
749 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
750 parameter with a single stmt_vec_info.
751 (vect_recog_func_ptr): Likewise.
752 (vect_recog_gather_scatter_pattern): Likewise, folding in...
753 (vect_try_gather_scatter_pattern): ...this.
754 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
755 the stmt_vec_info of the statement to be matched. Don't clear
756 STMT_VINFO_RELATED_STMT.
757 (vect_pattern_recog): Update call accordingly.
759 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
761 PR tree-optimization/85694
762 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
763 (uavgM3_ceil): Document new optabs.
764 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
765 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
767 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
768 (savg_ceil_optab): New optabs.
769 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
770 (vect_vect_recog_func_ptrs): Add it.
771 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
772 constant directly from the associated lhs.
774 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
776 * tree-vect-patterns.c (vect_split_statement): New function.
777 (vect_convert_input): Use it to try to split an existing cast.
779 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
781 * poly-int.h (print_hex): New function.
782 * dumpfile.h (dump_dec, dump_hex): Declare.
783 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
784 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
785 min_input_precision, operation_precision and operation_sign.
786 * tree-vect-patterns.c (vect_get_range_info): New function.
787 (vect_same_loop_or_bb_p, vect_single_imm_use)
788 (vect_operation_fits_smaller_type): Delete.
789 (vect_look_through_possible_promotion): Add an optional
790 single_use_p parameter.
791 (vect_recog_over_widening_pattern): Rewrite to use new
792 stmt_vec_info infomration. Handle one operation at a time.
793 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
794 (vect_truncatable_operation_p, vect_set_operation_type)
795 (vect_set_min_input_precision): New functions.
796 (vect_determine_min_output_precision_1): Likewise.
797 (vect_determine_min_output_precision): Likewise.
798 (vect_determine_precisions_from_range): Likewise.
799 (vect_determine_precisions_from_users): Likewise.
800 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
801 (vect_vect_recog_func_ptrs): Put over_widening first.
803 (vect_pattern_recog): Call vect_determine_precisions.
805 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
807 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
808 statements that have been replaced by further pattern statements.
809 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
811 2018-07-03 Richard Biener <rguenther@suse.de>
813 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
814 always set *dt. Dump vectype in vectype overload.
815 * dumpfile.h (dump_gimple_expr): New function.
816 (dump_gimple_expr_loc): Likewise.
817 * dumpfile.c (dump_gimple_expr): New function.
818 (dump_gimple_expr_loc): Likewise.
820 2018-07-02 Jeff Law <law@redhat.com>
822 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
823 the H8/300, H8/300H and H8/S variants into a single pattern.
824 (movhi_h8300, movqi_h8300hs): Similarly.
825 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
826 (QHI mode iterator): New.
828 * config/h8300/h8300.md: Remove trailing whitespace.
830 2018-07-02 Jim Wilson <jimw@sifive.com>
832 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
833 instead of emit_insn for interrupt returns.
834 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
835 (riscv_sret, riscv_uret): Likewise.
837 2018-07-02 David Malcolm <dmalcolm@redhat.com>
839 * pretty-print.c (selftest::test_pp_format): Move save and restore
840 of quotes to class auto_fix_quotes, and add an instance.
841 * selftest.c: Include "intl.h".
842 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
843 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
844 * selftest.h (selftest::auto_fix_quotes): New class.
846 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
848 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
849 (aarch64_sve_prepare_conditional_op): Remove.
850 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
851 Allow aarch64_simd_reg_or_zero as select operand; remove
852 the aarch64_sve_prepare_conditional_op call.
853 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
854 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
855 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
856 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
857 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
858 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
859 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
860 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
861 and a splitters to match all of the *_any patterns.
862 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
864 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
865 (SVE_COND_FP_BINARY_REV): Remove.
866 (sve_int_op_rev, sve_fp_op_rev): New.
867 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
868 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
869 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
870 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
871 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
872 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
873 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
874 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
875 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
877 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
878 Remove match_dup 1 from the inner unspec.
879 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
881 * config/aarch64/aarch64.md (movprfx): New attr.
882 (length): Default movprfx to 8.
883 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
884 (*madd<SVE_I>, *msub<SVE_I): Likewise.
885 (*<su>mul<SVE_I>3_highpart): Likewise.
886 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
887 (*v<ASHIFT><SVE_I>3): Likewise.
888 (*<su><MAXMIN><SVE_I>3): Likewise.
889 (*<su><MAXMIN><SVE_F>3): Likewise.
890 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
891 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
892 (*div<SVE_F>4): Likewise.
894 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
896 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
899 2018-07-02 Richard Biener <rguenther@suse.de>
901 PR tree-optimization/86363
902 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
903 memset argument refers to a non-variable address.
905 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
907 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
908 {PLUS,MINUS}_EXPR code to...
909 (adjust_symbolic_bound): ...here,
910 (combine_bound): ...here,
911 (set_value_range_with_overflow): ...and here.
913 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
915 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
917 (extract_range_from_abs_expr): ...here.
919 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
921 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
922 -fno-omit-frame-pointer when not optimizing.
924 2018-07-02 Martin Liska <mliska@suse.cz>
927 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
928 (propagate_nothrow): Likewise.
930 2018-07-02 Martin Liska <mliska@suse.cz>
933 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
935 2018-07-02 David Malcolm <dmalcolm@redhat.com>
937 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
938 function in r262149, changing "loc" param from source_location to
939 const dump_location_t &.
940 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
941 declaration, as above.
943 2018-07-01 Paul Koning <ni1d@arrl.net>
945 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
946 -munit-asm, -mgnu-asm, -mdec-asm.
947 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
948 (pdp11_output_labelref): New.
949 (pdp11_output_def): New.
950 (pdp11_output_addr_vec_elt): New.
951 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
952 %# and %@ format codes.
953 (pdp11_option_override): New.
954 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
955 (pdp11_output_ident): New.
956 (pdp11_asm_named_section): New.
957 (pdp11_asm_init_sections): New.
958 (pdp11_file_start): New.
959 (pdp11_file_end): New.
960 (output_ascii): Use .ascii/.asciz for -mdec-asm.
961 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
962 %o, like %c but octal.
963 (pdp11_option_override): New.
964 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
966 (DATA_SECTION_ASM_OP): Ditto.
967 (READONLY_DATA_SECTION_ASM_OP): New.
968 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
969 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
970 (ASM_OUTPUT_LABELREF): Ditto.
971 (ASM_OUTPUT_DEF): Ditto.
972 (ASM_OUTPUT_EXTERNAL): New.
973 (ASM_OUTPUT_SOURCE_FILENAME): New.
974 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
975 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
976 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
977 %# and %@ format codes.
978 * config/pdp11/pdp11.opt (mgnu-asm): New.
979 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
980 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
981 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
983 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
985 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
986 dereferencing path[] beyond its length.
988 (debug_all_paths): New.
989 (rewire_first_differing_edge): New.
990 (adjust_paths_after_duplication): New.
991 (duplicate_thread_path): Call adjust_paths_after_duplication.
993 (thread_through_all_blocks): Add new argument to
994 duplicate_thread_path.
996 2018-06-30 Jim Wilson <jimw@sifive.com>
998 * config/riscv/predicates.md (p2m1_shift_operand): New.
999 (high_mask_shift_operand): New.
1000 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1001 pattern using p2m1_shift_operand.
1002 (lshsi3_zero_extend_3+2): New combiner pattern using
1003 high_mask_shift_operand.
1005 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1007 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1009 (vect_recog_rotate_pattern): ...here.
1010 (vect_convert_input): Try to insert casts of invariants in the
1012 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1013 preheader to be empty.
1015 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1017 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1018 vector type. If given, install it in the new statement's
1020 (vect_element_precision): New function.
1021 (vect_unpromoted_value): New struct.
1022 (vect_unpromoted_value::vect_unpromoted_value): New function.
1023 (vect_unpromoted_value::set_op): Likewise.
1024 (vect_look_through_possible_promotion): Likewise.
1025 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1026 (vect_widened_op_tree, vect_convert_input): Likewise.
1027 (vect_convert_inputs, vect_convert_output): Likewise.
1028 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1029 to handle the optional cast of the multiplication result and
1030 vect_widened_op_tree to detect the widened multiplication itself.
1031 Do not require the input and output of promotion casts to have
1032 the same sign, but base the signedness of the operation on the
1033 input rather than the result. If the pattern includes two
1034 promotions, check that those promotions have the same sign.
1035 Do not restrict the MULT_EXPR handling to a double-width result;
1036 handle quadruple-width results and wider. Use vect_convert_inputs
1037 to convert the inputs to the common type.
1038 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1039 to handle the optional cast of the ABS result. Also allow a sign
1040 change or a sign extension between the ABS and MINUS.
1041 Use vect_widened_op_tree to detect the widened subtraction and use
1042 vect_convert_inputs to convert the inputs to the common type.
1043 (vect_handle_widen_op_by_const): Delete.
1044 (vect_recog_widen_op_pattern): New function.
1045 (vect_recog_widen_mult_pattern): Use it.
1046 (vect_recog_widen_shift_pattern): Likewise.
1047 (vect_recog_widen_sum_pattern): Use
1048 vect_look_through_possible_promotion to handle the promoted
1051 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1053 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1054 the containing gimple_seq *.
1055 * gimple-iterator.h (gsi_for_stmt): Declare it.
1056 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1057 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1058 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1059 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1060 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1062 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1064 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1065 statement being replaced is part of an existing pattern
1066 definition sequence, inserting the new pattern statements before
1068 (vect_pattern_recog_1): Don't return a bool. If the statement
1069 is already part of a pattern, instead apply pattern matching
1070 to the pattern definition statements. Don't clear the
1071 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1072 (vect_pattern_recog): Don't break after the first match;
1073 continue processing the pattern definition statements instead.
1074 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1076 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1078 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1079 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1080 (vect_recog_widen_sum_pattern): Use it.
1082 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1084 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1085 phi is not a pattern statement and has not been replaced by
1086 a pattern statement.
1087 * tree-vect-patterns.c (type_conversion_p): Don't check
1088 STMT_VINFO_IN_PATTERN_P.
1089 (vect_recog_vector_vector_shift_pattern): Likewise.
1090 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1091 the pattern statement rather than the original statement; check
1092 directly for a WIDEN_MULT_EXPR here.
1093 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1094 vect_is_simple_use to return the pattern statement rather
1095 than the original statement; use is_pattern_stmt_p to check
1096 for such a pattern statement.
1097 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1098 to return the pattern statement rather than the original statement;
1099 don't do the same transformation here.
1100 (vect_is_simple_use): If the defining statement has been replaced
1101 by a pattern statement, return the pattern statement instead.
1102 Remove the corresponding (local) transformation from the vectype
1105 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1107 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1108 end and default to null.
1109 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1110 (vectorizable_reduction): Update calls accordingly, dropping the
1111 gimple ** argument if the passed-back statement isn't needed.
1112 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1113 (vect_recog_rotate_pattern): Likewise.
1114 (vect_recog_mask_conversion_pattern): Likewise.
1115 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1116 (vect_mask_constant_operand_p): Likewise.
1117 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1118 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1119 (get_group_load_store_type, get_load_store_type): Likewise.
1120 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1121 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1122 (vectorizable_conversion, vectorizable_assignment): Likewise.
1123 (vectorizable_shift, vectorizable_operation): Likewise.
1124 (vectorizable_store, vect_is_simple_cond): Likewise.
1125 (vectorizable_condition, vectorizable_comparison): Likewise.
1126 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1127 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1128 and move it to the end. Cope with null def_stmt_outs.
1130 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1132 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1134 2018-06-29 Jeff Law <law@redhat.com>
1136 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1137 displacements for TARGET_V850E2V3 and newer.
1138 (TARGET_LRA_P): Remove. Defaults to LRA now.
1139 * config/v850/v850.md (sign23byte_load): Remove.
1140 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1141 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1143 2018-06-29 Martin Liska <mliska@suse.cz>
1146 * coverage.c (coverage_init): Mangle full path name.
1147 * doc/invoke.texi: Document the change.
1148 * gcov-io.c (mangle_path): New.
1149 * gcov-io.h (mangle_path): Likewise.
1150 * gcov.c (mangle_name): Use mangle_path for path mangling.
1152 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1154 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1155 if starting source register is not even.
1157 2018-06-29 Martin Liska <mliska@suse.cz>
1159 PR tree-optimization/86263
1160 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1161 Make edge redirection.
1163 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1165 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1166 (dump_scope_depth): New variable.
1167 (get_dump_scope_depth): New function.
1168 (dump_begin_scope): New function.
1169 (dump_end_scope): New function.
1170 * dumpfile.h (get_dump_scope_depth): New declaration.
1171 (dump_begin_scope): New declaration.
1172 (dump_end_scope): New declaration.
1173 (class auto_dump_scope): New class.
1174 (AUTO_DUMP_SCOPE): New macro.
1175 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1178 2018-06-29 Richard Biener <rguenther@suse.de>
1180 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1181 compute_all_dependences succeeds.
1182 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1183 exceed --param loop-max-datarefs-for-datadeps.
1185 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1187 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1189 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1192 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1193 alternative 0 in preferred_for_speed attribute.
1195 2018-06-28 Paul Koning <ni1d@arrl.net>
1197 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1198 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1199 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1200 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1201 for shift insn patterns.
1203 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1205 * cgraph.c (cgraph_node::get_body): Replace assignments to
1206 "dump_file" with calls to set_dump_file.
1207 * dumpfile.c (alt_dump_file): Make static, and group with...
1208 (alt_flags): ...this definition.
1209 (dumps_are_enabled): New variable.
1210 (refresh_dumps_are_enabled): New function.
1211 (set_dump_file): New function.
1212 (set_alt_dump_file): New function.
1213 (gcc::dump_manager::dump_start): Replace assignments to
1214 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1216 (gcc::dump_manager::dump_finish): Likewise.
1217 * dumpfile.h (alt_dump_file): Delete decl.
1218 (dumps_are_enabled): New variable decl.
1219 (set_dump_file): New function decl.
1220 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1222 * tree-nested.c (lower_nested_functions): Replace assignments to
1223 "dump_file" with calls to set_dump_file.
1225 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1227 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1228 goto_locus of each outgoing edge of each basic block.
1230 2018-06-28 Richard Biener <rguenther@suse.de>
1232 * dwarf2out.c (decl_scope_table): Remove.
1233 (push_decl_scope): Likewise.
1234 (pop_decl_scope): Likewise.
1235 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1236 (gen_struct_or_union_type_die): Likewise.
1237 (gen_tagged_type_die): Likewise.
1238 (dwarf2out_init): Do not initialize decl_scope_table.
1239 (dwarf2out_c_finalize): Do not free it.
1241 2018-06-28 Richard Biener <rguenther@suse.de>
1243 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1244 deciding whether to not re-use a DIE.
1246 2018-06-28 Richard Biener <rguenther@suse.de>
1248 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1249 DW_AT_abstract_origin attribute.
1251 2018-06-28 Martin Liska <mliska@suse.cz>
1253 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1254 Use newly introduced constants.
1255 * tree-switch-conversion.h (struct jump_table_cluster):
1256 Define max_ratio_for_size and max_ratio_for_speed.
1258 2018-06-28 Martin Liska <mliska@suse.cz>
1260 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1261 Add new checking assert to catch invalid state.
1262 (jump_table_cluster::can_be_handled): Handle single case
1264 (jump_table_cluster::is_beneficial): Bail out for such case.
1265 (bit_test_cluster::find_bit_tests):
1266 Add new checking assert to catch invalid state.
1267 (bit_test_cluster::can_be_handled): Handle single case
1269 (bit_test_cluster::is_beneficial): Bail out for such case.
1270 (switch_decision_tree::analyze_switch_statement):
1273 2018-06-28 Martin Liska <mliska@suse.cz>
1275 * common.opt: Introduce -completion option.
1276 * gcc.c (driver_handle_option): Handle it.
1277 (driver::main): Print completions if completion
1279 * opt-suggestions.c (option_proposer::get_completions):
1281 (option_proposer::suggest_completion): Likewise.
1282 (option_proposer::find_param_completions): Likewise.
1283 (verify_autocompletions): Likewise.
1284 (test_completion_valid_options): Likewise.
1285 (test_completion_valid_params): Likewise.
1286 (in_completion_p): Likewise.
1287 (empty_completion_p): Likewise.
1288 (test_completion_partial_match): Likewise.
1289 (test_completion_garbage): Likewise.
1290 (opt_proposer_c_tests): Likewise.
1291 * opt-suggestions.h: Declare new functions.
1292 * opts.c (common_handle_option): Handle OPT__completion_.
1293 * selftest-run-tests.c (selftest::run_tests): Add
1294 opt_proposer_c_tests.
1295 * selftest.c (assert_str_startswith): New.
1296 * selftest.h (assert_str_startswith): Likewise.
1297 (opt_proposer_c_tests): New.
1298 (ASSERT_STR_STARTSWITH): Likewise.
1300 2018-06-28 Martin Liska <mliska@suse.cz>
1302 * Makefile.in: Add opt-suggestions.o.
1303 * gcc-main.c: Include opt-suggestions.h.
1304 * gcc.c (driver::driver): Likewise.
1305 (driver::~driver): Remove m_option_suggestions.
1306 (driver::build_option_suggestions): Moved to option_proposer.
1307 (driver::suggest_option): Likewise.
1308 (driver::handle_unrecognized_options): Use option_proposer.
1309 * gcc.h (class driver): Add new memver m_option_proposer.
1310 * opt-suggestions.c: New file.
1311 * opt-suggestions.h: New file.
1313 2018-06-28 Martin Liska <mliska@suse.cz>
1315 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1316 (auto_string_vec::~auto_string_vec): Likewise.
1318 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1320 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1321 prevent_decl_creation_for_types fields up and add reset_location field.
1322 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1323 statement if id->reset_location is true.
1324 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1325 id->reset_location is true.
1326 (copy_phis_for_bb): Force input_location on the arguments if
1327 id->reset_location is true.
1328 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1329 is set on the function to be inlined.
1331 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1333 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1335 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1337 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1338 registers for Pmode.
1339 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1340 hard registers for the clobbered pseudo.
1342 2018-06-27 Paul Koning <ni1d@arrl.net>
1344 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1345 mutually exclusive options.
1346 * config/pdp11/constraints.md (h): New constraint.
1347 (O): Update definition to match shift code generation.
1348 (D): New constraint.
1349 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1351 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1353 (output_jump): Change arguments.
1354 (pdp11_fixed_cc_regs): New function.
1355 (pdp11_cc_mode): Ditto.
1356 (pdp11_expand_shift): Ditto.
1357 (pdp11_assemble_shift): Ditto.
1358 (pdp11_small_shift): Ditto.
1359 (pdp11_branch_cost): Remove.
1360 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1362 (pdp11_register_move_cost): Update for CC registers.
1363 (pdp11_rtx_costs): Add case for LSHIFTRT.
1364 (pdp11_output_jump): Add CCNZ mode conditional branches.
1365 (notice_update_cc_on_set): Remove.
1366 (pdp11_cc_mode): New function.
1367 (simple_memory_operand): Correct pre/post decrement case.
1368 (no_side_effect_operand): New function.
1369 (pdp11_regno_reg_class): Add CC_REGS class.
1370 (pdp11_fixed_cc_regs): New function.
1371 (pdp11_small_shift): New function.
1372 (pdp11_expand_shift): New function to expand shift insns.
1373 (pdp11_assemble_shift): New function to output shifts.
1374 (pdp11_branch_cost): Remove.
1375 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1376 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1377 (WCHAR_TYPE): Ditto.
1378 (PTRDIFF_TYPE): Ditto.
1379 (ADJUST_INSN_LENGTH): New macro.
1380 (FIXED_REGISTERS): Add CC registers.
1381 (CALL_USED_REGISTERS): Ditto.
1383 (REG_CLASS_NAMES): Ditto.
1384 (REG_CLASS_CONTENTS): Ditto.
1385 (SELECT_CC_MODE): Use new function.
1386 (TARGET_FLAGS_REGNUM): New macro.
1387 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1388 (cc0_reg_rtx): Remove.
1389 (CC_STATUS_MDEP): Remove.
1390 (CC_STATUS_MDEFP_INIT): Remove.
1391 (CC_IN_FPU): Remove.
1392 (NOTICE_UPDATE_CC): Remove.
1393 (REGISTER_NAMES): Add CC registers.
1394 (BRANCH_COST): Change to constant 1.
1395 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1397 * config/pdp11/pdp11.opt (mbcopy): Remove.
1398 (mbcopy-builtin): Remove.
1399 (mbranch-cheap): Remove.
1400 (mbranch-expensive): Remove.
1401 * config/pdp11/predicates.md (expand_shift_operand): Update to
1402 match shift code generation.
1403 (ccnz_operator): New predicate.
1404 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1405 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1406 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1408 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1409 description of O constraint.
1411 2018-06-27 Jeff Law <law@redhat.com>
1412 Austin Law <austinklaw@gmail.com>
1414 * config/v850/v850.md (addsi3_set_flags): New pattern.
1415 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1416 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1417 (zero_extendhisi2_v850_set_flags): Likewise.
1418 (zero_extendqisi2_v850_set_flags): Likewise.
1419 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1420 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1421 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1423 * config/v850/v850-protos.h (notice_update_cc): Remove.
1424 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1425 (v850_print_operand): Handle 'D' and "d".
1426 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1427 Add handling of arithmetic/logical operations compared against zero.
1428 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1429 Do not look at v850_compare_op, instead get mode from last argument.
1430 (v850_gen_compare): Remove
1431 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1432 after reload for prologue insns.
1433 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1435 (construct_save_jarl): Likewise.
1436 (TARGET_FLAGS_REGNUM): Define.
1437 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1438 (NOTICE_UPDATE_CC): Remove.
1439 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1440 than cc0. Conditionalize on reload_completed.
1441 (cmpsi_insn, setfcc_insn): Likewise.
1442 (tst1 splitter): Turn into define_and_split which sets the flags
1444 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1445 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1446 (cstoresf4, cstoredf4): Clobber the flags.
1447 (cmpsi, cmpsf, cmpdf): Remove expanders.
1448 (setf_insn): Remove pattern.
1449 (addsi3): Turn into define_and_split which clobbers the flags after
1450 reload and a suitable pattern (addsi3_clobber_flags) for use after
1452 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1453 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1454 (ashrsi3, ashrsi3_v850e2): Likewise.
1455 (bins): Clobber the flags.
1456 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1457 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1458 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1459 (call_value_internal_short, call_value_internal_long): Likewise.
1460 (callt_save_interrupt, callt_return_interrupt): Likewise.
1461 (save_interrupt, return_interrupt): Likewise.
1462 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1463 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1464 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1465 (All FP comparisons): Only allow after reload has completed.
1467 (divh, divhu): Tweak output template.
1468 (branch_z_normal, branch_z_invert): Remove
1469 (branch_nz_normal, branch_nz_invert): Likewise.
1470 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1472 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1473 * config/v850/v850.c (notice_update_cc): Remove.
1474 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1475 (CC_NO_CARRY): Likewise.
1476 (NOTICE_UPDATE_CC): Define to nothing.
1477 * config/v850/v850.md: Remove block comment on cc0 handling
1478 Remove "cc" attribute from all patterns. Remove cc_status handling
1479 from all patterns. Minor formatting fixes.
1481 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1483 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1484 (cortex-a76.cortex-a55): Likewise.
1485 * config/aarch64/aarch64-tune.md: Regenerate.
1486 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1487 cortex-a76.cortex-a55.
1489 2018-06-27 Jeff Law <law@redhat.com>
1491 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1492 (MULTILIB_DIRNAMES): Similarly.
1494 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1496 * gimple.h (gimple_return_retbnd): Delete.
1497 (gimple_return_set_retbnd): Likewise.
1498 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1499 gimple_return_set_retbnd.
1500 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1501 gimple_return_retbnd and adjust.
1502 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1503 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1504 Explicitly return NULL in a couple more cases. Move assertion
1505 on debug statements and remove unreachable code.
1506 (reset_debug_binding): Do not test id->retbnd.
1507 (expand_call_inline): Do not set it.
1509 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1511 * configure.ac: Add --disable-gcov option.
1512 * configure: Regenerate.
1513 * Makefile.in: Honour @enable_gcov@.
1514 * doc/install.texi: Document --disable-gcov.
1516 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1518 * config/arm/arm-cpus.in (cortex-a76): New entry.
1519 (cortex-a76.cortex-a55): Likewise.
1520 * config/arm/arm-tables.opt: Regenerate.
1521 * config/arm/arm-tune.md: Likewise.
1522 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1523 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1524 cortex-a76.cortex-a55.
1526 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1529 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1531 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1533 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1535 (EPILOGUE_USES): Likewise.
1537 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1539 * tree-inline.c (remap_location): New function extracted from...
1540 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1541 (copy_phis_for_bb): ...here. Call remap_location.
1542 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1544 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1546 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1547 unaligned vsx for 16B memset.
1549 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1552 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1553 ieee128_float_type_node to long_double_type_node unless
1554 TARGET_LONG_DOUBLE_128 is set.
1556 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1558 * cfgloop.c (get_loop_location): Convert return type from
1559 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1560 by implicit construction from rtx_insn *, and using
1561 dump_user_location_t::from_function_decl for the fallback case.
1562 * cfgloop.h (get_loop_location): Convert return type from
1563 location_t to dump_user_location_t.
1564 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1565 dump_printf_loc to pass in a dump_location_t rather than a
1566 location_t, via the gimple stmt.
1567 * coverage.c (get_coverage_counts): Update calls to
1568 dump_printf_loc to pass in dump_location_t rather than a
1570 * doc/optinfo.texi (Dump types): Convert example of
1571 dump_printf_loc from taking "locus" to taking "insn". Update
1572 description of the "_loc" calls to cover dump_location_t.
1573 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1575 (dump_user_location_t::dump_user_location_t): New constructors,
1576 from gimple *stmt and rtx_insn *.
1577 (dump_user_location_t::from_function_decl): New function.
1578 (dump_loc): Make static.
1579 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1580 const dump_location_t &.
1581 (dump_generic_expr_loc): Delete.
1582 (dump_printf_loc): Convert param "loc" from location_t to
1583 const dump_location_t &.
1584 (selftest::test_impl_location): New function.
1585 (selftest::dumpfile_c_tests): New function.
1586 * dumpfile.h: Include "profile-count.h".
1587 (class dump_user_location_t): New class.
1588 (struct dump_impl_location_t): New struct.
1589 (class dump_location_t): New class.
1590 (dump_printf_loc): Convert 2nd param from source_location to
1591 const dump_location_t &.
1592 (dump_generic_expr_loc): Delete.
1593 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1594 const dump_location_t &.
1595 * gimple-fold.c (fold_gimple_assign): Update call to
1596 dump_printf_loc to pass in a dump_location_t rather than a
1597 location_t, via the gimple stmt.
1598 (gimple_fold_call): Likewise.
1599 * gimple-loop-interchange.cc
1600 (loop_cand::analyze_iloop_reduction_var): Update for change to
1601 check_reduction_path.
1602 (tree_loop_interchange::interchange): Update for change to
1604 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1605 change in return-type of find_loop_location.
1606 (graphite_regenerate_ast_isl): Likewise.
1607 * graphite-optimize-isl.c (optimize_isl): Likewise.
1608 * graphite.c (graphite_transform_loops): Likewise.
1609 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1610 pass in a dump_location_t rather than a location_t, via the
1612 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1613 * ipa.c (walk_polymorphic_call_targets): Likewise.
1614 * loop-unroll.c (report_unroll): Convert "locus" param from
1615 location_t to dump_location_t.
1616 (decide_unrolling): Update for change to get_loop_location's
1618 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1619 location_t to dump_user_location_t.
1620 (grid_find_single_omp_among_assignments_1): Updates calls to
1621 dump_printf_loc to pass in a dump_location_t rather than a
1622 location_t, via the gimple stmt.
1623 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1624 location_t to dump_location_t. Updates calls to dump_printf_loc
1625 to pass in a dump_location_t rather than a location_t, via the
1627 (grid_inner_loop_gridifiable_p): Likewise.
1628 (grid_dist_follows_simple_pattern): Likewise.
1629 (grid_gfor_follows_tiling_pattern): Likewise.
1630 (grid_target_follows_gridifiable_pattern): Likewise.
1631 (grid_attempt_target_gridification): Convert initialization
1632 of local "grid" from memset to zero-initialization; FIXME: does
1633 this require C++11? Update call to dump_printf_loc to pass in a
1634 optinfo_location rather than a location_t, via the gimple stmt.
1635 * profile.c (read_profile_edge_counts): Updates call to
1636 dump_printf_loc to pass in a dump_location_t rather than a
1638 (compute_branch_probabilities): Likewise.
1639 * selftest-run-tests.c (selftest::run_tests): Call
1641 * selftest.h (dumpfile_c_tests): New decl.
1642 * tree-loop-distribution.c (pass_loop_distribution::execute):
1643 Update for change in return type of find_loop_location.
1644 * tree-parloops.c (parallelize_loops): Likewise.
1645 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1646 "locus" from location_t to dump_user_location_t.
1647 (canonicalize_loop_induction_variables): Likewise.
1648 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1649 for change in return type of find_loop_location.
1650 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1651 to dump_printf_loc to pass in a dump_location_t rather than a
1652 location_t, via the stmt.
1653 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1655 * tree-vect-loop-manip.c (find_loop_location): Convert return
1656 type from source_location to dump_user_location_t.
1657 (vect_do_peeling): Update for above change.
1658 (vect_loop_versioning): Update for change in type of
1660 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1661 from location_t to dump_user_location_t.
1662 (vect_estimate_min_profitable_iters): Update for change in type
1664 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1665 location_t to dump_location_t.
1666 (vect_slp_bb): Update for change in type of vect_location.
1667 * tree-vectorizer.c (vect_location): Convert from source_location
1668 to dump_user_location_t.
1669 (try_vectorize_loop_1): Update for change in vect_location's type.
1670 (vectorize_loops): Likewise.
1671 (increase_alignment): Likewise.
1672 * tree-vectorizer.h (vect_location): Convert from source_location
1673 to dump_user_location_t.
1674 (find_loop_location): Convert return type from source_location to
1675 dump_user_location_t.
1676 (check_reduction_path): Convert 1st param from location_t to
1677 dump_user_location_t.
1678 * value-prof.c (check_counter): Update call to dump_printf_loc to
1679 pass in a dump_user_location_t rather than a location_t; update
1680 call to error_at for change in type of "locus".
1681 (check_ic_target): Update call to dump_printf_loc to
1682 pass in a dump_user_location_t rather than a location_t, via the
1685 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1687 * config/s390/s390.h (enum processor_flags): Do not use
1688 default tune parameter when -march was specified.
1690 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1693 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1694 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1697 2018-06-26 Richard Biener <rguenther@suse.de>
1699 PR tree-optimization/86287
1701 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1702 (vect_analyze_loop): Initialize n_stmts.
1704 2018-06-26 Richard Biener <rguenther@suse.de>
1707 * fold-const.c (fold_convertible_p): Pointer extension
1710 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1713 * dwarf2out.c (loc_list_has_views): Adjust comments.
1714 (dw_loc_list): Split single cross-partition range with
1717 2018-06-25 Jeff Law <law@redhat.com>
1719 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1720 on -mbig-switch by default.
1722 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1724 (const_float_0_operand): Remove unused predicate.
1725 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1726 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1727 (recipsf2): New expander. Original pattern now called
1729 (recipdf2, recipdf2_insn): Similarly.
1730 (rsqrtsf2, rsqrtsf2_insn): Similarly
1731 (rsqrtdf2, rsqrtdf2_insn): Similarly
1733 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1735 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1736 Simplify logic for FreeBSD (twice).
1738 2018-06-25 Martin Sebor <msebor@redhat.com>
1740 PR tree-optimization/86204
1741 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1742 a strnlen result if it's less than the length of the string.
1744 2018-06-25 Martin Sebor <msebor@redhat.com>
1746 PR tree-optimization/85700
1747 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1748 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1749 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1751 2018-06-25 Martin Sebor <msebor@redhat.com>
1753 * doc/extend.texi (Zero-length arrays): Update and clarify.
1755 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1757 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1758 added IEEE/IBM long double multilib support on PowerPC little
1759 endian Linux systems.
1760 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1761 (MULTILIB_DEFAULTS): Likewise.
1762 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1764 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1765 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1766 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1768 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1771 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1772 (REORDER_45): Likewise.
1774 2018-06-25 Jeff Law <law@redhat.com>
1776 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1777 dividend to 32 bits. Adjust length.
1778 (udivmodhi4): Cleanup output template. Fix length.
1780 2018-06-25 Carl Love <cel@us.ibm.com>
1782 * config/rs6000/vsx.md: Change word selector to prefered location.
1784 2018-06-25 Richard Biener <rguenther@suse.de>
1786 PR tree-optimization/86304
1787 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1788 epilogue-if-converted loops as well.
1790 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1792 * lto-section-out.c (lto_begin_section): Do not print section
1793 name for noaddr and unnumbered dumps.
1795 2018-06-25 Richard Biener <rguenther@suse.de>
1797 * tree-vectorizer.h (struct vec_info_shared): New structure
1798 with parts split out from struct vec_info and loop_nest from
1799 struct _loop_vec_info.
1800 (struct vec_info): Adjust accordingly.
1801 (struct _loop_vec_info): Likewise.
1802 (LOOP_VINFO_LOOP_NEST): Adjust.
1803 (LOOP_VINFO_DATAREFS): Likewise.
1804 (LOOP_VINFO_DDRS): Likewise.
1805 (struct _bb_vec_info): Likewise.
1806 (BB_VINFO_DATAREFS): Likewise.
1807 (BB_VINFO_DDRS): Likewise.
1808 (struct _stmt_vec_info): Add dr_aux member.
1809 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1810 (DR_MISALIGNMENT_UNINITIALIZED): New.
1811 (set_dr_misalignment): Adjust.
1812 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1813 (vect_analyze_loop): Adjust prototype.
1814 (vect_analyze_loop_form): Likewise.
1815 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1816 Compute dependences lazily.
1817 (vect_record_base_alignments): Use shared datarefs/ddrs.
1818 (vect_verify_datarefs_alignment): Likewise.
1819 (vect_analyze_data_refs_alignment): Likewise.
1820 (vect_analyze_data_ref_accesses): Likewise.
1821 (vect_analyze_data_refs): Likewise.
1822 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1823 constructor parameter for shared part.
1824 (vect_analyze_loop_form): Pass in shared part and adjust.
1825 (vect_analyze_loop_2): Pass in storage for the number of
1826 stmts. Move loop nest finding to the caller. Compute
1828 (vect_analyze_loop): Pass in shared part.
1829 (vect_transform_loop): Verify shared datarefs are unchanged.
1830 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1831 constructor parameter for shared part.
1832 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1833 (vect_slp_bb): Verify shared datarefs are unchanged before
1835 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1837 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1838 DR_MISALIGNMENT_UNINITIALIZED.
1839 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1840 parameter for shared part.
1841 (vec_info::~vec_info): Adjust.
1842 (vec_info_shared::vec_info_shared): New.
1843 (vec_info_shared::~vec_info_shared): Likewise.
1844 (vec_info_shared::save_datarefs): Likewise.
1845 (vec_info_shared::check_datarefs): Likewise.
1846 (try_vectorize_loop_1): Construct shared part live for analyses
1847 of a single loop for multiple vector sizes.
1848 * tree-parloops.c (gather_scalar_reductions): Adjust.
1850 2018-06-25 Richard Biener <rguenther@suse.de>
1852 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1853 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1854 (vect_analyze_data_refs): Remove similar code from here and
1855 simplify accordingly.
1857 2018-06-25 Richard Biener <rguenther@suse.de>
1859 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1860 for reverse storage order accesses rather than asserting
1861 they cannot happen here.
1863 2018-06-25 Tom de Vries <tdevries@suse.de>
1866 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1867 Use data16 instead of .byte for insn prefix.
1869 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1872 * parser.c (make_char_string_pack): Pass this literal chars
1873 through cpp_interpret_string.
1874 (cp_parser_userdef_numeric_literal): Check the result of
1875 make_char_string_pack.
1877 2018-06-24 Maya Rashish <coypu@sdf.org>
1879 * ginclude/stddef.h: Simplify conditions around avoiding
1880 re-definition of __size_t.
1882 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1884 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1885 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1887 2018-06-22 Maya Rashish <coypu@sdf.org>
1889 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1890 made to OpenBSD and NetBSD through the years.
1892 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1894 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1895 behavior of vec_pack (vector double, vector double) to match
1896 behavior of vec_float2 (vector double, vector double).
1898 2018-06-22 Olivier Hainque <hainque@adacore.com>
1900 * gimplify.c (gimplify_function_tree): Prevent creation
1901 of a trampoline for the address of the current function
1902 passed to entry/exit instrumentation hooks.
1904 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1907 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1910 2018-06-22 Martin Liska <mliska@suse.cz>
1912 PR tree-optimization/86263
1913 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1914 Bail out if is_enabled is false.
1915 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1917 (jump_table_cluster::is_enabled): New function.
1919 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1921 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1922 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1923 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1924 (lto_input_ts_binfo_tree_pointers): Likewise.
1925 * tree-streamer-out.c (streamer_write_tree_bitfields,
1926 write_ts_binfo_tree_pointers): Likewise.
1927 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1929 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1931 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1933 2018-06-22 Martin Liska <mliska@suse.cz>
1935 * symbol-summary.h (get): Make it pure and inline move
1936 functionality from ::get function.
1937 (get): Remove and inline into ::get and ::get_create.
1938 (get_create): Move code from ::get function.
1940 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1943 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1944 -x assembler-with-cpp.
1946 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1948 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1949 _FILE_OFFSET_BITS=64 for C++.
1951 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1953 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1954 conversion insn that shows up when pr85657-3.c is compiled using
1955 IEEE 128-bit long double.
1956 (neg<mode>2_internal): Use the correct mode to check whether the
1957 mode is IBM extended.
1958 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1959 multiply and divide external functions from being created more
1962 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1964 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1966 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1967 the edge can be forwarded.
1968 (cfg_layout_merge_blocks): Likewise.
1970 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1972 * except.c (finish_eh_generation): Commit edge insertions only after
1973 the EH edges have been redirected from post-landing to landing pads.
1975 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1977 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1978 create_tmp_var_for to create the FRAME decl.
1979 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1981 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1983 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1984 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1985 debug statement when resetting its value.
1986 (expand_call_inline): Copy the locus of the call onto the assignment
1987 of the return value, if any. Use local variable in more cases.
1989 2018-06-21 Martin Liska <mliska@suse.cz>
1991 * ipa-pure-const.c (propagate_nothrow): Use
1992 funct_state_summaries->get.
1993 (dump_malloc_lattice): Likewise.
1994 (propagate_malloc): Likewise.
1996 2018-06-21 Richard Biener <rguenther@suse.de>
1998 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1999 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2000 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2002 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2003 BLOCK_ABSTRACT_ORIGIN unconditionally.
2005 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2007 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2009 * ipa-reference.c (ipa_reference_c_finalize): Delete
2010 ipa_ref_opt_sum_summaries and set it to NULL.
2012 2018-06-21 Tom de Vries <tdevries@suse.de>
2014 PR tree-optimization/85859
2015 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2016 test with comment from bb_no_side_effects_p.
2018 2018-06-21 Richard Biener <rguenther@suse.de>
2020 PR tree-optimization/86232
2021 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2022 max for constant niter.
2024 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2026 * config/aarch64/aarch64-simd.md
2027 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2029 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2031 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2032 Make opernads of the unspec commutative.
2034 2018-06-21 Richard Biener <rguenther@suse.de>
2036 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2037 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2038 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2039 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2041 (vect_record_base_alignments): Do not record base alignment
2042 for gathers or scatters.
2043 (vect_compute_data_ref_alignment): Drop return value that is always
2044 true. Bail out early for gathers or scatters.
2045 (vect_enhance_data_refs_alignment): Bail out early for gathers
2047 (vect_find_same_alignment_drs): Likewise.
2048 (vect_analyze_data_refs_alignment): Remove dead code.
2049 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2050 (vect_analyze_data_refs): For possible gathers or scatters do
2051 not create an alternate DR, just check their possible validity
2052 and mark them. Adjust DECL_NONALIASED handling to not rely
2054 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2055 update inits of gathers or scatters.
2056 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2057 Also copy gather/scatter flag to pattern vinfo.
2059 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2061 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2062 behavior of vec_packsu (vector unsigned long long, vector unsigned
2063 long long) to match behavior of vec_packs with same signature.
2065 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2066 Thomas Schwinge <thomas@codesourcery.com>
2067 Cesar Philippidis <cesar@codesourcery.com>
2069 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2070 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2071 (gimplify_adjust_omp_clauses): Likewise.
2072 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2073 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2074 (gimplify_omp_target_update): Update handling of acc update and
2076 * omp-low.c (install_var_field): Remove unused parameter
2077 base_pointers_restrict.
2078 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2079 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2081 (omp_target_base_pointers_restrict_p): Delete.
2082 (scan_omp_target): Update call to scan_sharing_clauses.
2083 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2085 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2086 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2087 (convert_local_omp_clauses): Likewise.
2088 * tree-pretty-print.c (dump_omp_clause): Likewise.
2089 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2091 (omp_clause_code_name): Likewise.
2093 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2096 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2099 PR tree-optimization/86231
2100 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2101 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2103 2018-06-20 Tom de Vries <tdevries@suse.de>
2105 PR tree-optimization/86097
2106 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2107 iv type if signedness of iv type is not the same as that of *nit.
2109 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2111 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2112 EDGE_EH edges, verify they are all EDGE_EH.
2114 2018-06-20 Maya Rashish <coypu@sdf.org>
2116 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2118 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2120 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2121 * config/aarch64/aarch64.c (xgene1_tunings): Add
2122 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2123 (aarch64_mode_valid_for_sched_fusion_p):
2124 Allow 16-byte modes.
2125 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2126 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2128 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2130 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2131 * config/aarch64/iterators.md (VQ2): New mode iterator.
2133 2018-06-20 Martin Liska <mliska@suse.cz>
2135 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2136 Change default ratio from 10 to 8.
2138 2018-06-20 Martin Liska <mliska@suse.cz>
2140 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2142 (bit_test_cluster::find_bit_tests): Likewise.
2143 (switch_decision_tree::analyze_switch_statement): Find clusters.
2144 * tree-switch-conversion.h (struct jump_table_cluster): Document
2147 2018-06-20 Martin Liska <mliska@suse.cz>
2149 * tree-switch-conversion.c (switch_conversion::collect):
2150 Record m_uniq property.
2151 (switch_conversion::expand): Bail out for special conditions.
2152 (group_cluster::~group_cluster): New.
2153 (group_cluster::group_cluster): Likewise.
2154 (group_cluster::dump): Likewise.
2155 (jump_table_cluster::emit): New.
2156 (switch_decision_tree::fix_phi_operands_for_edges): New.
2157 (struct case_node): Remove struct.
2158 (jump_table_cluster::can_be_handled): New.
2159 (case_values_threshold): Moved to header.
2160 (reset_out_edges_aux): Likewise.
2161 (jump_table_cluster::is_beneficial): New.
2162 (bit_test_cluster::can_be_handled): Likewise.
2163 (add_case_node): Remove.
2164 (bit_test_cluster::is_beneficial): New.
2165 (case_bit_test::cmp): New.
2166 (bit_test_cluster::emit): New.
2167 (expand_switch_as_decision_tree_p): Remove.
2168 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2169 (fix_phi_operands_for_edge): Likewise.
2170 (switch_decision_tree::analyze_switch_statement): New.
2171 (compute_cases_per_edge): Move ...
2172 (switch_decision_tree::compute_cases_per_edge): ... here.
2173 (try_switch_expansion): Likewise.
2174 (switch_decision_tree::try_switch_expansion): Likewise.
2175 (record_phi_operand_mapping): Likewise.
2176 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2177 (emit_case_decision_tree): Likewise.
2178 (switch_decision_tree::emit): Likewise.
2179 (balance_case_nodes): Likewise.
2180 (switch_decision_tree::balance_case_nodes): Likewise.
2181 (dump_case_nodes): Likewise.
2182 (switch_decision_tree::dump_case_nodes): Likewise.
2183 (emit_jump): Likewise.
2184 (switch_decision_tree::emit_jump): Likewise.
2185 (emit_cmp_and_jump_insns): Likewise.
2186 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2187 (emit_case_nodes): Likewise.
2188 (switch_decision_tree::emit_case_nodes): Likewise.
2189 (conditional_probability): Remove.
2190 * tree-switch-conversion.h (enum cluster_type): New.
2192 (struct cluster): Likewise.
2193 (cluster::cluster): Likewise.
2194 (struct simple_cluster): Likewise.
2195 (simple_cluster::simple_cluster): Likewise.
2196 (struct group_cluster): Likewise.
2197 (struct jump_table_cluster): Likewise.
2198 (struct bit_test_cluster): Likewise.
2199 (struct min_cluster_item): Likewise.
2200 (struct case_tree_node): Likewise.
2201 (case_tree_node::case_tree_node): Likewise.
2202 (jump_table_cluster::case_values_threshold): Likewise.
2203 (struct case_bit_test): Likewise.
2204 (struct switch_decision_tree): Likewise.
2205 (struct switch_conversion): Likewise.
2206 (switch_decision_tree::reset_out_edges_aux): Likewise.
2208 2018-06-20 Martin Liska <mliska@suse.cz>
2210 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2211 (hoist_edge_and_branch_if_true): Likewise.
2212 (expand_switch_using_bit_tests_p): Likewise.
2213 (struct case_bit_test): Likewise.
2214 (case_bit_test_cmp): Likewise.
2215 (emit_case_bit_tests): Likewise.
2216 (switch_conversion::switch_conversion): New class.
2217 (struct switch_conv_info): Remove old struct.
2218 (collect_switch_conv_info): More to ...
2219 (switch_conversion::collect): ... this.
2220 (check_range): Likewise.
2221 (switch_conversion::check_range): Likewise.
2222 (check_all_empty_except_final): Likewise.
2223 (switch_conversion::check_all_empty_except_final): Likewise.
2224 (check_final_bb): Likewise.
2225 (switch_conversion::check_final_bb): Likewise.
2226 (create_temp_arrays): Likewise.
2227 (switch_conversion::create_temp_arrays): Likewise.
2228 (free_temp_arrays): Likewise.
2229 (gather_default_values): Likewise.
2230 (switch_conversion::gather_default_values): Likewise.
2231 (build_constructors): Likewise.
2232 (switch_conversion::build_constructors): Likewise.
2233 (constructor_contains_same_values_p): Likewise.
2234 (switch_conversion::contains_same_values_p): Likewise.
2235 (array_value_type): Likewise.
2236 (switch_conversion::array_value_type): Likewise.
2237 (build_one_array): Likewise.
2238 (switch_conversion::build_one_array): Likewise.
2239 (build_arrays): Likewise.
2240 (switch_conversion::build_arrays): Likewise.
2241 (gen_def_assigns): Likewise.
2242 (switch_conversion::gen_def_assigns): Likewise.
2243 (prune_bbs): Likewise.
2244 (switch_conversion::prune_bbs): Likewise.
2245 (fix_phi_nodes): Likewise.
2246 (switch_conversion::fix_phi_nodes): Likewise.
2247 (gen_inbound_check): Likewise.
2248 (switch_conversion::gen_inbound_check): Likewise.
2249 (process_switch): Use the newly created class.
2250 (switch_conversion::expand): New.
2251 (switch_conversion::~switch_conversion): New.
2252 * tree-switch-conversion.h: New file.
2254 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2256 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2257 tree-vect-patterns.c.
2258 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2259 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2260 (vect_recog_sad_pattern): Likewise.
2261 (vect_recog_widen_sum_pattern): Likewise.
2262 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2263 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2264 (vect_recog_rotate_pattern): Likewise.
2265 (vect_recog_mult_pattern): Likewise.
2266 (vect_recog_vector_vector_shift_pattern): Likewise.
2267 (vect_recog_divmod_pattern): Likewise.
2268 (vect_recog_mixed_size_cond_pattern): Likewise.
2269 (vect_recog_bool_pattern): Likewise.
2270 (vect_recog_mask_conversion_pattern): Likewise.
2271 (vect_try_gather_scatter_pattern): Likewise.
2272 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2273 (vect_recog_over_widening_pattern): Likewise.
2274 (vect_recog_gather_scatter_pattern): Likewise.
2275 (vect_recog_func_ptr): Move from tree-vectorizer.h
2276 (vect_vect_recog_func_ptrs): Move further down the file.
2277 (vect_recog_func): Likewise. Remove the third argument.
2278 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2279 (vect_pattern_recog_1): Expect the pattern function to do any
2280 necessary target tests. Also expect it to provide a vector type.
2281 Remove the type_in handling.
2283 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2285 * tree-vect-patterns.c (vect_pattern_detected): New function.
2286 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2287 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2288 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2289 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2290 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2291 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2292 (vect_recog_mask_conversion_pattern)
2293 (vect_try_gather_scatter_pattern): Likewise.
2295 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2297 * tree-vect-patterns.c (vect_get_internal_def): New function.
2298 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2299 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2300 (search_type_for_mask_1): Use it.
2302 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2304 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2305 redundant WIDEN_SUM_EXPR handling.
2306 (vect_recog_sad_pattern): Likewise.
2308 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2310 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2311 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2312 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2313 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2314 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2316 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2318 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2319 use the stmt_vec_info of the original bb statement for the
2320 new zero assignment, even if the call is part of a pattern.
2322 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2324 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2325 that the sequence is attached to the original statement rather
2326 than the pattern statement.
2327 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2328 PATTERN_DEF_SEQ from the original statement rather than
2329 the main pattern statement.
2330 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2331 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2332 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2334 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2336 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2337 definition statements before the early exit for statements that aren't
2339 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2341 (vect_transform_loop): ...here. Process pattern definition
2342 statements without first checking whether the main pattern
2343 statement is live or relevant.
2345 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2347 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2348 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2350 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2352 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2353 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2354 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2355 (expand_block_compare): Change select_block_compare_mode call.
2356 (expand_strncmp_align_check): Use new functions, fix comment.
2357 (emit_final_str_compare_gpr): New function.
2358 (expand_strn_compare): Refactor and clean up code.
2359 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2361 2018-06-19 Tony Reix <tony.reix@atos.com>
2362 Damien Bergamini <damien.bergamini@atos.com>
2363 David Edelsohn <dje.gcc@gmail.com>
2365 * collect2.c (static_obj): New variable.
2366 (static_libs): New variable.
2367 (is_in_list): Uncomment declaration.
2368 (main): Track AIX libraries linked statically.
2369 (is_in_list): Uncomment definition.
2370 (scan_prog_file): Don't add AIX shared libraries initializer
2371 to constructor list if linking statically.
2373 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2375 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2377 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2379 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2381 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2384 2018-06-19 Martin Liska <mliska@suse.cz>
2386 * config/i386/i386.c (ix86_can_inline_p): Do not use
2387 ipa_fn_summaries::get_create.
2388 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2390 (devirtualization_time_bonus): Likewise.
2391 (ipcp_propagate_stage): Likewise.
2392 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2393 (edge_set_predicate): Likewise.
2394 (evaluate_conditions_for_known_args): Likewise.
2395 (evaluate_properties_for_edge): Likewise.
2396 (ipa_call_summary::reset): Tranform to ...
2397 (ipa_call_summary::~ipa_call_summary): ... this.
2398 (ipa_fn_summary::reset): Transform to ...
2399 (ipa_fn_summary::~ipa_fn_summary): ... this.
2400 (ipa_fn_summary_t::remove): Rename to ...
2401 (ipa_fn_summary_t::remove_callees): ... this.
2402 (ipa_fn_summary_t::duplicate): Use placement new
2403 instead of memory copy.
2404 (ipa_call_summary_t::duplicate): Likewise.
2405 (ipa_call_summary_t::remove): Remove.
2406 (dump_ipa_call_summary): Change get_create to get.
2407 (ipa_dump_fn_summary): Dump only when summary exists.
2408 (analyze_function_body): Use symbol_summary::get instead
2410 (compute_fn_summary): Likewise.
2411 (estimate_edge_devirt_benefit): Likewise.
2412 (estimate_edge_size_and_time): Likewise.
2413 (inline_update_callee_summaries): Likewise.
2414 (remap_edge_change_prob): Likewise.
2415 (remap_edge_summaries): Likewise.
2416 (ipa_merge_fn_summary_after_inlining): Likewise.
2417 (write_ipa_call_summary): Likewise.
2418 (ipa_fn_summary_write): Likewise.
2419 (ipa_free_fn_summary): Likewise.
2420 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2421 (struct ipa_call_summary): Likewise.
2422 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2424 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2425 (estimate_size_after_inlining): Likewise.
2426 (estimate_growth): Likewise.
2427 (growth_likely_positive): Likewise.
2428 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2429 (inline_call): Likewise.
2430 * ipa-inline.c (caller_growth_limits): Likewise.
2431 (can_inline_edge_p): Likewise.
2432 (can_inline_edge_by_limits_p): Likewise.
2433 (compute_uninlined_call_time): Likewise.
2434 (compute_inlined_call_time): Likewise.
2435 (want_inline_small_function_p): Likewise.
2436 (edge_badness): Likewise.
2437 (update_caller_keys): Likewise.
2438 (update_callee_keys): Likewise.
2439 (inline_small_functions): Likewise.
2440 (inline_to_all_callers_1): Likewise.
2441 (dump_overall_stats): Likewise.
2442 (early_inline_small_functions): Likewise.
2443 (early_inliner): Likewise.
2444 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2445 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2446 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2447 * ipa-split.c (execute_split_functions): Likewise.
2448 * symbol-summary.h: Likewise.
2449 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2451 2018-06-19 Richard Biener <rguenther@suse.de>
2453 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2454 (vectorize_loops): ... here. Fix dbgcnt handling.
2455 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2457 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2460 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2461 ieee128 argument takes up only one (vector) register, not two (floating
2464 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2466 * gimplify.c (gimplify_init_constructor): Really never clear for an
2467 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2469 2018-06-19 Richard Biener <rguenther@suse.de>
2471 PR tree-optimization/86179
2472 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2473 after failed recognition.
2475 2018-06-18 Martin Sebor <msebor@redhat.com>
2478 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2479 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2480 Handle integer subtraction.
2481 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2482 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2484 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2486 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2487 param from rtx to rtx_insn *.
2488 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2490 (frv_ifcvt_modify_insn): Likwise.
2491 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2492 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2493 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2494 as_a <rtx_insn *> cast to local "unprotected_region" once
2495 it's been established that it's not NULL or pc_rtx.
2496 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2497 param "sethi" from rtx to rtx_insn *.
2498 (nds32_group_float_insns): Likewise for param "insn".
2499 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2501 (vax_output_int_subtract): Likewise.
2502 * config/vax/vax.c (vax_output_int_add): Likewise for param
2504 (vax_output_int_subtract): Likewise.
2505 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2506 (emit_pattern_after): Likewise for param "after".
2507 (emit_insn_after): Likewise.
2508 (emit_jump_insn_after): Likewise.
2509 (emit_call_insn_after): Likewise.
2510 (emit_debug_insn_after): Likewise.
2511 (emit_pattern_before): Likewise for param "before".
2512 (emit_insn_before): Likewise.
2513 (emit_jump_insn_before): Likewise.
2514 * final.c (get_insn_template): Likewise for param "insn", removing
2516 * output.h (get_insn_template): Likewise for 2nd param.
2517 * rtl.h (emit_insn_before): Likewise.
2518 (emit_jump_insn_before): Likewise.
2519 (emit_debug_insn_before_noloc): Likewise.
2520 (emit_insn_after): Likewise.
2521 (emit_jump_insn_after): Likewise.
2522 (emit_call_insn_after): Likewise.
2523 (emit_debug_insn_after): Likewise.
2524 (set_insn_deleted): Likewise for param.
2526 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2529 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2530 floating point modes, so that IFmode is numerically greater than
2531 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2532 to declare the ordering. This prevents IFmode from being
2533 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2534 machine. Include rs6000-modes.h to share the fractional values
2535 between genmodes* and the rest of the compiler.
2539 * config/rs6000/rs6000-modes.h: New file.
2540 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2541 meaning of rs6000_long_double_size so that 126..128 selects an
2542 appropriate 128-bit floating point type.
2543 (rs6000_option_override_internal): Likewise.
2544 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2545 (TARGET_LONG_DOUBLE_128): Change the meaning of
2546 rs6000_long_double_size so that 126..128 selects an appropriate
2547 128-bit floating point type.
2548 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2549 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2550 source and destination to match the standard usage.
2551 (truncifkf2): Likewise.
2552 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2553 ISA 2.07 to use an explicit clobber, instead of passing in a
2555 (copysign<mode>3_soft): Likewise.
2557 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2559 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2560 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2561 (vect_slp_analyze_instance_dependence): Likewise.
2562 (vect_enhance_data_refs_alignment): Likewise.
2563 (vect_analyze_data_refs_alignment): Likewise.
2564 (vect_slp_analyze_and_verify_instance_alignment
2565 (vect_analyze_data_ref_accesses): Likewise.
2566 (vect_prune_runtime_alias_test_list): Likewise.
2567 (vect_analyze_data_refs): Likewise.
2568 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2569 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2570 (vect_analyze_scalar_cycles_1): Likewise.
2571 (vect_get_loop_niters): Likewise.
2572 (vect_analyze_loop_form_1): Likewise.
2573 (vect_update_vf_for_slp): Likewise.
2574 (vect_analyze_loop_operations): Likewise.
2575 (vect_analyze_loop): Likewise.
2576 (vectorizable_induction): Likewise.
2577 (vect_transform_loop): Likewise.
2578 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2579 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2580 (vect_make_slp_decision): Likewise.
2581 (vect_detect_hybrid_slp): Likewise.
2582 (vect_slp_analyze_operations): Likewise.
2583 (vect_slp_bb): Likewise.
2584 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2585 (vectorizable_bswap): Likewise.
2586 (vectorizable_call): Likewise.
2587 (vectorizable_simd_clone_call): Likewise.
2588 (vectorizable_conversion): Likewise.
2589 (vectorizable_assignment): Likewise.
2590 (vectorizable_shift): Likewise.
2591 (vectorizable_operation): Likewise.
2592 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2594 2018-06-18 Martin Sebor <msebor@redhat.com>
2596 PR tree-optimization/81384
2597 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2598 * builtins.c (expand_builtin_strnlen): New function.
2599 (expand_builtin): Call it.
2600 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2601 * builtins.def (BUILT_IN_STRNLEN): New.
2602 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2603 Warn for bounds in excess of maximum object size.
2604 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2605 single-value ranges. Handle strnlen.
2606 (handle_builtin_strlen): Handle strnlen.
2607 (strlen_check_and_optimize_stmt): Same.
2608 * doc/extend.texi (Other Builtins): Document strnlen.
2610 2018-06-18 Maya Rashish <coypu@sdf.org>
2612 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2613 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2614 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2616 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2618 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2620 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2622 * tree.c (escaped_string::escape): Replace cast to char * by
2623 const_cast<char *> (unescaped).
2625 2018-06-18 Nick Clifton <nickc@redhat.com>
2628 * tree.c (escaped_string): New class. Converts an unescaped
2629 string into its escaped equivalent.
2630 (warn_deprecated_use): Use the new class to convert the
2631 deprecation message, if present.
2632 (test_escaped_strings): New self test.
2633 (test_c_tests): Add test_escaped_strings.
2634 * doc/extend.texi (deprecated): Add a note that the
2635 deprecation message is affected by the -fmessage-length
2636 option, and that control characters will be escaped.
2637 (#pragma GCC error): Document this pragma.
2638 (#pragma GCC warning): Likewise.
2639 * doc/invoke.texi (-fmessage-length): Document this option's
2640 effect on the #warning and #error preprocessor directives and
2641 the deprecated attribute.
2643 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2645 * tree.c (decl_value_expr_lookup): Revert latest change.
2646 (decl_value_expr_insert): Likewise.
2648 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2650 * gimplify.c (nonlocal_vlas): Delete.
2651 (nonlocal_vla_vars): Likewise.
2652 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2654 (gimplify_body): Do not create and destroy nonlocal_vlas.
2655 * tree-nested.c: Include diagnostic.h.
2656 (use_pointer_in_frame): Tweak.
2657 (lookup_field_for_decl): Add assertion and declare the transformation.
2658 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2659 internal error when the reference is in a wrong context. Do not
2660 create a debug decl by default.
2661 (note_nonlocal_block_vlas): Delete.
2662 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2663 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2664 create a debug decl by default.
2665 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2666 call is in a wrong context.
2667 (fixup_vla_decls): New function.
2668 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2669 debug variables were created.
2670 * tree.c (decl_value_expr_lookup): Add checking assertion.
2671 (decl_value_expr_insert): Likewise.
2673 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2676 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2677 * tree-scalar-evolution.c (interpret_expr): Likewise.
2678 (expression_expensive_p): Likewise.
2679 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2680 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2681 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2682 (ssa_defined_by_minus_one_stmt_p): New.
2684 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2687 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2688 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2689 * dojump.c (do_jump): Likewise.
2690 * expr.c (expand_expr_real_2): Check operand type's sign.
2691 * fold-const.c (const_unop): Handle ABSU_EXPR.
2692 (fold_abs_const): Likewise.
2693 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2694 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2695 (strip_sign_op_1): Likesise.
2696 * match.pd: Add new pattern to generate ABSU_EXPR.
2697 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2698 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2699 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2700 * tree-inline.c (estimate_operator_cost): Likewise.
2701 * tree-pretty-print.c (dump_generic_node): Likewise.
2702 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2703 * tree.def (ABSU_EXPR): New.
2705 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2708 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2709 documented as preserved for backward compatibility only.
2710 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2712 PR rtl-optimization/86108
2713 * bb-reorder.c (create_forwarder_block): Renamed to ...
2714 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2715 jump from new landing pad to the second part.
2716 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2719 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2722 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2723 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2724 Only call store_expr for halves if the mode is the same.
2727 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2728 Fix up comment formatting.
2730 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2732 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2733 (typed_splay_tree::closure,
2734 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2735 (typed_splay_tree::typed_splay_tree,
2736 typed_splay_tree::operator =): Declared private.
2737 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2738 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2739 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2740 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2741 typed_splay_tree::splay_tree_splay,
2742 typed_splay_tree::splay_tree_foreach_helper,
2743 typed_splay_tree::splay_tree_insert,
2744 typed_splay_tree::splay_tree_remove,
2745 typed_splay_tree::splay_tree_lookup,
2746 typed_splay_tree::splay_tree_predecessor,
2747 typed_splay_tree::splay_tree_successor,
2748 typed_splay_tree::splay_tree_min,
2749 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2750 (typed_splay_tree::root, typed_splay_tree::comp,
2751 typed_splay_tree::delete_key,
2752 typed_splay_tree::delete_value): New data members.
2753 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2754 typed_splay_tree::remove.
2756 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2758 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2759 -mginv and -mno-ginv to the assembler.
2760 * config/mips/mips.opt (-mcrc): New option.
2762 * doc/invoke.text (-mcrc): Document.
2765 2018-06-15 Nick Clifton <nickc@redhat.com>
2768 * tree.c (escaped_string): New class. Converts an unescaped
2769 string into its escaped equivalent.
2770 (warn_deprecated_use): Use the new class to convert the
2771 deprecation message, if present.
2772 (test_escaped_strings): New self test.
2773 (test_c_tests): Add test_escaped_strings.
2774 * doc/extend.texi (deprecated): Add a note that the
2775 deprecation message is affected by the -fmessage-length
2776 option, and that control characters will be escaped.
2777 (#pragma GCC error): Document this pragma.
2778 (#pragma GCC warning): Likewise.
2779 * doc/invoke.texi (-fmessage-length): Document this option's
2780 effect on the #warning and #error preprocessor directives and
2781 the deprecated attribute.
2783 2018-06-15 Richard Biener <rguenther@suse.de>
2785 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2786 here, also noting vector size used.
2787 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2788 size used in MSG_OPTIMIZED_LOCATIONS dump.
2789 (pass_slp_vectorize::execute): Adjust.
2791 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2794 * config/arc/arc.c (arc_return_address_register): Fix
2797 2018-06-15 Richard Biener <rguenther@suse.de>
2800 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2801 leave useless conversion stripping to force_gimple_operand_gsi.
2802 (gimplify_build2): Likewise.
2803 (gimplify_build1): Likewise.
2805 2018-06-15 Richard Biener <rguenther@suse.de>
2808 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2809 before adjusting their block.
2811 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2813 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2814 multilibs for *-*-rtems*.
2815 * config/riscv/t-rtems: New file.
2817 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2820 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2821 unsigned_type_for returns NULL.
2824 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2825 subregs of multi-word pseudos unless the float mode has word size.
2827 2018-06-14 Richard Biener <rguenther@suse.de>
2830 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2831 duplicate and harmful type_hash_canon.
2832 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2834 2018-06-14 Richard Biener <rguenther@suse.de>
2837 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2840 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2842 * config/rtems.h (STDINT_LONG32): Define.
2844 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2845 Prachi Godbole <prachi.godbole@imgtec.com>
2847 * config/mips/mips-cpus.def: Define P6600.
2848 * config/mips/mips-tables.opt: Regenerate.
2849 * config/mips/mips.c (mips_ucbranch_type): New enum.
2850 (mips_rtx_cost_data): Add support for P6600.
2851 (mips_issue_rate): Likewise.
2852 (mips_multipass_dfa_lookahead): Likewise.
2853 (mips_avoid_hazard): Likewise.
2854 (mips_reorg_process_insns): Likewise.
2855 (mips_classify_branch_p6600): New function.
2856 * config/mips/mips.h (TUNE_P6600): New define.
2857 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2858 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2859 * config/mips/mips.md: Include p6600.md.
2860 (processor): Add p6600.
2861 * config/mips/p6600.md: New file.
2862 * doc/invoke.texi: Add p6600 to supported architectures.
2864 2018-06-13 Martin Sebor <msebor@redhat.com>
2866 PR tree-optimization/86114
2867 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2869 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2871 2018-06-13 Richard Biener <rguenther@suse.de>
2873 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2874 Properly set vector type of the intermediate stmt.
2875 * tree-vect-stmts.c (vectorizable_operation): The destination
2876 var always has vectype_out type.
2878 2018-06-13 Jeff Law <law@redhat.com>
2880 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2881 integer 0 for argument to print_rtl_with_bb.
2882 (rl78_reorg): Likewise.
2884 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2886 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2887 from rtx to rtx_insn *.
2888 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2890 (add_sched_insns_for_speculation): Likewise for local "target",
2891 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2892 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2893 from rtx_insn ** to rtx_code_label **.
2894 (reorg_emit_nops): Likewise.
2895 (c6x_reorg): Likewise for local "call_labels".
2896 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2898 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2899 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2900 the loops over LABEL_REFS.
2901 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2903 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2904 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2905 (split_branches): Strengthen local "olabel" from rtx to
2906 rtx_insn *, adding a safe_as_a cast.
2907 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2909 (add_insn_after): Likewise for first two params.
2910 (add_insn_before): Likewise.
2911 (remove_insn): Likewise for param.
2912 (emit_pattern_before_noloc): Likewise for second and third params.
2913 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2914 (emit_call_insn_before_noloc): Likewise.
2915 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2917 (emit_barrier_before): Likewise.
2918 (emit_label_before): Strengthen "label" param from "rtx" to
2919 "rtx_code_label *". Strengthen "before" param from "rtx" to
2921 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2923 (emit_pattern_after_noloc): Likewise.
2924 (emit_insn_after_noloc): Likewise.
2925 (emit_jump_insn_after_noloc): Likewise.
2926 (emit_call_insn_after_noloc): Likewise.
2927 (emit_debug_insn_after_noloc): Likewise.
2928 (emit_barrier_after): Likewise.
2929 (emit_label_after): Likewise for both params.
2930 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2931 "loc" param from "int" to "location_t".
2932 (emit_insn_after_setloc): Likewise.
2933 (emit_jump_insn_after_setloc): Likewise.
2934 (emit_call_insn_after_setloc): Likewise.
2935 (emit_debug_insn_after_setloc): Likewise.
2936 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2937 "loc" param from "int" to "location_t".
2938 (emit_pattern_before): Convert NULL_RTX to NULL.
2939 (emit_insn_before_setloc): Convert "loc" param from "int" to
2941 (emit_jump_insn_before_setloc): Likewise.
2942 (emit_call_insn_before_setloc): Likewise.
2943 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2944 rtx_insn *. Convert "loc" param from "int" to "location_t".
2945 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2946 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2947 Convert 3rd param from "int" to "location_t".
2948 (emit_barrier_before, emit_barrier_after, next_real_insn):
2949 Strengthen param from rtx to rtx_insn *.
2950 (emit_label_before): Strengthen 1st param from "rtx" to
2951 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2953 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2954 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2955 Strengthen 2nd param from "rtx" to "rtx_insn *".
2956 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2957 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2958 Likewise. Convert 3rd param from "int" to "location_t".
2959 (emit_label_after): Strengthen 1st param from "rtx" to
2961 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2963 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2964 from "rtx" to "rtx_insn *".
2966 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2968 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2969 bodies streamed in with -Q.
2970 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2971 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2972 * gimple-streamer-out.c: Include gimple-pretty-print.h
2973 (output_bb): Dump stmts streamed.
2974 * lto-section-out.c: Include print-tree.h
2975 (lto_begin_section): Dump sections created.
2976 (lto_output_decl_index): Dump decl encoded.
2977 * lto-streamer-out.c: Include print-tree.h
2978 (create_output_block): Dump output block created.
2979 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2980 (output_function): Dump function output.
2981 (output_constructor): Dump constructor streamed.
2982 (write_global_stream): Output indexes encoded.
2983 (produce_asm_for_decls): Dump streams encoded.
2984 * lto-streamer.c (streamer_dump_file): New global var.
2985 * lto-streamer.h (streamer_dump_file): Declare.
2986 * passes.c (ipa_write_summaries): Initialize streamer dump.
2987 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2990 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2993 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2994 offsets for register save directives. Emit a second batch of save
2995 directives, if need be, when the function accesses prior frames.
2997 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2999 * config/arc/fpu.md (fmasf4): Force operand to register.
3000 (fnmasf4): Likewise.
3002 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3004 * config/arc/arc-protos.h (arc_pad_return): Remove.
3005 * config/arc/arc.c (machine_function): Remove force_short_suffix
3007 (arc_print_operand): Adjust printing of '&'.
3008 (arc_verify_short): Remove conditional printing of short suffix.
3009 (arc_final_prescan_insn): Remove reference to size_reason.
3010 (pad_return): New function.
3011 (arc_reorg): Call pad_return.
3012 (arc_pad_return): Remove.
3013 (arc_init_machine_status): Remove reference to force_short_suffix.
3014 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3015 (attr length): When attribute iscompact is true force to 2
3016 regardless; in the case of maybe check if we want to force the
3017 instruction to have 4 bytes length.
3018 (nopv): Change it to generate 4 byte long nop as well.
3019 (blockage): New pattern.
3020 (simple_return): Remove call to arc_pad_return.
3021 (p_return_i): Likewise.
3023 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3025 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3027 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3029 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3032 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3034 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3035 for ARC700 and ARCv2.
3037 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3040 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3041 operands[2] instead of operands[1].
3044 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3046 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3047 case, check whether the outer register overlaps an unallocatable
3048 register, not just whether it fits the required class.
3050 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3052 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3054 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3055 (memrefs_conflict_p): Likewise.
3056 (init_alias_analysis): Likewise.
3057 * cfgexpand.c (expand_debug_expr): Likewise.
3058 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3059 * cse.c (fold_rtx): Likewise.
3060 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3061 * expr.c (emit_block_move_hints): Likewise.
3062 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3063 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3064 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3065 (emit_group_store): Likewise.
3066 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3067 to read the PRE/POST_MODIFY increment.
3068 * calls.c (store_one_arg): Use strip_offset.
3069 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3071 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3073 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3074 (simplify_binary_operation_1): Extend CONST_INT handling to
3076 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3077 than a HOST_WIDE_INT.
3078 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3080 (adjust_mems, add_stores): Update accodingly.
3081 (vt_canonicalize_addr): Track polynomial offsets.
3082 (emit_note_insn_var_location): Likewise.
3083 (vt_add_function_parameter): Likewise.
3084 (vt_initialize): Likewise.
3086 2018-06-12 Jeff Law <law@redhat.com>
3088 * config.gcc (alpha*-*-freebsd*): Remove.
3089 * config/alpha/freebsd.h: Remove.
3091 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3094 * spellcheck-tree.c (levenshtein_distance): Rename to...
3095 (get_edit_distance): ...this, and update for underlying renaming.
3096 * spellcheck-tree.h (levenshtein_distance): Rename to...
3097 (get_edit_distance): ...this.
3098 * spellcheck.c (levenshtein_distance): Rename to...
3099 (get_edit_distance): ...this. Convert from Levenshtein distance
3100 to Damerau-Levenshtein distance by supporting transpositions of
3101 adjacent characters. Rename "v1" to "v_next" and "v0" to
3103 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3104 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3105 update for underlying renaming.
3106 (selftest::levenshtein_distance_unit_test): Rename to...
3107 (selftest::test_get_edit_distance_unit): ...this, and update for
3108 underlying renaming.
3109 (selftest::test_find_closest_string): Add example from PR 69968
3110 where transposition helps
3111 (selftest::test_metric_conditions): Update for renaming.
3112 (selftest::test_metric_conditions): Likewise.
3113 (selftest::spellcheck_c_tests): Likewise.
3114 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3115 (get_edit_distance): ...this.
3116 (best_match::consider): Update for renaming.
3118 2018-06-12 Martin Sebor <msebor@redhat.com>
3120 PR tree-optimization/85259
3121 * builtins.c (compute_objsize): Handle constant offsets.
3122 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3123 true iff a warning has been issued.
3124 * gimple.h (gimple_nonartificial_location): New function.
3125 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3126 gimple_nonartificial_location and handle -Wno-system-headers.
3127 (handle_builtin_stxncpy): Same.
3129 2018-06-12 Martin Sebor <msebor@redhat.com>
3132 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3134 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3136 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3137 BUILTIN_VEC_XST entries for pointer to double and long long.
3139 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3142 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3144 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3147 2018-06-12 Martin Liska <mliska@suse.cz>
3149 * doc/options.texi: Document IntegerRange.
3151 2018-06-12 Martin Liska <mliska@suse.cz>
3153 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3154 * opt-functions.awk: Handle Deprecated flag.
3155 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3157 (read_cmdline_option): Report warning for a deprecated option.
3158 * opts.h (struct cl_option): Add new field cl_deprecated.
3159 (CL_ERR_DEPRECATED): New.
3161 2018-06-12 Martin Liska <mliska@suse.cz>
3163 * doc/options.texi: Document Deprecated option flag.
3165 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3167 * config/arc/arc-arch.h (arc_extras): New enum.
3168 (arc_cpu_t):Add field extra.
3169 (arc_cpu_types): Consider the extras.
3170 * config/arc/arc-cpus.def: Add extras info.
3171 * config/arc/arc-opts.h (processor_type): Consider extra field.
3172 * config/arc/arc.c (arc_override_options): Handle extra field.
3174 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3176 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3177 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3179 * config/arc/arc.c (arc_init): Update pic warning.
3180 (irq_range): Update irq range parsing warnings.
3181 (arc_override_options): Update various warning messages.
3182 (arc_handle_aux_attribute): Likewise.
3184 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3186 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3188 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3190 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3191 compiler messages directives.
3193 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3195 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3196 * config/mips/mips-tables.opt: Regenerate.
3197 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3199 * doc/invoke.texi: Document -march=i6500.
3201 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3203 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3204 (i6400_gpmul): Add cpu_unit.
3205 (i6400_gpdiv): Likewise.
3206 (i6400_msa_add_d): Update reservations.
3207 (i6400_msa_int_add) Likewise.
3208 (i6400_msa_short_logic3) Likewise.
3209 (i6400_msa_short_logic2) Likewise.
3210 (i6400_msa_short_logic) Likewise.
3211 (i6400_msa_move) Likewise.
3212 (i6400_msa_cmp) Likewise.
3213 (i6400_msa_short_float2) Likewise.
3214 (i6400_msa_div_d) Likewise.
3215 (i6400_msa_long_logic1) Likewise.
3216 (i6400_msa_long_logic2) Likewise.
3217 (i6400_msa_mult) Likewise.
3218 (i6400_msa_long_float2) Likewise.
3219 (i6400_msa_long_float4) Likewise.
3220 (i6400_msa_long_float5) Likewise.
3221 (i6400_msa_long_float8) Likewise.
3222 (i6400_fpu_fadd): Include frint type.
3223 (i6400_fpu_store): New define_insn_reservation.
3224 (i6400_fpu_load): Likewise.
3225 (i6400_fpu_move): Likewise.
3226 (i6400_fpu_fcmp): Likewise.
3227 (i6400_fpu_fmadd): Likewise.
3228 (i6400_int_mult): Include imul3nc type and update reservation.
3229 (i6400_int_div): Include idiv3 type and update reservation.
3230 (i6400_int_load): Update to check type not move_type.
3231 (i6400_int_store): Likewise.
3232 (i6400_int_prefetch): Set zero latency.
3234 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3236 * gcc.c: Document new %@{...} sequence.
3237 (LINK_COMMAND_SPEC): Use it for the -L switches.
3238 (cpp_unique_options): Use it for the -I switches.
3239 (at_file_argbuf): New global variable.
3240 (in_at_file): Likewise.
3241 (alloc_args): Create at_file_argbuf.
3242 (clear_args): Truncate at_file_argbuf.
3243 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3244 (open_at_file): New function.
3245 (close_at_file): Likewise.
3246 (create_at_file): Delete.
3247 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3250 (validate_switches_from_spec): Deal with %@{...} sequence.
3251 (validate_switches): Likewise.
3252 (driver::finalize): Call clear_args.
3254 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3256 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3258 2018-06-11 Martin Sebor <msebor@redhat.com>
3260 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3261 (Wno-class-memaccess): Add @opindex.
3262 (Wno-templates, Wno-multiple-inheritance): Same.
3263 (Wno-virtual-inheritance, Wno-namespaces): Same.
3264 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3265 (Wno-format-overflow, Wno-format-truncation): Same.
3266 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3267 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3268 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3269 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3270 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3271 (Wno-misspelled-isr): Same.
3273 2018-06-11 Martin Sebor <msebor@redhat.com>
3275 * PR tree-optimization/86083
3276 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3278 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3280 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3282 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3285 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3286 on the correct operand.
3287 (*movdi_internal64): Ditto.
3289 2018-06-11 Martin Liska <mliska@suse.cz>
3291 PR tree-optimization/86089
3292 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3294 2018-06-11 Julia Koval <julia.koval@intel.com>
3296 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3297 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3298 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3300 2018-06-11 Olivier Hainque <hainque@adacore.com>
3302 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3303 for Ada with strict dwarf2.
3305 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3308 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3311 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3313 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3315 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3317 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3320 2018-06-08 Martin Liska <mliska@suse.cz>
3322 * tree-cfg.h (debug_function): Fix argument type to match
3325 2018-06-08 Martin Liska <mliska@suse.cz>
3327 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3328 Remove usage of MPX-related (and removed) fields.
3329 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3331 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3333 * cfg.c (debug): Use TDF_NONE rather than 0.
3334 * cfghooks.c (debug): Likewise.
3335 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3336 (struct dump_option_value_info): Convert to...
3337 (struct kv_pair): ...this template type.
3338 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3340 (optinfo_verbosity_options): Likewise.
3341 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3343 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3344 than int for "optgroup_flags" param.
3345 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3347 (dump_dec): Likewise.
3348 (dump_finish): Use TDF_NONE rather than 0.
3349 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3350 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3351 than 0. Update for change to option_ptr.
3352 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3353 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3354 0. Update for changes to optinfo_verbosity_options and
3356 (opt_info_switch_p): Convert optgroup_flags from int to
3358 (dump_basic_block): Use dump_flags_t rather than int
3359 for "dump_kind" param.
3360 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3361 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3362 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3363 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3364 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3365 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3366 TDF_NONE): Convert from macros to...
3367 (enum dump_flag): ...this new enum.
3368 (dump_flags_t): Update to use enum.
3369 (operator|, operator&, operator~, operator|=, operator&=):
3370 Implement for dump_flags_t.
3371 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3372 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3373 Convert from macros to...
3374 (enum optgroup_flag): ...this new enum.
3375 (optgroup_flags_t): New typedef.
3376 (operator|, operator|=): Implement for optgroup_flags_t.
3377 (struct dump_file_info): Convert field "alt_flags" to
3378 dump_flags_t. Convert field "optgroup_flags" to
3380 (dump_basic_block): Use dump_flags_t rather than int for param.
3381 (dump_generic_expr_loc): Likewise.
3382 (dump_dec): Likewise.
3383 (dump_register): Convert param "optgroup_flags" to
3385 (opt_info_enable_passes): Likewise.
3386 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3388 * gimple-pretty-print.c (debug): Likewise.
3389 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3390 (merged_store_group::apply_stores): Likewise.
3391 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3392 * gimple.c (verify_gimple_pp): Likewise.
3393 * graphite-poly.c (print_pbb_body): Likewise.
3394 * passes.c (pass_manager::register_one_dump_file): Convert
3395 local "optgroup_flags" to optgroup_flags_t.
3396 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3398 (debug_body): Likewise.
3399 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3400 to optgroup_flags_t.
3401 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3403 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3404 (convert_mult_to_fma): Likewise.
3405 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3406 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3407 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3408 "dump_kind" to dump_flags_t.
3410 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3412 * config/rs6000/rs6000.c (min, max): Delete.
3414 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3416 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3417 -mabi=spe and -mabi=no-spe.
3419 2018-06-08 Martin Liska <mliska@suse.cz>
3421 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3422 where we expect an existing summary.
3424 2018-06-08 Martin Liska <mliska@suse.cz>
3426 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3427 * ipa-inline.h (estimate_edge_growth): Likewise.
3429 2018-06-08 Martin Liska <mliska@suse.cz>
3431 * cgraph.c (function_version_hasher::hash): Use
3432 cgraph_node::get_uid ().
3433 (function_version_hasher::equal):
3434 * cgraph.h (cgraph_node::get_uid): New method.
3435 * ipa-inline.c (update_caller_keys): Use
3436 cgraph_node::get_uid ().
3437 (update_callee_keys): Likewise.
3438 * ipa-utils.c (searchc): Likewise.
3439 (ipa_reduced_postorder): Likewise.
3440 * lto-cgraph.c (input_node): Likewise.
3441 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3442 * symbol-summary.h (symtab_insertion): Likewise.
3443 (symtab_removal): Likewise.
3444 (symtab_duplication): Likewise.
3445 * tree-pretty-print.c (dump_function_header): Likewise.
3446 * tree-sra.c (convert_callers_for_node): Likewise.
3448 2018-06-08 Martin Liska <mliska@suse.cz>
3450 * cgraph.c (symbol_table::create_edge): Always assign a new
3452 (symbol_table::free_edge): Do not recycle numbers.
3453 * cgraph.h (cgraph_edge::get): New method.
3454 * symbol-summary.h (symtab_removal): Use it.
3455 (symtab_duplication): Likewise.
3456 (call_summary::hashable_uid): Remove.
3458 2018-06-08 Martin Liska <mliska@suse.cz>
3460 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3461 (initialize_growth_caches): Remove.
3462 (free_growth_caches): Likewise.
3463 (do_estimate_edge_time): Use edge_growth_cache.
3464 (do_estimate_edge_size): Likewise.
3465 (do_estimate_edge_hints): Likewise.
3466 * ipa-inline.c (reset_edge_caches): Likewise.
3467 (recursive_inlining): Likewise.
3468 (inline_small_functions): Likewise.
3469 * ipa-inline.h (initialize_growth_caches): Remove.
3470 (estimate_edge_size): Likewise.
3471 (estimate_edge_time): Likewise.
3472 (estimate_edge_hints): Likewise.
3473 (reset_edge_growth_cache): Likewise.
3474 * symbol-summary.h (call_summary::remove): New method.
3476 2018-06-08 Martin Liska <mliska@suse.cz>
3478 * ipa-cp.c (class edge_clone_summary): New summary.
3479 (grow_edge_clone_vectors): Remove.
3480 (ipcp_edge_duplication_hook): Remove.
3481 (class edge_clone_summary_t): New call_summary class.
3482 (ipcp_edge_removal_hook): Remove.
3483 (edge_clone_summary_t::duplicate): New function.
3484 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3485 (create_specialized_node): Likewise.
3486 (ipcp_driver): Initialize edge_clone_summaries and do not
3489 2018-06-08 Martin Liska <mliska@suse.cz>
3491 * symbol-summary.h (get): New function.
3492 (call_summary::m_initialize_when_cloning): New class member.
3494 2018-06-08 Martin Liska <mliska@suse.cz>
3496 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3497 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3498 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3499 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3500 (remove_cgraph_node_from_order): Use the removed_nodes set.
3501 (do_per_function_toporder): Likwise.
3502 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3503 instead of summary_uid.
3504 (symtab_removal): Likewise.
3505 (symtab_duplication): Likewise.
3507 2018-06-08 Martin Liska <mliska@suse.cz>
3509 * ipa-cp.c (ipcp_store_bits_results): Use
3510 ipcp_transformation_sum.
3511 (ipcp_store_vr_results): Likewise.
3512 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3514 (ipcp_transformation_initialize): ... this.
3515 (ipa_set_node_agg_value_chain):
3516 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3517 (write_ipcp_transformation_info): Likewise.
3518 (read_ipcp_transformation_info): Likewise.
3519 (ipcp_update_bits): Likewise.
3520 (ipcp_update_vr): Likewise.
3521 (ipcp_transform_function): Likewise.
3522 * ipa-prop.h: Rename ipcp_transformation_summary to
3523 ipcp_transformation.
3524 (class ipcp_transformation_t): New function summary.
3525 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3526 (ipa_get_agg_replacements_for_node): Likewise.
3528 2018-06-08 Martin Liska <mliska@suse.cz>
3530 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3532 (class funct_state_summary_t): New function_summary class.
3533 (has_function_state): Remove.
3534 (get_function_state): Likewise.
3535 (set_function_state): Likewise.
3536 (add_new_function): Likewise.
3537 (funct_state_summary_t::insert): New function.
3538 (duplicate_node_data): Remove.
3539 (remove_node_data): Remove.
3540 (funct_state_summary_t::duplicate): New function.
3541 (register_hooks): Create new funct_state_summaries.
3542 (pure_const_generate_summary): Use it.
3543 (pure_const_write_summary): Likewise.
3544 (pure_const_read_summary): Likewise.
3545 (propagate_pure_const): Likewise.
3546 (propagate_nothrow): Likewise.
3547 (dump_malloc_lattice): Likewise.
3548 (propagate_malloc): Likewise.
3549 (execute): Do not register hooks, just remove summary
3551 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3554 2018-06-08 Martin Liska <mliska@suse.cz>
3556 * ipa-reference.c (remove_node_data): Remove.
3557 (duplicate_node_data): Likewise.
3558 (class ipa_ref_var_info_summary_t): New class.
3559 (class ipa_ref_opt_summary_t): Likewise.
3560 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3561 (get_reference_optimization_summary): Use
3562 ipa_ref_opt_sum_summaries.
3563 (set_reference_vars_info): Remove.
3564 (set_reference_optimization_summary): Likewise.
3565 (ipa_init): Create summaries.
3566 (init_function_info): Use function summary.
3567 (ipa_ref_opt_summary_t::duplicate): New function.
3568 (ipa_ref_opt_summary_t::remove): New function.
3569 (get_read_write_all_from_node): Fix GNU coding style.
3570 (propagate): Use function summary.
3571 (write_node_summary_p): Fix GNU coding style.
3572 (stream_out_bitmap): Likewise.
3573 (ipa_reference_read_optimization_summary): Use function summary.
3574 (ipa_reference_c_finalize): Do not release hooks.
3576 2018-06-08 Martin Liska <mliska@suse.cz>
3578 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3579 (analyze_function_body): Extract multiple calls of get_create.
3580 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3581 * ipa-inline.c (recursive_inlining): Use ::get method.
3582 * ipa-inline.h (estimate_edge_growth): Likewise.
3584 2018-06-08 Martin Liska <mliska@suse.cz>
3586 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3588 (hsa_function_summary::hsa_function_summary): Use the new enum
3590 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3591 * hsa-gen.c (hsa_get_host_function): Likewise.
3592 (get_brig_function_name): Likewise.
3593 * ipa-hsa.c (process_hsa_functions): Likewise.
3594 (ipa_hsa_write_summary): Likewise.
3595 * symbol-summary.h (symtab_duplication): Use ::get function/
3596 (get): New function.
3598 2018-06-08 Martin Liska <mliska@suse.cz>
3600 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3602 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3603 (hsa_register_kernel): Likewise.
3604 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3605 * hsa-gen.c (hsa_get_host_function): Likewise.
3606 (get_brig_function_name): Likewise.
3607 (generate_hsa): Likewise.
3608 (pass_gen_hsail::execute): Likewise.
3609 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3610 (devirtualization_time_bonus): Likewise.
3611 (ipcp_propagate_stage): Likewise.
3612 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3613 (edge_set_predicate): Likewise.
3614 (evaluate_conditions_for_known_args): Likewise.
3615 (evaluate_properties_for_edge): Likewise.
3616 (ipa_fn_summary::reset): Likewise.
3617 (ipa_fn_summary_t::duplicate): Likewise.
3618 (dump_ipa_call_summary): Likewise.
3619 (ipa_dump_fn_summary): Likewise.
3620 (analyze_function_body): Likewise.
3621 (compute_fn_summary): Likewise.
3622 (estimate_edge_devirt_benefit): Likewise.
3623 (estimate_edge_size_and_time): Likewise.
3624 (estimate_calls_size_and_time): Likewise.
3625 (estimate_node_size_and_time): Likewise.
3626 (inline_update_callee_summaries): Likewise.
3627 (remap_edge_change_prob): Likewise.
3628 (remap_edge_summaries): Likewise.
3629 (ipa_merge_fn_summary_after_inlining): Likewise.
3630 (ipa_update_overall_fn_summary): Likewise.
3631 (read_ipa_call_summary): Likewise.
3632 (inline_read_section): Likewise.
3633 (write_ipa_call_summary): Likewise.
3634 (ipa_fn_summary_write): Likewise.
3635 (ipa_free_fn_summary): Likewise.
3636 * ipa-hsa.c (process_hsa_functions): Likewise.
3637 (ipa_hsa_write_summary): Likewise.
3638 (ipa_hsa_read_section): Likewise.
3639 * ipa-icf.c (sem_function::merge): Likewise.
3640 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3641 (do_estimate_edge_time): Likewise.
3642 (estimate_size_after_inlining): Likewise.
3643 (estimate_growth): Likewise.
3644 (growth_likely_positive): Likewise.
3645 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3646 (inline_call): Likewise.
3647 * ipa-inline.c (caller_growth_limits): Likewise.
3648 (can_inline_edge_p): Likewise.
3649 (can_inline_edge_by_limits_p): Likewise.
3650 (compute_uninlined_call_time): Likewise.
3651 (compute_inlined_call_time): Likewise.
3652 (want_inline_small_function_p): Likewise.
3653 (edge_badness): Likewise.
3654 (update_caller_keys): Likewise.
3655 (update_callee_keys): Likewise.
3656 (recursive_inlining): Likewise.
3657 (inline_small_functions): Likewise.
3658 (inline_to_all_callers_1): Likewise.
3659 (dump_overall_stats): Likewise.
3660 (early_inline_small_functions): Likewise.
3661 (early_inliner): Likewise.
3662 * ipa-inline.h (estimate_edge_growth): Likewise.
3663 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3664 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3665 * ipa-prop.h (IPA_NODE_REF): Likewise.
3666 (IPA_EDGE_REF): Likewise.
3667 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3668 (propagate_malloc): Likewise.
3669 * ipa-split.c (execute_split_functions): Likewise.
3670 * symbol-summary.h: Rename get to get_create.
3672 (get_create): Likewise.
3673 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3675 2018-06-08 Martin Liska <mliska@suse.cz>
3677 * symbol-summary.h (release): Move definition out of class
3679 (symtab_removal): Likewise.
3680 (symtab_duplication): Likewise.
3682 2018-06-08 Martin Liska <mliska@suse.cz>
3684 * symbol-summary.h (function_summary): Move constructor
3685 implementation out of class declaration.
3686 (release): Likewise.
3687 (symtab_insertion): Likewise.
3688 (symtab_removal): Likewise.
3689 (symtab_duplication): Likewise.
3692 2018-06-08 Martin Liska <mliska@suse.cz>
3694 * Makefile.in: Remove support for MPX (macros, related functions,
3695 fields in cgraph_node, ...).
3696 * builtin-types.def (BT_BND): Likewise.
3697 (BT_FN_BND_CONST_PTR): Likewise.
3698 (BT_FN_CONST_PTR_BND): Likewise.
3699 (BT_FN_VOID_PTR_BND): Likewise.
3700 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3701 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3702 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3703 (expand_builtin_mempcpy_with_bounds): Likewise.
3704 (expand_builtin_memset_with_bounds): Likewise.
3705 (expand_builtin_memset_args): Likewise.
3706 (std_expand_builtin_va_start): Likewise.
3707 (expand_builtin): Likewise.
3708 (expand_builtin_with_bounds): Likewise.
3709 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3710 (DEF_LIB_BUILTIN_CHKP): Likewise.
3711 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3712 (DEF_CHKP_BUILTIN): Likewise.
3713 (BUILT_IN_MEMCPY): Likewise.
3714 (BUILT_IN_MEMMOVE): Likewise.
3715 (BUILT_IN_MEMPCPY): Likewise.
3716 (BUILT_IN_MEMSET): Likewise.
3717 (BUILT_IN_STPCPY): Likewise.
3718 (BUILT_IN_STRCAT): Likewise.
3719 (BUILT_IN_STRCHR): Likewise.
3720 (BUILT_IN_STRCPY): Likewise.
3721 (BUILT_IN_STRLEN): Likewise.
3722 (BUILT_IN_MEMCPY_CHK): Likewise.
3723 (BUILT_IN_MEMMOVE_CHK): Likewise.
3724 (BUILT_IN_MEMPCPY_CHK): Likewise.
3725 (BUILT_IN_MEMSET_CHK): Likewise.
3726 (BUILT_IN_STPCPY_CHK): Likewise.
3727 (BUILT_IN_STRCAT_CHK): Likewise.
3728 (BUILT_IN_STRCPY_CHK): Likewise.
3729 * calls.c (store_bounds): Likewise.
3730 (emit_call_1): Likewise.
3731 (special_function_p): Likewise.
3732 (maybe_warn_nonstring_arg): Likewise.
3733 (initialize_argument_information): Likewise.
3734 (finalize_must_preallocate): Likewise.
3735 (compute_argument_addresses): Likewise.
3736 (expand_call): Likewise.
3737 * cfgexpand.c (expand_call_stmt): Likewise.
3738 (expand_return): Likewise.
3739 (expand_gimple_stmt_1): Likewise.
3740 (pass_expand::execute): Likewise.
3741 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3742 (cgraph_node::remove): Likewise.
3743 (cgraph_node::dump): Likewise.
3744 (cgraph_node::verify_node): Likewise.
3745 * cgraph.h (chkp_function_instrumented_p): Likewise.
3746 (symtab_node::get_alias_target): Likewise.
3747 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3748 (cgraph_local_p): Likewise.
3749 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3750 (cgraph_edge::rebuild_references): Likewise.
3751 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3752 (walk_polymorphic_call_targets): Likewise.
3753 (cgraph_node::expand_thunk): Likewise.
3754 (symbol_table::output_weakrefs): Likewise.
3755 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3756 (ix86_handle_option): Likewise.
3757 * config/i386/constraints.md: Likewise.
3758 * config/i386/i386-builtin-types.def (BND): Likewise.
3762 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3763 (BDESC_FIRST): Likewise.
3765 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3766 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3767 * config/i386/i386.c (enum reg_class): Likewise.
3768 (ix86_target_string): Likewise.
3769 (ix86_option_override_internal): Likewise.
3770 (ix86_conditional_register_usage): Likewise.
3771 (ix86_valid_target_attribute_inner_p): Likewise.
3772 (ix86_set_indirect_branch_type): Likewise.
3773 (ix86_set_current_function): Likewise.
3774 (ix86_function_arg_regno_p): Likewise.
3775 (init_cumulative_args): Likewise.
3776 (ix86_function_arg_advance): Likewise.
3777 (ix86_function_arg): Likewise.
3778 (ix86_pass_by_reference): Likewise.
3779 (ix86_function_value_regno_p): Likewise.
3780 (ix86_function_value_1): Likewise.
3781 (ix86_function_value_bounds): Likewise.
3782 (ix86_return_in_memory): Likewise.
3783 (ix86_setup_incoming_vararg_bounds): Likewise.
3784 (ix86_va_start): Likewise.
3785 (indirect_thunk_need_prefix): Likewise.
3786 (print_reg): Likewise.
3787 (ix86_print_operand): Likewise.
3788 (ix86_expand_call): Likewise.
3789 (ix86_output_function_return): Likewise.
3790 (reg_encoded_number): Likewise.
3791 (BDESC_VERIFYS): Likewise.
3792 (ix86_init_mpx_builtins): Likewise.
3793 (ix86_init_builtins): Likewise.
3794 (ix86_emit_cmove): Likewise.
3795 (ix86_emit_move_max): Likewise.
3796 (ix86_expand_builtin): Likewise.
3797 (ix86_builtin_mpx_function): Likewise.
3798 (ix86_get_arg_address_for_bt): Likewise.
3799 (ix86_load_bounds): Likewise.
3800 (ix86_store_bounds): Likewise.
3801 (ix86_load_returned_bounds): Likewise.
3802 (ix86_store_returned_bounds): Likewise.
3803 (ix86_class_likely_spilled_p): Likewise.
3804 (ix86_hard_regno_mode_ok): Likewise.
3805 (x86_order_regs_for_local_alloc): Likewise.
3806 (ix86_mitigate_rop): Likewise.
3807 (ix86_bnd_prefixed_insn_p): Likewise.
3808 (ix86_mpx_bound_mode): Likewise.
3809 (ix86_make_bounds_constant): Likewise.
3810 (ix86_initialize_bounds): Likewise.
3811 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3812 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3813 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3814 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3815 (TARGET_CHKP_BOUND_MODE): Likewise.
3816 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3817 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3818 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3819 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3820 * config/i386/i386.h (TARGET_MPX): Likewise.
3821 (TARGET_MPX_P): Likewise.
3822 (VALID_BND_REG_MODE): Likewise.
3823 (FIRST_BND_REG): Likewise.
3824 (LAST_BND_REG): Likewise.
3825 (enum reg_class): Likewise.
3826 (BND_REG_P): Likewise.
3827 (BND_REGNO_P): Likewise.
3828 (BNDmode): Likewise.
3829 (ADJUST_INSN_LENGTH): Likewise.
3830 * config/i386/i386.md: Likewise.
3831 * config/i386/i386.opt: Likewise.
3832 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3833 (defined): Likewise.
3834 (LINK_MPX): Likewise.
3835 (MPX_SPEC): Likewise.
3836 (LIBMPX_SPEC): Likewise.
3837 (LIBMPXWRAPPERS_SPEC): Likewise.
3838 (CHKP_SPEC): Likewise.
3839 * config/i386/predicates.md: Likewise.
3840 * dbxout.c (dbxout_type): Likewise.
3841 * doc/extend.texi: Likewise.
3842 * doc/invoke.texi: Likewise.
3843 * doc/md.texi: Likewise.
3844 * doc/tm.texi: Likewise.
3845 * doc/tm.texi.in: Likewise.
3846 * dwarf2out.c (is_base_type): Likewise.
3847 (gen_formal_types_die): Likewise.
3848 (gen_subprogram_die): Likewise.
3849 (gen_type_die_with_usage): Likewise.
3850 (gen_decl_die): Likewise.
3851 (dwarf2out_late_global_decl): Likewise.
3852 * expr.c (expand_assignment): Likewise.
3853 (emit_storent_insn): Likewise.
3854 (store_expr_with_bounds): Likewise.
3855 (store_expr): Likewise.
3856 (expand_expr_real_1): Likewise.
3857 * expr.h (store_expr_with_bounds): Likewise.
3858 * function.c (use_register_for_decl): Likewise.
3859 (struct bounds_parm_data): Likewise.
3860 (assign_parms_augmented_arg_list): Likewise.
3861 (assign_parm_find_entry_rtl): Likewise.
3862 (assign_parm_is_stack_parm): Likewise.
3863 (assign_parm_load_bounds): Likewise.
3864 (assign_bounds): Likewise.
3865 (assign_parms): Likewise.
3866 (expand_function_start): Likewise.
3867 * gcc.c (CHKP_SPEC): Likewise.
3868 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3869 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3870 (wrestrict_dom_walker::check_call): Likewise.
3871 * gimple.c (gimple_build_call_from_tree): Likewise.
3872 * gimple.h (enum gf_mask): Likewise.
3873 (gimple_call_with_bounds_p): Likewise.
3874 (gimple_call_set_with_bounds): Likewise.
3875 * gimplify.c (gimplify_init_constructor): Likewise.
3876 * ipa-cp.c (initialize_node_lattices): Likewise.
3877 (propagate_constants_across_call): Likewise.
3878 (find_more_scalar_values_for_callers_subset): Likewise.
3879 * ipa-hsa.c (process_hsa_functions): Likewise.
3880 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3881 * ipa-icf.c (sem_function::merge): Likewise.
3882 * ipa-inline.c (early_inliner): Likewise.
3883 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3884 (warn_function_cold): Likewise.
3885 (propagate_pure_const): Likewise.
3886 * ipa-ref.h (enum GTY): Likewise.
3887 * ipa-split.c (find_retbnd): Likewise.
3888 (consider_split): Likewise.
3889 (split_function): Likewise.
3890 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3891 * ipa.c (walk_polymorphic_call_targets): Likewise.
3892 (symbol_table::remove_unreachable_nodes): Likewise.
3893 (process_references): Likewise.
3894 (cgraph_build_static_cdtor_1): Likewise.
3895 * lto-cgraph.c (lto_output_node): Likewise.
3896 (output_refs): Likewise.
3897 (compute_ltrans_boundary): Likewise.
3898 (input_overwrite_node): Likewise.
3899 (input_node): Likewise.
3900 (input_cgraph_1): Likewise.
3901 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3902 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3903 (class pass_chkp_instrumentation_passes): Likewise.
3904 (make_pass_chkp_instrumentation_passes): Likewise.
3905 * passes.def: Likewise.
3906 * rtl.h (struct GTY): Likewise.
3907 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3908 * stor-layout.c (layout_type): Likewise.
3909 * symtab.c: Likewise.
3910 * target.def: Likewise.
3911 * targhooks.c (default_chkp_bound_type): Likewise.
3912 (default_chkp_bound_mode): Likewise.
3913 (default_builtin_chkp_function): Likewise.
3914 (default_chkp_function_value_bounds): Likewise.
3915 (default_chkp_make_bounds_constant): Likewise.
3916 (default_chkp_initialize_bounds): Likewise.
3917 * targhooks.h (default_chkp_bound_type): Likewise.
3918 (default_chkp_bound_mode): Likewise.
3919 (default_builtin_chkp_function): Likewise.
3920 (default_chkp_function_value_bounds): Likewise.
3921 (default_chkp_make_bounds_constant): Likewise.
3922 (default_chkp_initialize_bounds): Likewise.
3923 * toplev.c (compile_file): Likewise.
3924 (process_options): Likewise.
3925 * tree-core.h (DEF_BUILTIN): Likewise.
3926 (DEF_BUILTIN_CHKP): Likewise.
3927 * tree-inline.c (declare_return_variable): Likewise.
3928 (remap_gimple_stmt): Likewise.
3929 (copy_bb): Likewise.
3930 (initialize_inlined_parameters): Likewise.
3931 (expand_call_inline): Likewise.
3932 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3933 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3934 (make_pass_ipa_chkp_produce_thunks): Likewise.
3935 (make_pass_chkp): Likewise.
3936 (make_pass_chkp_opt): Likewise.
3937 (make_pass_chkp_instrumentation_passes): Likewise.
3938 * tree-pretty-print.c (dump_generic_node): Likewise.
3939 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3940 * tree-ssa-dce.c (propagate_necessity): Likewise.
3941 (eliminate_unnecessary_stmts): Likewise.
3942 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3943 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3944 * tree-ssa-sccvn.h: Likewise.
3945 * tree-ssa-strlen.c (get_string_length): Likewise.
3946 (valid_builtin_call): Likewise.
3947 (adjust_last_stmt): Likewise.
3948 (handle_builtin_strchr): Likewise.
3949 (handle_builtin_strcpy): Likewise.
3950 (handle_builtin_stxncpy): Likewise.
3951 (handle_builtin_memcpy): Likewise.
3952 (handle_builtin_strcat): Likewise.
3953 (strlen_check_and_optimize_stmt): Likewise.
3954 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3955 * tree-streamer-in.c: Likewise.
3956 * tree-streamer.c (record_common_node): Likewise.
3957 * tree.c (tree_code_size): Likewise.
3958 (wide_int_to_tree_1): Likewise.
3959 (type_contains_placeholder_1): Likewise.
3960 (build_common_tree_nodes): Likewise.
3961 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3962 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3963 (POINTER_BOUNDS_P): Likewise.
3964 (BOUNDED_TYPE_P): Likewise.
3965 (BOUNDED_P): Likewise.
3966 (CALL_WITH_BOUNDS_P): Likewise.
3967 (pointer_bounds_type_node): Likewise.
3968 * value-prof.c (gimple_ic): Likewise.
3969 * var-tracking.c (vt_add_function_parameters): Likewise.
3970 * varasm.c (make_decl_rtl): Likewise.
3971 (assemble_start_function): Likewise.
3972 (output_constant): Likewise.
3973 (maybe_assemble_visibility): Likewise.
3974 * varpool.c (ctor_for_folding): Likewise.
3975 * chkp-builtins.def: Remove.
3976 * ipa-chkp.c: Remove.
3977 * ipa-chkp.h: Remove.
3978 * rtl-chkp.c: Remove.
3979 * rtl-chkp.h: Remove.
3980 * tree-chkp-opt.c: Remove.
3981 * tree-chkp.c: Remove.
3982 * tree-chkp.h: Remove.
3984 2018-06-07 Carl Love <cel@us.ibm.com>
3986 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3987 vextract_fp_from_shortl): Add BE support.
3989 2018-06-07 Paul Koning <ni1d@arrl.net>
3991 * compare-elim.c (try_merge_compare): Don't merge compare if
3992 address contains a side effect.
3993 (try_eliminate_compare): Likewise.
3995 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3997 * config.gcc: Support "tremont".
3998 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3999 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4001 * config/i386/i386.c (m_TREMONT): Define.
4002 (processor_target_table): Add "tremont".
4003 (PTA_TREMONT): Define.
4004 (ix86_lea_outperforms): Add TARGET_TREMONT.
4005 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4006 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4007 and M_INTEL_GOLDMONT_PLUS.
4008 (fold_builtin_cpu): Add "tremont".
4009 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4010 (ix86_option_override_internal): Add "tremont".
4011 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4012 (processor_type): Add PROCESSOR_TREMONT.
4013 * config/i386/x86-tune.def: Add m_TREMONT.
4014 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4016 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4018 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4019 symbol defined for msp430i* devices to be lower case.
4021 gcc/testsuite/gcc.target/msp430/
4022 * msp430i-device-symbol.c: New test.
4023 * msp430f-device-symbol.c: New test.
4024 * msp430.h: New test header file.
4026 2018-06-07 Richard Biener <rguenther@suse.de>
4028 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4029 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4030 Properly wrap signed arithmetic if overflow wraps.
4032 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4034 PR tree-optimization/69615
4035 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4036 of a cast from a same precision integral SSA_NAME in a bb dominated
4037 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4038 cast to utype if rhs2 has already a compatible type.
4040 2018-06-07 Richard Biener <rguenther@suse.de>
4042 PR tree-optimization/85935
4043 * graphite-scop-detection.c (find_params_in_bb): Analyze
4044 condition operands with respect to the correct loop. Assert
4045 the analysis doesn't fail.
4047 2018-06-04 Carl Love <cel@us.ibm.com>
4049 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4050 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4051 as it is slightly cheaper.
4052 (first_match_or_eos_index_<mode>):
4053 Calculate index using natural element order.
4054 (first_match_index_<mode>):
4055 Calculate index using natural element order.
4056 (first_match_or_eos_index_<mode>):
4057 Calculate index using natural order.
4058 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4059 for BE and LE modes.
4060 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4061 P9V_BUILTIN_VCLZLSBB_V16QI.
4062 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4065 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4067 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4068 indentation and line wrap for many prototypes. Add missing
4069 @smallexample directives around block of prototypes for vec_xl and
4072 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4074 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4075 track if we pass or return IEEE 128-bit floating point.
4076 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4077 C++ mangling that is compatible with GCC 8.1.
4078 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4079 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4080 floating point types.
4081 (rs6000_function_arg_advance_1): Likewise.
4082 (rs6000_mangle_type): Optionally generate mangled names that match
4083 what GCC 8.1 generated for IEEE 128-bit floating point types.
4084 (rs6000_globalize_decl_name): If we have an external function that
4085 passes or returns IEEE 128-bit types, generate a weak reference
4086 from the mangled name used in GCC 8.1 to the current mangled
4088 (rs6000_init_builtins): Make __ibm128 use the long double type if
4089 long double is IBM extended double. Make __float128 use the long
4090 double type if long double is IEEE 128-bit.
4093 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4094 macro for __ibm128 built-in functions.
4095 (PACK_IF): Add __ibm128 pack/unpack functions.
4096 (UNPACK_IF): Likewise.
4097 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4098 enable long double built-in functions if long double is IEEE
4099 128-bit floating point.
4100 (rs6000_invalid_builtin): Update long double built-in function
4102 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4103 functions, adjust the built-in function to use the long double
4104 built-in function if __ibm128 and long double are the same type.
4105 * doc/extend.texi (PowerPC builtins): Update documention for
4106 __builtin_{,un}pack_longdouble. Add documentation for
4107 __builtin_{,un}pack_ibm128.
4109 2018-06-06 Jim Wilson <jimw@sifive.com>
4111 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4112 (struct machine_function): New field interrupt_mode.
4113 (riscv_handle_type_attribute): New function. Add forward declaration.
4114 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4115 (riscv_expand_epilogue): Check interrupt_mode field.
4116 (riscv_set_current_function): Check interrupt attribute args and
4117 set interrupt_mode field.
4118 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4119 (riscv_sret, riscv_uret): New.
4120 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4121 new arguments to interrupt attribute.
4123 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4126 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4127 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4129 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4131 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4132 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4133 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4134 several redundant entries.
4136 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4138 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4139 type from "rtx" to "rtx_insn *".
4140 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4141 for local "call_insn", removing cast.
4142 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4144 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4146 PR tree-optimization/86066
4147 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4148 for BIT_INSERT_EXPR stores.
4150 2018-06-06 Richard Biener <rguenther@suse.de>
4152 PR tree-optimization/86062
4153 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4154 component refs ontop
4155 of to be offsetted base.
4157 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4159 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4160 to be static and remove check on interrupt attribute name.
4162 gcc/testsuite/gcc.target/msp430/
4163 * function-attributes-4.c: New test.
4164 * static-interrupts.c: New test.
4166 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4168 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4169 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4171 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4174 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4176 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4177 Remove second argument, change how error is called.
4178 (aarch64_layout_arg): Remove second argument from
4179 aarch64_err_no_fpadvsimd call.
4180 (aarch64_init_cumulative_args): Ditto.
4181 (aarch64_gimplify_va_arg_expr): Ditto.
4182 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4184 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4186 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4187 (*simple_return_indirect_internal<mode>): Rename from
4188 simple_return_indirect_internal. Use W mode iterator.
4189 (rstorssp): New expander.
4190 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4191 (clrssbsy): New expander.
4192 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4194 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4196 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4198 (cmse_check_pointed_object): Likewise.
4200 2018-06-05 Martin Liska <mliska@suse.cz>
4202 PR gcov-profile/47618
4203 * doc/invoke.texi: Document how -fprofile-dir format
4206 2018-06-05 Richard Biener <rguenther@suse.de>
4208 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4209 removal pretend DOM info isn't available so we do not update
4210 it and only remove edges, not dominated blocks. Actually free
4211 DOM info in case we removed something. Remove unreachable blocks.
4212 (mfb_keep_latches): Work with either DOM info or marked backedges.
4213 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4214 first. Mark backedges if DOM info isn't available.
4215 (Re-)compute DOM info after cleanup_control_flow_pre.
4217 2018-06-05 Richard Biener <rguenther@suse.de>
4219 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4220 (locus_discrim_hasher::hash): Adjust.
4221 (locus_discrim_hasher::equal): Likewise.
4222 (next_discriminator_for_locus): Work on line directly.
4223 (same_line_p): Pass in expanded locus1 as well.
4224 (assign_discriminators): Avoid redundant location expansions.
4226 2018-06-05 Richard Biener <rguenther@suse.de>
4228 PR tree-optimization/86046
4229 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4230 if required after clearing TREE_ADDRESSABLE.
4232 2018-06-05 Richard Biener <rguenther@suse.de>
4234 PR tree-optimization/86047
4235 * tree-ssa-loop.c (for_each_index): Glob handling of all
4236 decls and constants and really handle all of them.
4238 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4241 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4242 qualifier_void_pointer and qualifier_const_void_pointer.
4243 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4244 (arm_init_builtins): Handle the above.
4245 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4246 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4247 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4250 2018-06-05 Martin Liska <mliska@suse.cz>
4252 * auto-profile.c (read_autofdo_file): Do not use
4253 gcov_ctr_summary struct.
4254 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4255 * coverage.c (struct counts_entry): Likewise.
4256 (read_counts_file): Read just single summary entry.
4257 (get_coverage_counts): Use gcov_summary struct.
4258 * coverage.h (get_coverage_counts): Likewise.
4259 * gcov-dump.c (dump_working_sets): Likewise.
4260 (tag_summary): Dump just single summary.
4261 * gcov-io.c (gcov_write_summary): Write just histogram
4263 (gcov_read_summary): Read just single summary.
4264 (compute_working_sets): Use gcov_summary struct.
4265 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4266 of GCOV_COUNTERS_SUMMABLE.
4267 (GCOV_COUNTERS_SUMMABLE): Remove.
4268 (GCOV_FIRST_VALUE_COUNTER): Replace with
4269 GCOV_COUNTER_V_INTERVAL.
4270 (struct gcov_ctr_summary): Remove.
4271 (struct gcov_summary): Directly use fields of former
4273 (compute_working_sets): Use gcov_summary struct.
4274 * gcov.c (read_count_file): Do not use ctrs fields.
4275 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4277 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4279 * profile.c: Likewise.
4280 * profile.h: Likewise.
4282 2018-06-05 Martin Liska <mliska@suse.cz>
4284 PR gcov-profile/84846
4285 * gcov.c (output_lines): Print working directory only
4286 in intermediate format.
4288 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4290 * config/s390/s390-builtin-types.def: Add void function type.
4291 * config/s390/s390-builtins.def: Use the function type for the
4294 2018-06-04 Jim Wilson <jimw@sifive.com>
4296 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4298 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4299 and maybe_eh_return. Change regno to unsigned int. Use new args to
4300 handle EH_RETURN_DATA_REGNO registers properly.
4301 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4302 (riscv_expand_epilogue): Update comment. Change argument name and
4303 type. Update code to use new name and type. Pass new args to
4304 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4306 * config/riscv/riscv.md (NORMAL_RETURN): New.
4307 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4308 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4309 (eh_return): Call gen_eh_return_internal and emit barrier.
4310 (eh_return_internal): Call riscv_expand_epilogue.
4312 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4314 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4315 bit_insertion field and declare can_be_merged_into method.
4316 (merged_store_group::can_be_merged_into): New method.
4317 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4318 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4319 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4321 2018-06-04 Richard Biener <rguenther@suse.de>
4323 PR tree-optimization/85955
4324 * builtins.c (fold_builtin_sincos): Convert pointers to
4325 destination to appropriate type before dereferencing.
4327 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4329 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4331 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4333 * expr.c (expand_expr_real_1): Force the operand into memory if
4334 its TYPE_MODE is BLKmode and if there is no integer mode for
4335 the number of bits being extracted.
4337 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4341 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4342 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4344 2018-06-04 Richard Biener <rguenther@suse.de>
4346 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4347 (cleanup_tree_cfg_noloop): ... single caller. Do
4348 start_recording_case_labels later.
4350 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4352 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4353 to _IMMINTRIN_H_INCLUDED.
4354 * config/i386/pconfigintrin.h: Ditto.
4355 * config/i386/waitpkgintrin.h: Ditto.
4356 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4357 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4358 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4359 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4360 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4361 waitpkgintrin.h and cldemoteintrin.h.
4363 2018-06-04 Richard Biener <rguenther@suse.de>
4365 PR tree-optimization/86038
4366 * tracer.c (find_best_successor): Check probability for
4367 being initialized, bail out if not.
4369 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4372 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4373 of bits to ignore when comparing architectures.
4375 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4377 PR tree-optimization/69615
4378 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4379 maximum or minimum of the type, try to merge it also as if
4380 range1 is + [-, x - 1] or + [x + 1, -].
4383 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4385 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4387 PR tree-optimization/86034
4388 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4389 the unsigned bitfield type in a bit insertion sequence if it does not
4390 have a larger precision than the bitfield size.
4391 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4393 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4395 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4397 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4399 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4400 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4401 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4402 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4404 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4406 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4407 Disable -fdelete-null-pointer-checks for ELF toolchain.
4409 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4410 Kito Cheng <kito.cheng@gmail.com>
4412 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4413 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4414 (nds32*-*-*): Add glibc and uclibc conditions.
4415 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4416 (TARGET_EXCEPT_UNWIND_INFO): Define.
4417 * config/nds32/elf.h: New file.
4418 * config/nds32/linux.h: New file.
4419 * config/nds32/nds32-elf.opt: New file.
4420 * config/nds32/nds32-linux.opt: New file.
4421 * config/nds32/nds32-fp-as-gp.c
4422 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4423 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4425 (nds32_asm_file_end): Ditto.
4426 (nds32_print_operand): Ditto.
4427 (nds32_insert_attributes): Ditto.
4428 (nds32_init_libfuncs): New function.
4429 (TARGET_HAVE_TLS): Define.
4430 (TARGET_INIT_LIBFUNCS): Define.
4431 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4433 (TARGET_ELF): Apply different mcmodel setting.
4434 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4435 been migrated into elf.h and linux.h files.
4436 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4437 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4438 (mcmodel): The content has been migrated into nds32-elf.opt and
4439 nds32-linux.opt files.
4440 * config/nds32/t-elf: New file.
4441 * config/nds32/t-linux: New file.
4443 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4444 Shiva Chen <shiva0217@gmail.com>
4446 * config/nds32/constants.md (unspec_volatile_element): Add
4447 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4448 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4450 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4451 (make_pass_nds32_fp_as_gp): Declare.
4452 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4454 (nds32_asm_function_end_prologue): Remove unused asm output.
4455 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4456 (nds32_asm_file_start): Output necessary fp_as_gp information.
4457 (nds32_option_override): Adjust register usage.
4458 (nds32_expand_prologue): Consider fp_as_gp situation.
4459 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4460 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4463 (simple_return): Ditto.
4464 (omit_fp_begin): Output special directive for fp_as_gp.
4465 (omit_fp_end): Output special directive for fp_as_gp.
4466 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4467 mforbid-fp-as-gp): New options.
4469 2018-06-01 Mark Wielaard <mark@klomp.org>
4471 * dwarf2out.c (dwarf2out_finish): Remove generation of
4472 DW_AT_loclists_base.
4474 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4476 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4477 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4478 (struct merged_store_group): Add bit_insertion field.
4479 (dump_char_array): Use standard hexadecimal format.
4480 (merged_store_group::merged_store_group): Set bit_insertion to false.
4481 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4482 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4483 also print the mask in the dump file.
4484 (pass_store_merging::gate): Minor tweak.
4485 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4486 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4487 stores with INTEGER_CST stores.
4488 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4489 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4490 and use it throughout. Generate bit insertion sequences if need be.
4491 (pass_store_merging::process_store): Remove redundant condition.
4492 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4494 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4496 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4497 the 128-bit floating point types. Fix function comment.
4499 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4501 * config/aarch64/aarch64-simd.md
4502 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4504 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4507 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4509 PR tree-optimization/85989
4510 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4512 (backprop::intersect_uses): Check it when deciding whether this
4513 is a backedge reference.
4514 (backprop::process_block): Add each phi to m_visited_phis
4515 after visiting it, then clear it at the end.
4517 2018-06-01 Richard Biener <rguenther@suse.de>
4519 * tree-vectorizer.h (vect_dr_stmt): New function.
4520 (vect_get_load_cost): Adjust.
4521 (vect_get_store_cost): Likewise.
4522 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4523 Use vect_dr_stmt instead of DR_SMTT.
4524 (vect_record_base_alignments): Likewise.
4525 (vect_calculate_target_alignment): Likewise.
4526 (vect_compute_data_ref_alignment): Likewise and make static.
4527 (vect_update_misalignment_for_peel): Likewise.
4528 (vect_verify_datarefs_alignment): Likewise.
4529 (vector_alignment_reachable_p): Likewise.
4530 (vect_get_data_access_cost): Likewise. Pass down
4531 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4532 (vect_get_peeling_costs_all_drs): Likewise.
4533 (vect_peeling_hash_get_lowest_cost): Likewise.
4534 (vect_enhance_data_refs_alignment): Likewise.
4535 (vect_find_same_alignment_drs): Likewise.
4536 (vect_analyze_data_refs_alignment): Likewise.
4537 (vect_analyze_group_access_1): Likewise.
4538 (vect_analyze_group_access): Likewise.
4539 (vect_analyze_data_ref_access): Likewise.
4540 (vect_analyze_data_ref_accesses): Likewise.
4541 (vect_vfa_segment_size): Likewise.
4542 (vect_small_gap_p): Likewise.
4543 (vectorizable_with_step_bound_p): Likewise.
4544 (vect_prune_runtime_alias_test_list): Likewise.
4545 (vect_analyze_data_refs): Likewise.
4546 (vect_supportable_dr_alignment): Likewise.
4547 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4548 (vect_gen_prolog_loop_niters): Likewise.
4549 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4550 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4552 (vect_recog_mask_conversion_pattern): Likewise.
4553 (vect_try_gather_scatter_pattern): Likewise.
4554 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4555 to vect_get_store_cost.
4556 (vect_get_store_cost): Get stmt_info instead of DR.
4557 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4558 (vect_get_load_cost): Get stmt_info instead of DR.
4560 2018-06-01 Richard Biener <rguenther@suse.de>
4563 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4564 vars wrapped in MEM_REFs.
4566 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4568 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4569 Fix subreg tests so that we only return a choice between
4570 GENERAL_REGS and FP_REGS if the original classes included both.
4572 2018-06-01 Richard Biener <rguenther@suse.de>
4575 * tree-ssa-structalias.c (get_function_part_constraint):
4576 Handle NULL fi->decl.
4577 (find_func_aliases_for_call): Properly handle indirect
4578 fi from direct call.
4579 (find_func_clobbers): Likewise.
4580 (ipa_pta_execute): Likewise.
4581 (create_variable_info_for): For functions that are ifunc_resolver
4582 resolve to a varinfo that contains the result of the resolver call.
4583 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4586 2018-05-31 Michael Collison <michael.collison@arm.com>
4588 * config/aarch64/aarch64.md:
4589 (*fix_to_zero_extenddfdi2): New pattern.
4590 * gcc.target/aarch64/fix_extend1.c: New testcase.
4592 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4596 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4597 and BUILT_IN_STRNCMP_EQ.
4598 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4599 BUILT_IN_STRNCMP_EQ.
4600 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4601 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4602 (gimple_fold_builtin): Likewise.
4603 * tree-ssa-strlen.c (compute_string_length): New function.
4604 (determine_min_obsize): New function.
4605 (handle_builtin_string_cmp): New function to handle calls to
4606 string compare functions.
4607 (strlen_optimize_stmt): Add handling to builtin string compare
4609 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4610 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4611 * tree.c (build_common_builtin_nodes): Add new defines of
4612 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4614 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4617 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4618 functions with naked attribute.
4620 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4622 * config/i386/sse.md (avx_vec_concat<mode>):
4623 Substitute concat_tg_mode mode attribute with xtg_mode.
4624 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4625 (concat_tg_mode): Remove mode attribute.
4627 2018-05-31 Martin Sebor <msebor@redhat.com>
4630 * calls.c (alloc_max_size): Correct a logic error/typo.
4631 Treat excessive arguments as infinite. Warn for invalid arguments.
4632 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4634 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4637 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4638 and movx for Haswell.
4640 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4641 Cesar Philippidis <cesar@codesourcery.com>
4644 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4645 when emitting error on private/firstprivate reductions.
4646 * omp-low.c (lower_omp_target): Avoid reference-type processing
4647 on pointers for firstprivate clause.
4649 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4651 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4654 * config/aarch64/aarch64-simd.md
4655 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4656 (aarch64_ld1_x3_<mode>): Likewise
4657 (aarch64_st1x2<VALLDIF:mode>): Likewise
4658 (aarch64_st1_x2_<mode>): Likewise
4659 (aarch64_st1x3<VALLDIF:mode>): Likewise
4660 (aarch64_st1_x3_<mode>): Likewise
4661 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4662 (vld1_s8_x3): Likewise.
4663 (vld1_u16_x3): Likewise.
4664 (vld1_s16_x3): Likewise.
4665 (vld1_u32_x3): Likewise.
4666 (vld1_s32_x3): Likewise.
4667 (vld1_u64_x3): Likewise.
4668 (vld1_s64_x3): Likewise.
4669 (vld1_f16_x3): Likewise.
4670 (vld1_f32_x3): Likewise.
4671 (vld1_f64_x3): Likewise.
4672 (vld1_p8_x3): Likewise.
4673 (vld1_p16_x3): Likewise.
4674 (vld1_p64_x3): Likewise.
4675 (vld1q_u8_x3): Likewise.
4676 (vld1q_s8_x3): Likewise.
4677 (vld1q_u16_x3): Likewise.
4678 (vld1q_s16_x3): Likewise.
4679 (vld1q_u32_x3): Likewise.
4680 (vld1q_s32_x3): Likewise.
4681 (vld1q_u64_x3): Likewise.
4682 (vld1q_s64_x3): Likewise.
4683 (vld1q_f16_x3): Likewise.
4684 (vld1q_f32_x3): Likewise.
4685 (vld1q_f64_x3): Likewise.
4686 (vld1q_p8_x3): Likewise.
4687 (vld1q_p16_x3): Likewise.
4688 (vld1q_p64_x3): Likewise.
4689 (vst1_s64_x2): Likewise.
4690 (vst1_u64_x2): Likewise.
4691 (vst1_f64_x2): Likewise.
4692 (vst1_s8_x2): Likewise.
4693 (vst1_p8_x2): Likewise.
4694 (vst1_s16_x2): Likewise.
4695 (vst1_p16_x2): Likewise.
4696 (vst1_s32_x2): Likewise.
4697 (vst1_u8_x2): Likewise.
4698 (vst1_u16_x2): Likewise.
4699 (vst1_u32_x2): Likewise.
4700 (vst1_f16_x2): Likewise.
4701 (vst1_f32_x2): Likewise.
4702 (vst1_p64_x2): Likewise.
4703 (vst1q_s8_x2): Likewise.
4704 (vst1q_p8_x2): Likewise.
4705 (vst1q_s16_x2): Likewise.
4706 (vst1q_p16_x2): Likewise.
4707 (vst1q_s32_x2): Likewise.
4708 (vst1q_s64_x2): Likewise.
4709 (vst1q_u8_x2): Likewise.
4710 (vst1q_u16_x2): Likewise.
4711 (vst1q_u32_x2): Likewise.
4712 (vst1q_u64_x2): Likewise.
4713 (vst1q_f16_x2): Likewise.
4714 (vst1q_f32_x2): Likewise.
4715 (vst1q_f64_x2): Likewise.
4716 (vst1q_p64_x2): Likewise.
4717 (vst1_s64_x3): Likewise.
4718 (vst1_u64_x3): Likewise.
4719 (vst1_f64_x3): Likewise.
4720 (vst1_s8_x3): Likewise.
4721 (vst1_p8_x3): Likewise.
4722 (vst1_s16_x3): Likewise.
4723 (vst1_p16_x3): Likewise.
4724 (vst1_s32_x3): Likewise.
4725 (vst1_u8_x3): Likewise.
4726 (vst1_u16_x3): Likewise.
4727 (vst1_u32_x3): Likewise.
4728 (vst1_f16_x3): Likewise.
4729 (vst1_f32_x3): Likewise.
4730 (vst1_p64_x3): Likewise.
4731 (vst1q_s8_x3): Likewise.
4732 (vst1q_p8_x3): Likewise.
4733 (vst1q_s16_x3): Likewise.
4734 (vst1q_p16_x3): Likewise.
4735 (vst1q_s32_x3): Likewise.
4736 (vst1q_s64_x3): Likewise.
4737 (vst1q_u8_x3): Likewise.
4738 (vst1q_u16_x3): Likewise.
4739 (vst1q_u32_x3): Likewise.
4740 (vst1q_u64_x3): Likewise.
4741 (vst1q_f16_x3): Likewise.
4742 (vst1q_f32_x3): Likewise.
4743 (vst1q_f64_x3): Likewise.
4744 (vst1q_p64_x3): Likewise.
4746 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4748 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4749 user_label_prefix to name.
4751 * tree-core.h: Update comment about the format of NAME string
4752 passed to handler in attribute_spec.
4754 * config/msp430/msp430.md: Remove erroneous subreg expression from
4755 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4756 zero_extend{q,h}isi2.
4758 2018-05-30 Borislav Petkov <bp@suse.de>
4760 * doc/extend.texi: Document some architecture specific
4761 constraints and sort entries.
4763 2018-05-30 Martin Sebor <msebor@redhat.com>
4766 * builtins.c (expand_builtin_stpcpy_1): New function.
4767 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4768 only if the former succeeds.
4770 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4772 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4775 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4777 * doc/invoke.texi (-flinker-output): Document
4779 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4781 * passes.c (ipa_write_summaries): Only modify statements if body
4783 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4784 incrementally linking.
4785 (ipa_passes): Likewise.
4786 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4787 pass down resolution info.
4788 * common.opt (flag_incremental_link): Update info.
4789 * gcc.c (plugin specs): Turn flinker-output=* to
4790 -plugin-opt=-linker-output-known
4791 * toplev.c (compile_file): Also cut compilation when doing incremental
4793 * flag-types. (enum lto_partition_model): Add
4794 LTO_LINKER_OUTPUT_NOLTOREL.
4795 (invoke.texi): Add -flinker-output docs.
4796 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4797 link same way as WPA; do not stream in dead initializers.
4799 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4800 darf2out_register_external_decl): Support incremental link.
4802 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4804 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4806 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4808 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4809 it down to simple_object_copy_lto_debug_sections.
4810 (run_gcc): Determine incremental LTO link time and configure
4811 lto1 into non-wpa mode, disable renaming of debug sections.
4813 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4815 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4816 descriptions of various incorrectly documented functions.
4818 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4821 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4822 address check not strict.
4824 2018-05-30 Richard Biener <rguenther@suse.de>
4826 PR tree-optimization/85964
4827 * tracer.c (better_p): Drop initialized count check, we only
4828 call the function with initialized counts now.
4829 (find_best_successor): Do find a best edge if one
4830 has uninitialized count.
4831 (find_best_predecessor): Likewise. Do BB frequency check only
4832 if count is initialized.
4834 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4836 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4837 (aarch64_ldrstr_offset_compare): New.
4838 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4839 load/store orderings.
4840 (aarch64_gen_adjusted_ldpstp): Likewise.
4842 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4844 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4845 Check for subset of GENERAL_REGS and FP_REGS.
4846 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4849 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4851 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4852 and wi::to_poly_offset. Add the current offset and then check
4853 whether the sum fits, rather than using an unchecked addition of
4854 a checked term. Check for a shwi rather than a uhwi.
4855 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4856 (store_constructor): Use poly_int_tree_p.
4857 (expand_expr_real_1): Likewise.
4858 * function.c (assign_temp): Likewise.
4859 * fold-const.c (const_binop): Use poly_int_tree_p and
4861 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4863 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4864 to_poly_offset for MEM offsets.
4865 * ipa-icf.c (sem_variable::equals): Likewise.
4866 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4867 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4868 wi::to_poly_offset for BIT_FIELD_REF offsets.
4869 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4871 * var-tracking.c (emit_note_insn_var_location): Use
4872 tree_to_poly_uint64.
4874 2018-05-29 Jim Wilson <jimw@sifive.com>
4876 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4878 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4881 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4882 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4884 (sse4_1_round<mode>2): Use nonimmediate_operand
4885 for operand 1 predicate.
4887 2018-05-29 Martin Sebor <msebor@redhat.com>
4888 Richard Biener <rguenther@suse.de>
4891 * calls.c (get_size_range): Call determine_value_range instead
4892 of get_value_range..
4893 * tree-vrp.h (determine_value_range): Declared new function.
4894 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4896 2018-05-29 Richard Biener <rguenther@suse.de>
4898 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4899 sure to use non-pattern stmts for get_earlier_stmt arguments.
4900 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4901 called on pattern stmts.
4902 (get_later_stmt): Likewise.
4904 2018-05-29 Martin Liska <mliska@suse.cz>
4906 PR gcov-profile/85759
4907 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4910 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4912 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4914 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4918 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4919 VEC_PACK_FLOAT_EXPR): New tree codes.
4920 * tree-pretty-print.c (op_code_prio): Handle
4921 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4922 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4923 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4924 * tree-inline.c (estimate_operator_cost): Likewise.
4925 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4926 * fold-const.c (const_binop): Likewise.
4927 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4928 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4929 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4930 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4931 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4932 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4933 * expr.c (expand_expr_real_2): Likewise.
4934 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4935 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4936 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4938 * optabs.c (expand_widen_pattern_expr): For
4939 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4940 sign from result type rather than operand's type.
4941 (expand_binop_directly): For vec_packu_float_optab and
4942 vec_packs_float_optab allow result type to be different from operand's
4944 * optabs-tree.c (optab_for_tree_code): Handle
4945 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4946 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4947 * tree-vect-generic.c (expand_vector_operations_1): Handle
4948 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4949 VEC_PACK_FLOAT_EXPR.
4950 * tree-vect-stmts.c (supportable_widening_operation): Handle
4952 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4953 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4954 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4955 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4956 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4958 (vec_pack<floatprefix>_float_<mode>): New expander.
4959 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4961 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4962 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4963 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4964 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4965 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4967 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4968 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4969 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4970 VEC_PACK_FLOAT_EXPR): Document.
4972 2018-05-29 Richard Biener <rguenther@suse.de>
4974 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4976 (stmt_vec_info_vec): Make pointer.
4977 (init_stmt_vec_info_vec): Remove.
4978 (free_stmt_vec_info_vec): Likewise.
4979 (set_stmt_vec_info_vec): New function.
4980 (free_stmt_vec_infos): Likewise.
4981 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4982 (set_vinfo_for_stmt): Likewise.
4983 (get_earlier_stmt): Likewise.
4984 (get_later_stmt): Likewise.
4985 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4986 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4987 (vec_info::~vec_info): Free stmt_vec_infos.
4988 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4989 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4990 (pass_slp_vectorize::execute): Likewise.
4991 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4992 (free_stmt_vec_info_vec): Likewise.
4993 (set_stmt_vec_info_vec): New function.
4994 (free_stmt_vec_infos): Likewise.
4995 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4996 the global stmt_vec_info_vec.
4997 * tree-parloops.c (gather_scalar_reductions): Use
4998 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5001 2018-05-29 Richard Biener <rguenther@suse.de>
5003 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5005 2018-05-29 Martin Liska <mliska@suse.cz>
5006 David Malcolm <dmalcolm@redhat.com>
5008 * vec.c (test_reverse): New.
5009 (vec_c_tests): Add new test.
5010 * vec.h (vl_ptr>::reverse): New function.
5012 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5014 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5016 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5019 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5021 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5023 2018-05-28 Richard Biener <rguenther@suse.de>
5025 PR tree-optimization/85933
5026 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5027 look at stmts marked as vectorizable.
5029 2018-05-28 Richard Biener <rguenther@suse.de>
5031 PR tree-optimization/85934
5032 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5033 vector boolean check before scalar optimization.
5035 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5037 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5040 2018-05-28 Mark Wielaard <mark@klomp.org>
5042 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5043 if it is an expression containing a minus sign.
5045 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5047 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5049 2018-05-27 Paul Koning <ni1d@arrl.net>
5051 * config/pdp11/pdp11.md (truncsihi2): Remove.
5053 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5054 Chung-Ju Wu <jasonwucj@gmail.com>
5056 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5058 (unaligned_store_dw): Ditto.
5059 * config/nds32/nds32-memory-manipulation.c
5060 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5061 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5062 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5063 (emit_setmem_word_loop): Rename to ...
5064 (emit_setmem_doubleword_loop): ... this.
5065 (nds32_gen_dup_4_byte_to_word_value): New function.
5066 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5067 (nds32_expand_setmem_loop): Refine implementation.
5068 (nds32_expand_setmem_loop_v3m): Ditto.
5069 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5072 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5074 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5076 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5078 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5079 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5080 machine->attr_no_prologue_p.
5081 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5082 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5083 (nds32_expand_epilogue): Consider attr_naked_p.
5084 (nds32_expand_epilogue_v3pop): Likewise.
5085 (nds32_can_use_return_insn): Likewise.
5086 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5087 attr_no_prologue_p fields.
5088 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5090 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5093 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5095 * config/i386/sse.md
5096 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5098 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5100 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5102 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5104 (*<floatsuffix>floatv2div2sf2): Rename to ...
5105 (*float<floatunssuffix>v2div2sf2): ... this.
5106 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5107 (float<floatunssuffix>v2div2sf2_mask): ... this.
5108 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5109 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5110 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5112 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5114 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5116 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5118 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5120 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5122 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5123 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5124 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5125 gen_ufix_truncv8dfv8si2.
5126 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5127 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5128 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5129 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5130 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5131 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5132 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5133 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5135 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5139 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5141 2018-05-25 Jim Wilson <jimw@sifive.com>
5143 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5144 * config/riscv/riscv.c (struct machine_function): Add
5145 interrupt_handler_p and attribute_checked_p fields.
5146 (riscv_attribute_table): Add interrupt.
5147 (riscv_interrupt_type_p): New.
5148 (riscv_save_reg_p): Save extra regs for interrupt handler.
5149 (riscv_use_save_libcall): Return false for interrupt handler.
5150 (riscv_first_stack_step): Add forward declaration.
5151 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5152 for interrupt handler with large frame. Use it for saved reg list.
5153 (riscv_expand_prologue): Move flag_stack_usage_info support to
5154 eliminate duplication.
5155 (riscv_expand_epilogue): Generate mret for interrupt handler.
5156 (riscv_epilogue_uses): New.
5157 (riscv_can_use_return_insn): Return false for interrupt handler.
5158 (riscv_function_ok_for_sibcall): Likewise.
5159 (riscv_set_current_function): Add interrupt handler support.
5160 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5161 * config/riscv/riscv.md (UNSPECV_MRET): New.
5163 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5165 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5167 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5169 PR tree-optimization/85712
5170 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5171 this candidate has already been replaced in-situ by a copy.
5173 2018-05-25 Jason Merrill <jason@redhat.com>
5175 PR c++/80485 - inline function non-zero address.
5176 * symtab.c (nonzero_address): Check DECL_COMDAT.
5178 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5181 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5182 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5183 extension of SImode operation. Use const123_operand predicate.
5184 (*saddsi_1): Remove.
5185 (*saddl_se_1): Ditto.
5187 (*ssubl_se_1): Ditto.
5188 * config/alpha/predicates.md (const123_operand): New predicate.
5189 * config/alpha/constraints.md (P): Use IN_RANGE.
5191 2018-05-25 Richard Biener <rguenther@suse.de>
5193 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5195 (ref_maybe_used_by_stmt_p): Likewise.
5196 (stmt_may_clobber_ref_p): Likewise.
5197 (stmt_may_clobber_ref_p_1): Likewise.
5198 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5200 (ref_maybe_used_by_stmt_p): Likewise.
5201 (stmt_may_clobber_ref_p): Likewise.
5202 (stmt_may_clobber_ref_p_1): Likewise.
5203 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5204 the alias oracle to disambiguate DRs with stmts DR analysis
5206 (vect_analyze_data_refs): Do not give up on not analyzable
5207 DRs for BB vectorization. Remove code truncating the dataref
5210 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5213 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5214 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5215 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5217 2018-05-25 Richard Biener <rguenther@suse.de>
5219 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5220 function, combining stmt data ref gathering and fatal analysis
5222 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5223 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5224 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5225 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5226 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5228 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5230 PR tree-optimization/85720
5231 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5232 SCC if all partitions are builtins.
5233 (version_loop_by_alias_check): New parameter. Generate cancelable
5234 runtime alias check if all partitions are builtins.
5235 (distribute_loop): Update call to above function.
5237 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5239 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5240 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5241 (parm_default_def_partition_arg): Ditto.
5242 (set_parm_default_def_partition): Ditto.
5243 (get_parm_default_def_partitions): Ditto and make it static.
5244 (get_undefined_value_partitions): Ditto and make it static.
5245 (remove_ssa_form): Refactor call to init_var_map here.
5246 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5247 computation for loop region.
5248 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5249 (register_default_def): Delete.
5250 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5251 (parm_default_def_partition_arg): Ditto.
5252 (set_parm_default_def_partition): Ditto.
5253 (get_parm_default_def_partitions): Ditto and make it static.
5254 (get_undefined_value_partitions): Ditto and make it static.
5255 (coalesce_with_default, coalesce_with_default): Update comment.
5256 (create_coalesce_list_for_region): New func factored out from
5257 create_outofssa_var_map.
5258 (populate_coalesce_list_for_outofssa): New func factored out from
5259 create_outofssa_var_map and coalesce_ssa_name.
5260 (create_outofssa_var_map): Delete.
5261 (coalesce_ssa_name): Refactor to support live range computation.
5262 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5263 (get_parm_default_def_partitions): Delete.
5264 (get_undefined_value_partitions): Ditto.
5265 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5266 computation for loop region.
5267 (new_tree_live_info, loe_visit_block): Ditto.
5268 (live_worklist, set_var_live_on_entry): Ditto.
5269 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5270 * tree-ssa-live.h (struct _var_map): New fields.
5271 (init_var_map): Change decl.
5272 (region_contains_p): New.
5274 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5276 * tree-ssa-live.h (live_merge_and_clear): Delete.
5278 2018-05-25 Richard Biener <rguenther@suse.de>
5281 * tree-dump.c (dequeue_and_dump): Remove access to removed
5282 operand 2 of a SWITCH_EXPR.
5284 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5286 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5287 multiplication and division.
5288 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5289 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5290 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5291 (cond_udiv_optab, cond_umod_optab): New optabs.
5292 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5293 (IFN_COND_RDIV): New internal functions.
5294 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5295 TRUNC_MOD_EXPR and RDIV_EXPR.
5296 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5297 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5299 (SVE_INT_BINARY): Include mult.
5300 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5301 (optab, sve_int_op): Handle mult.
5302 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5304 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5305 for SVE_INT_BINARY_SD.
5307 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5309 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5310 (optab, sve_int_op): Handle div and udiv.
5311 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5312 for SVE_INT_BINARY_SD.
5313 (*<optab><mode>3): New insn for the same.
5315 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5317 * tree-vect-patterns.c: Include predict.h.
5318 (vect_recog_divmod_pattern): Restrict check for division support
5319 to when optimizing for size.
5321 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5323 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5324 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5325 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5326 (gimple_match_op::set_op): Likewise.
5327 (gimple_resimplify4): Declare.
5328 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5329 (expr::gen_transform): Likewise.
5330 (decision_tree::gen): Generate a simplification routine for 4 operands.
5331 * gimple-match-head.c (gimple_simplify): Add an overload for
5332 4 operands. In the top-level function, handle up to 4 call
5333 arguments and call gimple_resimplify4.
5334 (gimple_resimplify4): New function.
5335 (build_call_internal): Pass a fourth operand.
5336 (maybe_push_to_seq): Likewise.
5337 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5338 Fold VEC_COND_EXPRs of an operation and a default value into
5339 an IFN_COND_* function if possible.
5340 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5342 (SVE_COND_FP_BINARY): Include them.
5343 (optab, sve_fp_op): Handle them.
5344 (SVE_INT_BINARY_REV): New code iterator.
5345 (SVE_COND_FP_BINARY_REV): New int iterator.
5346 (commutative): New int attribute.
5347 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5349 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5351 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5352 (*cond_<optab><mode>): New patterns for reversed operands.
5354 2018-05-25 Richard Biener <rguenther@suse.de>
5356 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5357 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5358 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5359 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5360 * tree-vect-data-refs.c (everywhere): Adjust users.
5361 * tree-vect-loop.c (everywhere): Likewise.
5362 * tree-vect-slp.c (everywhere): Likewise.
5363 * tree-vect-stmts.c (everywhere): Likewise.
5364 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5366 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5368 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5370 (gcc_cv_as_section_exclude): ... this.
5371 Try Solaris as #exclude syntax.
5372 * configure: Regenerate.
5373 * config.in: Regenerate.
5374 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5376 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5377 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5379 * varasm.c (default_elf_asm_named_section): Don't check if
5380 HAVE_GAS_SECTION_EXCLUDE is defined.
5382 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5384 * doc/md.texi: Update the documentation of the cond_* optabs
5385 to mention the new final operand. Fix GET_MODE_NUNITS call.
5386 Describe the scalar case too.
5387 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5388 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5390 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5391 (get_conditional_internal_fn): Update comment.
5392 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5393 accumulator value as a final argument to conditional functions.
5394 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5395 a define_expand and add an "else" operand. Assert for now that
5396 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5397 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5398 (*cond_<optab><mode>): New patterns.
5399 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5400 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5401 (UNSPEC_COND_EOR): Delete.
5402 (optab): Remove associated mappings.
5403 (SVE_INT_BINARY): New code iterator.
5404 (sve_int_op): Remove int attribute and add "minus" to the code
5406 (SVE_COND_INT_OP): Delete.
5407 (SVE_COND_FP_OP): Rename to...
5408 (SVE_COND_FP_BINARY): ...this.
5410 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5412 * optabs.c (can_reuse_operands_p): New function.
5413 (maybe_legitimize_operands): Try to reuse the results for
5416 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5418 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5419 Add {q} suffix to insn mnemonic.
5421 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5423 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5424 (msp430_warn_func_return): New.
5426 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5428 * fold-const.c (tree_nonzero_bits): New function.
5429 * fold-const.h (tree_nonzero_bits): Likewise.
5430 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5431 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5433 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5437 * varasm.c (assemble_alias): Check ifunc_resolver only on
5440 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5443 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5444 when memory input operand is handled.
5446 2018-05-24 Luis Machado <luis.machado@linaro.org>
5448 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5450 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5452 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5454 * match.pd: Delay FMA folds until after vectorization.
5456 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5459 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5460 address check not strict.
5462 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5464 * gimple-match.h (gimple_match_op): New class.
5465 (mprts_hook): Replace parameters with a gimple_match_op *.
5466 (maybe_build_generic_op): Likewise.
5467 (gimple_simplified_result_is_gimple_val): Replace parameters with
5468 a const gimple_match_op *.
5469 (gimple_simplify): Replace code_helper * and tree * parameters with
5470 a gimple_match_op * parameter.
5471 (gimple_resimplify1): Replace code_helper *, tree and tree *
5472 parameters with a gimple_match_op * parameter.
5473 (gimple_resimplify2): Likewise.
5474 (gimple_resimplify3): Likewise.
5475 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5476 parameters with a gimple_match_op * parameter.
5477 * gimple-match-head.c (gimple_simplify): Change prototypes of
5478 auto-generated functions to take a gimple_match_op * instead of
5479 separate code_helper * and tree * parameters. Make the same
5480 change in the top-level overload and update calls to the
5481 gimple_resimplify routines. Update calls to the auto-generated
5482 functions and to maybe_push_res_to_seq in the publicly-facing
5483 operation-specific gimple_simplify overloads.
5484 (gimple_match_op::MAX_NUM_OPS): Define.
5485 (gimple_resimplify1): Replace rcode and ops with a single res_op
5486 parameter. Update call to gimple_simplify.
5487 (gimple_resimplify2): Likewise.
5488 (gimple_resimplify3): Likewise.
5489 (mprts_hook): Replace parameters with a gimple_match_op *.
5490 (maybe_build_generic_op): Likewise.
5491 (build_call_internal): Replace type, nargs and ops with
5492 a gimple_match_op *.
5493 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5494 with a single gimple_match_op *. Update calls to mprts_hook,
5495 build_call_internal and gimple_simplified_result_is_gimple_val.
5496 Factor out code that is common to the tree_code and combined_fn cases.
5497 * genmatch.c (expr::gen_transform): Replace tem_code and
5498 tem_ops with a gimple_match_op called tem_op. Update calls
5499 to the gimple_resimplify functions and maybe_push_res_to_seq.
5500 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5501 res_ops. Update call to the gimple_resimplify functions.
5502 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5503 (decision_tree::gen): Make the functions take a gimple_match_op *
5504 called res_op instead of separate res_code and res_ops parameters.
5505 Update call accordingly.
5506 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5507 and ops with a single res_op parameter. Update calls to
5508 maybe_build_generic_op and maybe_push_res_to_seq.
5509 (fold_stmt_1): Update calls to gimple_simplify and
5510 replace_stmt_with_simplification.
5511 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5512 and gimple_simplified_result_is_gimple_val.
5513 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5515 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5516 with a gimple_match_op *.
5517 (vn_nary_build_or_lookup): Likewise. Update call to
5518 vn_nary_build_or_lookup_1.
5519 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5520 gimple_match_op *. Update calls to the gimple_resimplify routines
5521 and to gimple_simplified_result_is_gimple_val.
5522 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5523 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5524 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5525 (visit_nary_op): Likewise.
5526 (visit_reference_op_load): Likewise.
5528 2018-05-23 Luis Machado <luis.machado@linaro.org>
5530 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5531 modifier for printing the step amount.
5533 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5536 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5539 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5541 * doc/sourcebuild.texi (Endianness): New subsubsection.
5543 2018-05-23 Luis Machado <luis.machado@linaro.org>
5545 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5546 <prefetch_dynamic_strides>: New const bool field.
5547 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5548 prefetch_dynamic_strides.
5549 (exynosm1_prefetch_tune): Likewise.
5550 (thunderxt88_prefetch_tune): Likewise.
5551 (thunderx_prefetch_tune): Likewise.
5552 (thunderx2t99_prefetch_tune): Likewise.
5553 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5555 (aarch64_override_options_internal): Update to set
5556 PARAM_PREFETCH_DYNAMIC_STRIDES.
5557 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5558 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5559 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5560 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5561 prefetch-dynamic-strides setting.
5563 2018-05-23 Luis Machado <luis.machado@linaro.org>
5565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5566 <minimum_stride>: New const int field.
5567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5568 minimum_stride field defaulting to -1.
5569 (exynosm1_prefetch_tune): Likewise.
5570 (thunderxt88_prefetch_tune): Likewise.
5571 (thunderx_prefetch_tune): Likewise.
5572 (thunderx2t99_prefetch_tune): Likewise.
5573 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5574 <default_opt_level>: Set to 3.
5575 (aarch64_override_options_internal): Update to set
5576 PARAM_PREFETCH_MINIMUM_STRIDE.
5577 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5578 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5579 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5580 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5581 stride is constant and is below the minimum stride threshold.
5583 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5585 * config/arm/arm-cpus.in (mode26): Delete.
5586 (armv4): Delete mode26 reference.
5587 * config/arm/arm.c (arm_configure_build_target): Delete use of
5590 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5592 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5594 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5595 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5597 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5598 (floatunsdidf2): Ditto.
5600 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5602 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5603 (fixuns_trunc<mode>si2_avx512f): Ditto.
5604 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5605 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5606 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5608 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5610 PR rtl-optimization/79985
5611 * df-scan.c (df_insn_refs_collect): Remove special case for
5612 global registers and asm statements.
5614 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5616 * extend.texi (Global Register Variables): Rewrite the bullet list.
5617 Note that the register is available for allocation. Note that access
5618 via inline asm must use constraints. Add note about async-signal
5619 handlers. Remove paragraph about automagic register selection.
5621 2018-05-23 Richard Biener <rguenther@suse.de>
5623 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5624 of fixed offset from memset VN.
5626 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5628 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5630 (alloc_cand_and_find_basis): Initialize first_interp field.
5631 (slsr_process_mul): Modify first_interp field.
5632 (slsr_process_add): Likewise.
5633 (slsr_process_cast): Modify first_interp field for each new
5635 (slsr_process_copy): Likewise.
5636 (dump_candidate): Dump first_interp field.
5637 (replace_mult_candidate): Process all interpretations, not just
5639 (replace_rhs_if_not_dup): Likewise.
5640 (replace_one_candidate): Likewise.
5642 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5644 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5646 (aarch64_needs_frame_chain): New function.
5647 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5649 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5652 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5653 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5654 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5655 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5656 as true for strict-align.
5657 (aarch64_can_inline_p): Perform checks even when callee has no
5658 attributes to check for strict alignment.
5659 * doc/extend.texi (AArch64 Function Attributes): Document
5661 * doc/invoke.texi: (AArch64 Options): Likewise.
5663 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5665 PR tree-optimization/85853
5666 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5667 the handling of the root of the node to...
5668 (vect_slp_analyze_node_operations_1): ...this new function,
5669 and run the whole thing with the child nodes' def types
5670 set according to their SLP node's def type.
5672 2018-05-23 Richard Biener <rguenther@suse.de>
5675 * tree-data-ref.c (create_runtime_alias_checks): Defer
5676 and ignore overflow warnings.
5678 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5680 PR tree-optimization/85822
5681 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5684 2018-05-23 Richard Biener <rguenther@suse.de>
5686 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5687 memset constants via native_interpret_expr.
5689 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5692 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5694 (cgraph_node::create_alias): Likewise.
5695 (cgraph_node::get_availability): Check ifunc_resolver instead
5696 of looking up ifunc attribute.
5697 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5698 * varasm.c (do_assemble_alias): Likewise.
5699 (assemble_alias): Likewise.
5700 (default_binds_local_p_3): Likewise.
5701 * cgraph.h (cgraph_node): Add ifunc_resolver.
5702 (cgraph_node::only_called_directly_or_aliased_p): Return false
5704 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5706 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5707 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5708 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5709 instead of looking up ifunc attribute.
5711 2018-05-22 Luis Machado <luis.machado@linaro.org>
5713 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5715 2018-05-22 Martin Sebor <msebor@redhat.com>
5718 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5719 only when expasion succeeds.
5720 (expand_builtin_strcmp): Same.
5721 (expand_builtin_strncmp): Same.
5723 2018-05-22 Martin Sebor <msebor@redhat.com>
5725 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5727 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5728 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5730 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5731 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5733 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5734 Add check for legitimate_address.
5735 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5736 (aarch64_swap_ldrstr_operands): New.
5737 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5740 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5741 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5743 * config/aarch64/aarch64.md: New patterns to generate stp
5745 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5746 single words and double words.
5747 (load_pair_sw, load_pair_dw): Likewise.
5748 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5750 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5752 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5753 for different mode ldpstp and add peephole for merged zero stores.
5755 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5757 (aarch64_gen_store_pair): Rename calls to match new patterns.
5758 (aarch64_gen_load_pair): Rename calls to match new patterns.
5759 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5760 (load_pair<DREG:mode><DREG2:mode>): ... This.
5761 (store_pair<mode>): Rename to...
5762 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5763 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5765 (V_INT_EQUIV): Handle SImode.
5766 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5769 2018-05-22 Martin Sebor <msebor@redhat.com>
5772 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5773 or ajust the presumed bound on an operation to avoid unnecessary
5776 2018-05-22 Martin Sebor <msebor@redhat.com>
5778 PR tree-optimization/85826
5779 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5780 assuming that a DECL necesarily has a constant size.
5782 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5785 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5787 2018-05-22 Richard Biener <rguenther@suse.de>
5789 PR tree-optimization/85834
5790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5791 non-constant and non-zero memset arguments.
5793 2018-05-22 Martin Liska <mliska@suse.cz>
5796 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5798 2018-05-22 Richard Biener <rguenther@suse.de>
5800 PR tree-optimization/85863
5801 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5802 comparisons when vectype is specified.
5803 (vectorizable_condition): Do not specify vectype for
5804 vect_is_simple_cond when SLP vectorizing.
5806 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5809 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5810 define __ibm128 as long double.
5811 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5812 as a distinct type when IEEE 128-bit support is enabled.
5813 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5814 128-bit types to use the correct functions.
5815 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5816 convert between 128-bit floating point types that have different
5817 modes but the same representation, instead of using gen_lowpart to
5819 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5821 (IFKF_reg): New attributes to give the register constraints for
5823 (extend<mode>tf2_internal): New insns to mark an explicit
5824 conversion between 128-bit floating point types that have a
5825 different mode but share the same representation.
5827 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5829 PR tree-optimization/85814
5830 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5831 a null return from get_strinfo when unsharing the next
5832 strinfo in the chain.
5834 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5837 * varasm.c (weak_finish): Clean up weak_decls.
5839 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5841 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5842 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5843 UNSPEC_UADALP values.
5844 * config/aarch64/iterators.md (ABAL): New int iterator.
5847 (sur): Add mappings for the above.
5848 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5850 (aarch64_<sur>abal<mode>_4): Likewise.
5851 (aarch64_<sur>adalp<mode>_3): Likewise.
5852 (<sur>sadv16qi): New define_expand.
5854 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5856 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5857 (*movdf_internal): Ditto.
5858 (*rcpsf2_sse): Ditto.
5859 (*rsqrtsf2_sse): Ditto.
5860 (*sqrt<mode>2_sse): Ditto.
5862 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5864 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5866 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5867 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5868 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5869 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5871 * config/aarch64/arm_neon.h: Likewise.
5872 * config/aarch64/iterators.md (VQ_I): New.
5874 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5876 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5878 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5881 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5883 * config/nds32/constraints.md (S): New constraint.
5884 * config/nds32/nds32.md (call_internal): Use constraint S.
5885 (call_value_internal): Likewise.
5886 (sibcall_internal): Likewise.
5887 (sibcall_value_internal): Likewise.
5889 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5890 Chung-Ju Wu <jasonwucj@gmail.com>
5892 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5895 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5896 Chung-Ju Wu <jasonwucj@gmail.com>
5898 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5899 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5900 (nds32_rtx_costs_impl): Simplify.
5901 (nds32_address_cost_impl): Simplify.
5902 (nds32_init_rtx_costs): New function.
5903 (nds32_rtx_costs_speed_prefer): Likewise.
5904 (nds32_rtx_costs_size_prefer): Likewise.
5905 (nds32_address_cost_speed_prefer): Likewise.
5906 (nds32_address_cost_speed_fwprop): Likewise.
5907 (nds32_address_cost_size_prefer): Likewise.
5908 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5909 * config/nds32/nds32.c (nds32_option_override): Use
5910 nds32_init_rtx_costs function.
5912 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5914 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5915 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5916 (TARGET_PIPELINE_N8): Likewise.
5917 (TARGET_PIPELINE_N10): Likewise.
5918 (TARGET_PIPELINE_N13): Likewise.
5919 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5921 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5923 * config/nds32/nds32-fpu.md: Update copyright year.
5925 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5927 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5929 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5931 * config/nds32/nds32.c
5932 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5933 * config/nds32/nds32.opt (minline-asm-r15): New option.
5935 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5937 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5939 * config/nds32/nds32.md (abssi2): New pattern.
5941 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5943 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5944 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5945 Merge insn pattern from sse_cvtsi2ss<round_name> and
5946 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5947 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5948 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5949 using SWI48 mode iterator.
5950 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5951 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5952 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5953 pattern from sse_cvttss2si<round_saeonly_name>
5954 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5955 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5956 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5957 using SWI48 mode iterator.
5958 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5959 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5960 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5961 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5962 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5963 using SWI48 mode iterator.
5964 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5965 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5966 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5967 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5968 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5969 SWI48 mode iterator.
5970 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5971 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5972 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5973 pattern from sse_cvttsd2si<round_saeonly_name>
5974 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5976 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5978 * config/nds32/nds32-md-auxiliary.c
5979 (nds32_valid_smw_lwm_base_p): Refine.
5980 (nds32_output_smw_single_word): Refine.
5981 (nds32_output_smw_double_word): New.
5982 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5984 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5986 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5987 (nds32_output_stack_pop): Refine.
5988 (nds32_expand_unaligned_load): Refine.
5989 (nds32_expand_unaligned_store): Refine.
5991 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5992 Chung-Ju Wu <jasonwucj@gmail.com>
5994 * config/nds32/constants.md: Add TP_REGNUM constant.
5995 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5996 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5998 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5999 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6000 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6001 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6002 and PIC code generation.
6003 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6005 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6007 * config/nds32/nds32.md: Support TLS and PIC.
6008 * config/nds32/nds32.c: Support TLS and PIC.
6009 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6010 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6013 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6015 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6016 mode with E_ prefix.
6018 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6019 Chung-Ju Wu <jasonwucj@gmail.com>
6021 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6022 * config/nds32/nds32-md-auxiliary.c
6023 (symbolic_reference_mentioned_p): New.
6024 (nds32_legitimize_ict_address): New.
6025 (nds32_expand_ict_move): New.
6026 (nds32_indirect_call_referenced_p): New.
6027 (nds32_symbol_binds_local_p): Delete.
6028 (nds32_long_call_p): Modify.
6029 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6030 * config/nds32/nds32-protos.h
6031 (symbolic_reference_mentioned_p): Declare.
6032 (nds32_legitimize_ict_address): Declare.
6033 (nds32_expand_ict_move): Declare.
6034 (nds32_indirect_call_referenced_p): Declare.
6035 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6036 (nds32_relax_group): Use nds32_ict_const_p as condition.
6037 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6038 (nds32_asm_file_start): Output ict_model directive in asm code.
6039 (nds32_legitimate_address_p): Consider indirect call.
6040 (nds32_print_operand): Consider indirect call.
6041 (nds32_print_operand_address): Consider indirect call.
6042 (nds32_insert_attributes): Handle "indirect_call" attribute.
6043 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6044 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6045 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6046 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6047 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6048 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6049 (TARGET_ICT_MODEL_SMALL): Define.
6050 (TARGET_ICT_MODEL_LARGE): Define.
6051 * config/nds32/nds32.md (movsi): Consider ict model.
6052 (call, call_value): Consider ict model.
6053 (sibcall, sibcall_value): Consider ict model.
6054 * config/nds32/nds32.opt (mict-model): New option.
6055 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6058 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6059 Monk Chiang <sh.chiang04@gmail.com>
6060 Jim Wilson <jimw@sifive.com>
6062 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6063 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6064 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6065 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6066 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6067 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6068 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6069 compute save_libcall_adjustment properly.
6070 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6071 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6072 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6073 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6074 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6075 (ABI_SPEC): Handle mabi=ilp32e.
6076 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6077 (RVE): Add RVE mask.
6078 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6079 <-march>: Add rv32e as an example.
6081 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6084 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6085 (intra_create_variable_infos): Handle C++ constructors.
6087 2018-05-18 Martin Liska <mliska@suse.cz>
6089 * passes.def: Remove a redundant pass.
6091 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6094 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6096 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6098 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6100 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6102 (armv2, armv2a, armv3, armv3m): Delete architectures.
6103 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6104 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6105 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6107 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6108 (*mulsidi3adddi): Likewise.
6109 (mulsidi3): Likewise.
6110 (*mulsidi3_nov6): Likewise.
6111 (umulsidi3): Likewise.
6112 (umulsidi3_nov6): Likewise.
6113 (umaddsidi4): Likewise.
6114 (*umulsidi3adddi): Likewise.
6115 (smulsi3_highpart): Likewise.
6116 (*smulsi3_highpart_nov6): Likewise.
6117 (umulsi3_highpart): Likewise.
6118 (*umulsi3_highpart_nov6): Likewise.
6119 * config/arm/arm.h (arm_arch3m): Delete.
6120 * config/arm/arm.c (arm_arch3m): Delete.
6121 (arm_option_override_internal): Update armv3-related comment.
6122 (arm_configure_build_target): Delete use of isa_bit_mode32.
6123 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6124 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6125 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6127 (mulusa3): Likewise.
6128 * config/arm/arm-protos.h (arm_arch3m): Delete.
6129 * config/arm/arm-tables.opt: Regenerate.
6130 * config/arm/arm-tune.md: Likewise.
6131 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6132 deleted architectures.
6134 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6136 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6137 (armv5t, armv5te): New features.
6138 (ARMv5, ARMv5e): Delete fgroups.
6139 (ARMv5t, ARMv5te): Adjust for above changes.
6141 (armv5, armv5e): Delete arches.
6142 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6144 (*call_reg_arm): Likewise.
6145 (*call_value_reg_armv5): Likewise.
6146 (*call_value_reg_arm): Likewise.
6147 (*call_symbol): Likewise.
6148 (*call_value_symbol): Likewise.
6149 (*sibcall_insn): Likewise.
6150 (*sibcall_value_insn): Likewise.
6152 (prefetch): Likewise.
6153 (define_split and define_peephole2 dependent on arm_arch5):
6155 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6157 (TARGET_ARM_QBIT): Likewise.
6158 (TARGET_DSP_MULTIPLY): Likewise.
6159 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6160 (arm_arch5, arm_arch5e): Delete.
6161 (arm_arch5t, arm_arch5te): Declare.
6162 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6163 (arm_arch5t): Declare.
6164 (arm_option_reconfigure_globals): Update for the above.
6165 (arm_options_perform_arch_sanity_checks): Update comment, replace
6166 use of arm_arch5 with arm_arch5t.
6167 (use_return_insn): Likewise.
6168 (arm_emit_call_insn): Likewise.
6169 (output_return_instruction): Likewise.
6170 (arm_final_prescan_insn): Likewise.
6171 (arm_coproc_builtin_available): Likewise.
6172 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6173 arm_arch5e with arm_arch5t and arm_arch5te.
6174 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6175 (arm_arch5t, arm_arch5te): Declare.
6176 * config/arm/arm-tables.opt: Regenerate.
6177 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6178 * config/arm/t-multilib: Likewise.
6179 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6180 instead of arm_arch5.
6181 (*call_reg_thumb1): Likewise.
6182 (*call_value_reg_thumb1_v5): Likewise.
6183 (*call_value_reg_thumb1): Likewise.
6184 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6186 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6188 2018-05-18 Martin Liska <mliska@suse.cz>
6190 PR gcov-profile/84846
6191 * doc/gcov.texi: Document -t option of gcov tool.
6193 2018-05-18 Martin Liska <mliska@suse.cz>
6195 PR gcov-profile/84846
6196 * gcov.c (print_usage): Add new -t option.
6197 (process_args): Handle the option.
6198 (generate_results): Use stdout as output when requested by
6201 2018-05-18 Martin Liska <mliska@suse.cz>
6203 PR gcov-profile/84846
6204 * coverage.c (coverage_init): Write PWD to .gcno file.
6205 * doc/gcov.texi: Document how working directory is printed.
6206 * gcov-dump.c (dump_gcov_file): Print PWD.
6207 * gcov.c (output_intermediate_file): Likewise.
6208 (read_graph_file): Read PWD string.
6209 (output_lines): Print PWD.
6211 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6214 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6215 for retval and return false if all args to phi are zero.
6217 2018-05-18 Richard Biener <rguenther@suse.de>
6219 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6221 (evrp_dom_walker::before_dom_children): Call it.
6223 2018-05-18 Richard Biener <rguenther@suse.de>
6225 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6226 results when processing array refs with variable index.
6228 2018-05-18 Toon Moene <toon@moene.org>
6230 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6231 directly after that of -floop-interchange. Indicate that both
6232 options are enabled by default when specifying -O3.
6234 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6236 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6237 iterator. Delete separate integer-mode vec_set<mode> expander.
6238 (aarch64_simd_vec_setv2di): Delete.
6239 (vec_setv2di): Delete.
6240 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6241 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6242 the "w, r" alternative.
6244 2018-05-18 Martin Liska <mliska@suse.cz>
6246 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6247 * tree-pass.h (make_pass_lower_switch_O0): New function.
6248 * tree-switch-conversion.c (node_has_low_bound): Remove.
6249 (node_has_high_bound): Likewise.
6250 (node_is_bounded): Likewise.
6251 (class pass_lower_switch): Make it a template type and create
6253 (pass_lower_switch::execute): Add template argument.
6254 (make_pass_lower_switch): New function.
6255 (make_pass_lower_switch_O0): New function.
6256 (do_jump_if_equal): Remove.
6257 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6258 all the hard work to tree optimization passes.
6260 2018-05-18 Martin Liska <mliska@suse.cz>
6262 * dbgcnt.c (limit_low): Renamed from limit.
6263 (limit_high): New variable.
6264 (dbg_cnt_is_enabled): Check for upper limit.
6265 (dbg_cnt): Adjust dumping.
6266 (dbg_cnt_set_limit_by_index): Add new argument for high
6268 (dbg_cnt_set_limit_by_name): Likewise.
6269 (dbg_cnt_process_single_pair): Parse new format.
6270 (dbg_cnt_process_opt): Use strtok.
6271 (dbg_cnt_list_all_counters): Remove 'value' and add
6273 * doc/invoke.texi: Document changes.
6275 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6277 * doc/sourcebuild.texi (scalar_all_fma): Document.
6278 * tree.def (FMA_EXPR): Delete.
6279 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6280 * internal-fn.c (ternary_direct): New macro.
6281 (expand_ternary_optab_fn): Likewise.
6282 (direct_ternary_optab_supported_p): Likewise.
6283 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6284 * builtins.c (fold_builtin_fma): Delete.
6285 (fold_builtin_3): Don't call it.
6286 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6287 * expr.c (expand_expr_real_2): Likewise.
6288 * fold-const.c (operand_equal_p): Likewise.
6289 (fold_ternary_loc): Likewise.
6290 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6291 * gimple.c (DEFTREECODE): Likewise.
6292 * gimplify.c (gimplify_expr): Likewise.
6293 * optabs-tree.c (optab_for_tree_code): Likewise.
6294 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6295 * tree-eh.c (operation_could_trap_p): Likewise.
6296 (stmt_could_throw_1_p): Likewise.
6297 * tree-inline.c (estimate_operator_cost): Likewise.
6298 * tree-pretty-print.c (dump_generic_node): Likewise.
6299 (op_code_prio): Likewise.
6300 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6301 * tree-ssa-operands.c (get_expr_operands): Likewise.
6302 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6303 * fold-const-call.h (fold_fma): Delete.
6304 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6305 CFN_FNMA and CFN_FNMS.
6307 * genmatch.c (combined_fn): New enum.
6308 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6309 (commutative_op): New function.
6310 (commutate): Use it. Handle more than 2 operands.
6311 (dt_operand::gen_gimple_expr): Use commutative_op.
6312 (parser::parse_expr): Allow :c to be used with non-binary
6313 operators if the commutative operand is known.
6314 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6315 CFN_FMS, CFN_FNMA and CFN_FNMS.
6316 (backprop::process_assign_use): Remove FMA_EXPR handling.
6317 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6318 (gen_hsa_fma): New function.
6319 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6320 IFN_FNMA and IFN_FNMS.
6321 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6322 * gimple-fold.h (follow_all_ssa_edges): Declare.
6323 * gimple-fold.c (follow_all_ssa_edges): New function.
6324 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6325 gimple_build interface and use follow_all_ssa_edges to fold the result.
6326 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6327 instead of checking for optabs directly.
6328 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6329 rather than FMA_EXPRs.
6330 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6331 call to IFN_FMA instead of an FMA_EXPR.
6333 2018-05-17 Jim Wilson <jimw@sifive.com>
6335 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6336 sign extended promoted subreg, and the range does not have the sign bit
6337 set, then do a sign extend.
6339 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6340 test, check for sign extended subreg and/or constant operands, and
6341 do a sign extend in that case.
6343 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6345 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6346 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6348 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6349 Change logics_shift_reg to logics_shift_imm.
6350 (thunderx2t99_fp_loadpair_basic): Delete.
6351 (thunderx2t99_fp_storepair_basic): Delete.
6352 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6353 (thunderx2t99_asimd_polynomial): Delete.
6354 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6355 and neon_fp_mul_d_scalar_q.
6356 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6357 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6358 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6359 (thunderx2t99_asimd_lut): Add missing tbl types.
6360 (thunderx2t99_asimd_ext): Delete.
6361 (thunderx2t99_asimd_load1_1_mult): Delete.
6362 (thunderx2t99_asimd_load1_2_mult): Delete.
6363 (thunderx2t99_asimd_load1_ldp): New.
6364 (thunderx2t99_asimd_load1): New.
6365 (thunderx2t99_asimd_load2): Add missing *load2* types.
6366 (thunderx2t99_asimd_load3): New.
6367 (thunderx2t99_asimd_load4): New.
6368 (thunderx2t99_asimd_store1_1_mult): Delete.
6369 (thunderx2t99_asimd_store1_2_mult): Delete.
6370 (thunderx2t99_asimd_store2_mult): Delete.
6371 (thunderx2t99_asimd_store2_onelane): Delete.
6372 (thunderx2t99_asimd_store_stp): New.
6373 (thunderx2t99_asimd_store1): New.
6374 (thunderx2t99_asimd_store2): New.
6375 (thunderx2t99_asimd_store3): New.
6376 (thunderx2t99_asimd_store4): New.
6378 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6380 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6381 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6383 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6384 Segher Boessenkool <segher@kernel.crashing.org>
6387 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6390 2018-05-17 Richard Biener <rguenther@suse.de>
6392 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6393 for pruning loop and prune defs feeding only already visited PHIs.
6395 2018-05-17 Richard Biener <rguenther@suse.de>
6397 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6399 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6400 Richard Biener <rguenther@suse.de>
6402 PR tree-optimization/85793
6403 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6404 for VMAT_ELEMENTWISE.
6406 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6408 * internal-fn.h (lookup_internal_fn): Declare
6409 * internal-fn.c (lookup_internal_fn): New function.
6410 * gimple.c (gimple_build_call_from_tree): Handle calls to
6412 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6413 internal function names.
6414 * tree-pretty-print.c (dump_generic_node): Likewise.
6415 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6417 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6419 * gimple-fold.h (gimple_build): Make the function forms take
6420 combined_fn rather than built_in_function.
6421 (gimple_simplify): Likewise.
6422 * gimple-match-head.c (gimple_simplify): Likewise.
6423 * gimple-fold.c (gimple_build): Likewise.
6424 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6425 rather than gimple_build_call_internal.
6426 (get_initial_defs_for_reduction): Likewise.
6427 (vect_create_epilog_for_reduction): Likewise.
6428 (vectorizable_live_operation): Likewise.
6430 2018-05-17 Martin Liska <mliska@suse.cz>
6432 * gimple-ssa-sprintf.c (format_directive): Do not use
6433 space in between 'G_' and '('.
6435 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6438 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6439 even if the mask is not all ones.
6442 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6444 (ix86_gimple_fold_builtin): Likewise.
6447 * config/i386/i386.c: Include tree-vector-builder.h.
6448 (ix86_vector_shift_count): New function.
6449 (ix86_fold_builtin): Fold shift builtins by scalar count.
6450 (ix86_gimple_fold_builtin): Likewise.
6452 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6453 _mm512_setzero): New intrinsics.
6455 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6456 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6458 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6459 code generation for cases where splatting a value is not useful.
6460 * simplify-rtx.c (simplify_ternary_operation): Simplify
6461 vec_merge across a vec_duplicate and a paradoxical subreg forming
6462 a vector mode to a vec_concat.
6464 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6466 * config.gcc: Support "goldmont-plus".
6467 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6469 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6470 PROCESSOR_GOLDMONT_PLUS.
6471 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6472 (processor_target_table): Add "goldmont-plus".
6473 (PTA_GOLDMONT_PLUS): Define.
6474 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6475 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6476 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6477 (fold_builtin_cpu): Add "goldmont-plus".
6478 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6479 (ix86_option_override_internal): Add "goldmont-plus".
6480 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6481 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6482 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6483 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6485 2018-05-17 Richard Biener <rguenther@suse.de>
6487 PR tree-optimization/85757
6488 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6489 remove defs that only feed that PHI from further processing.
6491 2018-05-16 Jim Wilson <jimw@sifive.com>
6493 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6495 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6497 2018-05-16 Mark Wielaard <mark@klomp.org>
6499 * dwarf2out.c (count_index_strings): New function.
6500 (output_indirect_strings): Call count_index_strings and generate
6501 header for dwarf_version >= 5.
6503 2018-05-16 Mark Wielaard <mark@klomp.org>
6505 * dwarf2out.c (dwarf_FORM): New function.
6506 (set_indirect_string): Use dwarf_FORM.
6507 (reset_indirect_string): Likewise.
6508 (size_of_die): Likewise.
6509 (value_format): Likewise.
6510 (output_die): Likewise.
6511 (add_skeleton_AT_string): Likewise.
6512 (output_macinfo_op): Likewise.
6513 (index_string): Likewise.
6514 (output_index_string_offset): Likewise.
6515 (output_index_string): Likewise.
6516 (count_index_strings): Likewise.
6518 2018-05-16 Carl Love <cel@us.ibm.com>
6520 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6521 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6523 2018-05-16 Martin Jambor <mjambor@suse.cz>
6525 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6526 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6528 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6530 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6531 (fnma<mode>4): Likewise.
6532 (fms<mode>4): Likewise.
6533 (fnms<mode>4): Likewise.
6534 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6535 (aarch64_fnma<mode>4): Likewise.
6536 (aarch64_fms<mode>4): Likewise.
6537 (aarch64_fnms<mode>4): Likewise.
6538 (aarch64_fnmadd<mode>4): Likewise.
6540 2018-05-16 Jason Merrill <jason@redhat.com>
6542 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6544 2018-05-16 Richard Biener <rguenther@suse.de>
6546 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6547 (dump_stmt_cost): Declare.
6548 (add_stmt_cost): Dump cost we add.
6549 (add_stmt_costs): New function.
6550 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6552 (vect_analyze_stmt): Adjust prototype.
6553 (vectorizable_condition): Likewise.
6554 (vectorizable_live_operation): Likewise.
6555 (vectorizable_reduction): Likewise.
6556 (vectorizable_induction): Likewise.
6557 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6558 cost vector to pass to vectorizable_ and record afterwards.
6559 (vect_model_reduction_cost): Take cost vector argument and adjust.
6560 (vect_model_induction_cost): Likewise.
6561 (vectorizable_reduction): Likewise.
6562 (vectorizable_induction): Likewise.
6563 (vectorizable_live_operation): Likewise.
6564 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6565 SLP_TREE_NUMBER_OF_VEC_STMTS.
6566 (vect_analyze_slp_cost_1): Remove.
6567 (vect_analyze_slp_cost): Likewise.
6568 (vect_slp_analyze_node_operations): Take visited args and
6569 a target cost vector. Avoid processing already visited stmt sets.
6570 (vect_slp_analyze_operations): Use a local cost vector to gather
6571 costs and register those of non-discarded instances.
6572 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6573 (vect_schedule_slp_instance): Remove copying of
6574 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6576 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6577 adding cost. Record cost entry location.
6578 (vect_prologue_cost_for_slp_op): Function to compute cost of
6579 a constant or invariant generated for SLP vect in the prologue,
6580 split out from vect_analyze_slp_cost_1.
6581 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6582 (vect_model_promotion_demotion_cost): Likewise.
6583 (vect_model_store_cost): Likewise, make static.
6584 (vect_model_load_cost): Likewise.
6585 (vectorizable_bswap): Add cost vector arg and adjust.
6586 (vectorizable_call): Likewise.
6587 (vectorizable_simd_clone_call): Likewise.
6588 (vectorizable_conversion): Likewise.
6589 (vectorizable_assignment): Likewise.
6590 (vectorizable_shift): Likewise.
6591 (vectorizable_operation): Likewise.
6592 (vectorizable_store): Likewise.
6593 (vectorizable_load): Likewise.
6594 (vectorizable_condition): Likewise.
6595 (vectorizable_comparison): Likewise.
6596 (can_vectorize_live_stmts): Likewise.
6597 (vect_analyze_stmt): Likewise.
6598 (vect_transform_stmt): Adjust calls to vectorizable_*.
6599 * tree-vectorizer.c: Include gimple-pretty-print.h.
6600 (dump_stmt_cost): New function.
6602 2018-05-16 Richard Biener <rguenther@suse.de>
6604 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6605 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6606 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6607 (check_name): New callback.
6608 (dse_classify_store): Track cycles via a visited bitmap of PHI
6609 defs and simplify handling of in-loop and across loop dead stores
6610 and properly fail for loop-variant refs. Handle byte-tracking with
6611 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6614 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6616 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6617 (vect_get_mask_type_for_stmt): Likewise.
6618 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6620 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6621 to determine the statement's vector type and the vector type that
6622 should be used for calculating nunits. Deal with cases in which
6623 the type has to be deferred.
6624 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6625 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6626 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6627 (vect_determine_vf_for_stmt): New functions, split out from...
6628 (vect_determine_vectorization_factor): ...here.
6629 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6630 (vect_get_mask_type_for_stmt): New functions, split out from
6631 vect_determine_vectorization_factor.
6633 2018-05-16 Richard Biener <rguenther@suse.de>
6635 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6636 verify the [VEC_]COND_EXPR embedded comparison.
6638 2018-05-15 Martin Sebor <msebor@redhat.com>
6640 PR tree-optimization/85753
6641 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6642 RECORD_TYPE in addition to ARRAY_TYPE.
6644 2018-05-15 Martin Sebor <msebor@redhat.com>
6647 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6649 2018-05-15 Richard Biener <rguenther@suse.de>
6651 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6652 add by_clobber_p one. Change algorithm to collect all defs
6653 representing uses we need to walk and try reducing them to
6654 a single one before failing.
6655 (dse_dom_walker::dse_optimize_stmt): Adjust.
6657 2018-05-13 Mark Wielaard <mark@klomp.org>
6659 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6660 (size_of_loc_descr): Likewise.
6661 (output_loc_operands): Likewise.
6662 (output_loc_operands_raw): Likewise.
6663 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6664 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6665 (hash_loc_operands): Likewise.
6666 (compare_loc_operands): Likewise.
6668 2018-05-14 Mark Wielaard <mark@klomp.org>
6670 * dwarf2out.c (count_index_addrs): New function.
6671 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6673 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6675 PR tree-optimization/83648
6676 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6677 return value as malloc candidate.
6679 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6682 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6683 param as true in call to suggest_attribute.
6685 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6687 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6688 -mreadonly-in-sdata.
6690 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6692 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6694 (aarch64_crypto_aesd_fused): Likewise.
6696 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6698 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6699 (movsi_aarch64): Likewise.
6700 (load_pairsi): Likewise.
6701 (load_pairdi): Likewise.
6702 (store_pairsi): Likewise.
6703 (store_pairdi): Likewise.
6704 (load_pairsf): Likewise.
6705 (load_pairdf): Likewise.
6706 (store_pairsf): Likewise.
6707 (store_pairdf): Likewise.
6708 (zero_extend): Likewise.
6709 (trunc): Swap alternatives.
6710 (fcvt_target): Add '?' to prefer w over r.
6712 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6715 * config/i386/i386.md: Disallow non-commutative arithmetics in
6716 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6717 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6718 in the peephole2 before it.
6720 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6722 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6723 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6724 (ix86_handle_option): Handle -mcldemote.
6725 * config.gcc: New header.
6726 * config/i386/cldemoteintrin.h: New file.
6727 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6728 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6730 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6731 OPTION_MASK_ISA_CLDEMOTE.
6732 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6733 (ix86_valid_target_attribute_inner_p): Ditto.
6734 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6735 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6736 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6737 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6738 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6740 * config/i386/i386.opt: Add -mcldemote.
6741 * config/i386/x86intrin.h: New header.
6742 * doc/invoke.texi: Add -mcldemote.
6744 2018-05-14 Richard Biener <rguenther@suse.de>
6746 * doc/match-and-simplify.texi: Adjust :s documentation.
6748 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6750 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6751 intended memcpy size.
6752 (REORDER_45): Likewise.
6754 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6756 * sort.cc: New file.
6757 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6758 * vec.c (qsort_chk): Use gcc_qsort.
6759 * Makefile.in (OBJS-libcommon): Add sort.o.
6760 (build/sort.o): New target. Use it...
6761 (BUILD_RTL): ... here, and...
6762 (build/gencfn-macros): ... here, and...
6763 (build/genmatch): ... here.
6765 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6766 Chung-Ju Wu <jasonwucj@gmail.com>
6768 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6769 * config/nds32/nds32-graywolf.md: New file.
6770 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6771 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6773 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6774 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6775 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6776 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6777 * config/nds32/pipelines.md: Include n15 settings.
6779 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6780 Chung-Ju Wu <jasonwucj@gmail.com>
6782 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6783 * config/nds32/nds32-n13.md: New file.
6784 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6785 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6787 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6788 * config/nds32/nds32.md (pipeline_model): Add n13.
6789 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6790 * config/nds32/pipelines.md: Include n13 settings.
6792 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6793 Chung-Ju Wu <jasonwucj@gmail.com>
6795 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6796 * config/nds32/nds32-n10.md: New file.
6797 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6798 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6800 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6801 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6802 * config/nds32/nds32.md (pipeline_model): Add n10.
6803 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6804 * config/nds32/pipelines.md: Include n10 settings.
6806 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6807 Kito Cheng <kito.cheng@gmail.com>
6808 Chung-Ju Wu <jasonwucj@gmail.com>
6810 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6811 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6812 Add enum values for DSP extension instructions.
6813 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6815 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6816 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6818 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6819 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6820 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6821 * config/nds32/nds32-intrinsic.md: Likewise.
6822 * config/nds32/nds32_intrinsic.h: Likewise.
6823 * config/nds32/nds32-md-auxiliary.c: Likewise.
6824 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6825 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6826 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6827 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6828 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6829 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6830 TYPE_DMAC in switch statement.
6831 * config/nds32/nds32.c: New checking and implementation for DSP
6832 extension instructions.
6833 * config/nds32/nds32.h: Likewise.
6834 * config/nds32/nds32.md: Likewise.
6835 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6836 * config/nds32/predicates.md: Implement new predicates for DSP
6839 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6841 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6842 Reformat alternatives and attributes so it is easier to identify
6843 which constraints/attributes go with which instruction.
6844 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6845 (mov<mode>_softfloat32, FMOVE64): Likewise.
6846 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6847 (mov<mode>_softfloat64, FMOVE64): Likewise.
6849 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6851 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6853 (Basic PowerPC Built-in Functions): The new name of the
6854 subsection previously known as "PowerPC Built-in Functions".
6855 (Basic PowerPC Built-in Functions Available on all Configurations):
6857 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6858 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6859 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6860 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6862 2018-05-11 Martin Jambor <mjambor@suse.cz>
6865 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6868 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6871 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6873 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6875 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6876 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6877 (ix86_handle_option): Handle -mwaitpkg.
6878 * config.gcc: New header.
6879 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6880 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6881 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6883 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6884 OPTION_MASK_ISA_WAITPKG.
6885 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6886 (ix86_option_override_internal): Add PTA_WAITPKG.
6887 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6888 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6889 IX86_BUILTIN_TPAUSE.
6890 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6891 __builtin_ia32_umwait and __builtin_ia32_tpause.
6892 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6893 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6894 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6895 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6896 UNSPECV_TPAUSE): New.
6897 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6898 * config/i386/i386.opt: Add -mwaitpkg.
6899 * config/i386/waitpkgintrin.h: New file.
6900 * config/i386/x86intrin.h: New header.
6901 * doc/invoke.texi: Add -mwaitpkg.
6903 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6906 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6908 (cortex-m0): Use armv6s-m isa.
6909 (cortex-m0plus): Likewise.
6910 (cortex-m1): Likewise.
6911 (cortex-m0.small-multiply): Likewise.
6912 (cortex-m0plus.small-multiply): Likewise.
6913 (cortex-m1.small-multiply): Likewise.
6915 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6916 Jakub Jelinek <jakub@redhat.com>
6918 PR tree-optimization/85692
6919 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6920 source permute as well.
6922 2018-05-11 Martin Liska <mliska@suse.cz>
6925 * doc/extend.texi: Document LLVM style format for no_sanitize
6928 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6930 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6931 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6932 (mode_supports_vsx_dform_quad): Likewise.
6933 (mode_supports_vmx_dform): Move these functions to be next to the
6934 other mode_supports functions.
6935 (mode_supports_dq_form): Likewise.
6936 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6937 mode_supports_dq_form.
6938 (reg_offset_addressing_ok_p): Likewise.
6939 (offsettable_ok_by_alignment): Likewise.
6940 (rs6000_legitimate_offset_address_p): Likewise.
6941 (legitimate_lo_sum_address_p): Likewise.
6942 (rs6000_legitimize_address): Likewise.
6943 (rs6000_legitimize_reload_address): Likewise.
6944 (rs6000_secondary_reload_inner): Likewise.
6945 (rs6000_preferred_reload_class): Likewise.
6946 (rs6000_output_move_128bit): Likewise.
6948 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6950 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6951 Generate SImode target register for null target.
6952 <case IX86_BUILTIN_XGETBV>: Ditto.
6953 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6954 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6956 2018-05-10 Carl Love <cel@us.ibm.com>
6958 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6959 dcbtt and dcbtstt if operands[2] is 0.
6961 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6964 * config/i386/sse.md (usadv64qi): New expander.
6966 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6968 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6969 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6970 -maltivec=be support.
6971 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6972 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6973 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6974 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6975 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6976 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6977 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6978 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6979 altivec_vsumsws): Adjust.
6980 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6981 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6982 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6984 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6985 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6986 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6987 (altivec_lve<VI_char>x): Delete expand.
6988 (*altivec_lve<VI_char>x_internal): Rename to...
6989 (altivec_lve<VI_char>x): ... this.
6990 (altivec_lvxl_<mode>): Delete expand.
6991 (*altivec_lvxl_<mode>_internal): Rename to ...
6992 (altivec_lvxl_<mode>): ... this.
6993 (altivec_stvxl_<mode>): Delete expand.
6994 (*altivec_stvxl_<mode>_internal): Rename to ...
6995 (altivec_stvxl_<mode>): ... this.
6996 (altivec_stve<VI_char>x): Delete expand.
6997 (*altivec_stve<VI_char>x_internal): Rename to ...
6998 (altivec_stve<VI_char>x): ... this.
6999 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7000 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7001 reduc_plus_scal_<mode>): Adjust.
7002 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7004 (rs6000_cpu_cpp_builtins): Adjust.
7005 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7006 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7007 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7008 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7009 -maltivec=be support.
7010 (rs6000_split_vec_extract_var): Adjust.
7011 (rs6000_split_v4si_init): Adjust.
7012 (swap_selector_for_mode): Delete.
7013 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7014 altivec_expand_stvex_be): Delete.
7015 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7016 -maltivec=be support.
7017 (rs6000_gimple_fold_builtin): Ditto.
7018 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7020 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7021 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7022 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7023 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7024 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7025 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7026 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7027 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7028 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7029 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7030 anonymous split): Adjust.
7031 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7032 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7034 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7036 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7037 when --with-gxx-include-dir is also specified.
7038 * configure: Regenerate.
7040 2018-05-09 Jim Wilson <jimw@sifive.com>
7043 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7044 * config/riscv/t-withmultilib: New.
7045 * config/riscv/withmultilib.h: New.
7046 * doc/install.texi: Document RISC-V --with-multilib-list support.
7048 2018-05-09 Richard Biener <rguenther@suse.de>
7050 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7052 (vect_bb_vectorization_profitable_p): Adjust. Compute
7053 actual scalar cost using the cost vector and the add_stmt_cost
7056 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7058 PR rtl-optimization/85645
7059 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7060 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7062 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7064 PR rtl-optimization/85645
7065 * shrink-wrap.c (spread_components): Return a boolean saying if
7066 anything was changed.
7067 (try_shrink_wrapping_separate): Iterate spread_components until
7068 nothing changes anymore.
7070 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7072 PR rtl-optimization/85645
7073 * regrename.c (build_def_use): Also kill the chains that include the
7074 destination of a REG_CFA_REGISTER note.
7076 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7078 PR rtl-optimization/85645
7079 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7080 insn that has a REG_CFA_REGISTER note.
7082 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7084 * cfgexpand.c (expand_clobber): New function.
7085 (expand_gimple_stmt_1): Use it.
7086 * tree-vect-stmts.c (vect_clobber_variable): New function,
7088 (vectorizable_simd_clone_call): ...here.
7089 (vectorizable_store): Emit a clobber either side of an
7090 IFN_STORE_LANES sequence.
7091 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7093 2018-05-09 Tom de Vries <tom@codesourcery.com>
7096 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7097 (define_insn "trap_if_false"): Add exit after trap.
7099 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7101 PR rtl-optimization/85638
7102 * bb-reorder.c: Include common/common-target.h.
7103 (create_forwarder_block): New function extracted from...
7104 (fix_up_crossing_landing_pad): ...here. Rename into...
7105 (dw2_fix_up_crossing_landing_pad): ...this.
7106 (sjlj_fix_up_crossing_landing_pad): New function.
7107 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7108 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7109 from both partitions and exit the loop after one iteration.
7111 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7114 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7116 (Basic PowerPC Built-in Functions): The new name of the
7117 subsection previously known as "PowerPC Built-in Functions".
7118 (Basic PowerPC Built-in Functions Available on all Configurations):
7120 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7122 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7124 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7126 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7129 2018-05-08 Jim Wilson <jimw@sifive.com>
7131 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7132 (LD_EMUL_SUFFIX): New.
7133 (LINK_SPEC): Use it.
7135 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7137 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7139 (Basic PowerPC Built-in Functions): The new name of the
7140 subsection previously known as "PowerPC Built-in Functions".
7141 (Basic PowerPC Built-in Functions Available on all Configurations):
7143 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7145 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7147 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7149 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7152 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7155 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7156 after cmpelim optimization.
7158 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7160 * config.gcc: Support "goldmont".
7161 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7162 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7164 * config/i386/i386.c (m_GOLDMONT): Define.
7165 (processor_target_table): Add "goldmont".
7166 (PTA_GOLDMONT): Define.
7167 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7168 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7169 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7170 (fold_builtin_cpu): Add "goldmont".
7171 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7172 (ix86_option_override_internal): Add "goldmont".
7173 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7174 (processor_type): Add PROCESSOR_GOLDMONT.
7175 * config/i386/i386.md: Add CPU "glm".
7176 * config/i386/glm.md: New file.
7177 * config/i386/x86-tune.def: Add m_GOLDMONT.
7178 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7180 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7183 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7185 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7186 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7187 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7190 * config/i386/i386.c (ix86_fold_builtin): Handle
7191 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7194 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7195 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7197 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7200 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7201 (check_arch): Likewise.
7202 (check_fpu): Return the result rather than printing it.
7203 (end arch): Fix operator precedence.
7204 (end cpu): Likewise.
7205 (END): Print the result from check_fpu.
7207 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7208 Alan Hayward <alan.hayward@arm.com>
7209 David Sherwood <david.sherwood@arm.com>
7211 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7212 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7213 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7214 (*fcmuo<mode>_and): New patterns.
7216 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7218 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7219 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7220 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7221 (cmp_op, sve_imm_con): New code attributes.
7222 (SVE_COND_INT_CMP, imm_con): Delete.
7223 (cmp_op): Remove above unspecs from int attribute.
7224 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7226 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7227 comparison-specific unspecs.
7228 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7229 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7230 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7231 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7232 (*vec_fcm<cmp_op><mode>): Rename to...
7233 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7234 (*vec_fcmuo<mode>): Rename to...
7235 (*fcmuo<mode>): ...this and adjust likewise.
7236 (*pred_fcm<cmp_op><mode>): New pattern.
7237 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7238 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7240 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7242 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7243 (aarch64_emit_sve_predicated_cond): New function.
7244 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7245 (aarch64_emit_unspec_cond_or): Replace with...
7246 (aarch64_emit_sve_or_conds): ...this new function. Use
7247 aarch64_emit_sve_ptrue_op for the individual comparisons and
7248 aarch64_emit_binop to OR them together.
7249 (aarch64_emit_inverted_unspec_cond): Replace with...
7250 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7251 aarch64_emit_sve_ptrue_op for the comparison and
7252 aarch64_emit_unop to invert the result.
7253 (aarch64_expand_sve_vec_cmp_float): Update after the above
7254 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7256 2018-05-07 Nathan Sidwell <nathan@acm.org>
7258 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7259 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7260 (Backwards Compatibility): Likewise.
7262 2018-05-07 Luis Machado <luis.machado@linaro.org>
7266 2018-05-07 Luis Machado <luis.machado@linaro.org>
7268 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7269 <prefetch_dynamic_strides>: New const bool field.
7270 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7271 prefetch_dynamic_strides.
7272 (exynosm1_prefetch_tune): Likewise.
7273 (thunderxt88_prefetch_tune): Likewise.
7274 (thunderx_prefetch_tune): Likewise.
7275 (thunderx2t99_prefetch_tune): Likewise.
7276 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7278 (aarch64_override_options_internal): Update to set
7279 PARAM_PREFETCH_DYNAMIC_STRIDES.
7280 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7281 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7282 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7283 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7284 prefetch-dynamic-strides setting.
7286 2018-05-07 Luis Machado <luis.machado@linaro.org>
7288 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7289 <minimum_stride>: New const int field.
7290 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7291 minimum_stride field.
7292 (exynosm1_prefetch_tune): Likewise.
7293 (thunderxt88_prefetch_tune): Likewise.
7294 (thunderx_prefetch_tune): Likewise.
7295 (thunderx2t99_prefetch_tune): Likewise.
7296 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7297 (aarch64_override_options_internal): Update to set
7298 PARAM_PREFETCH_MINIMUM_STRIDE.
7299 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7300 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7301 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7302 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7303 stride is constant and is below the minimum stride threshold.
7305 2018-05-07 Luis Machado <luis.machado@linaro.org>
7307 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7310 2018-05-07 Luis Machado <luis.machado@linaro.org>
7312 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7313 <prefetch_dynamic_strides>: New const bool field.
7314 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7315 prefetch_dynamic_strides.
7316 (exynosm1_prefetch_tune): Likewise.
7317 (thunderxt88_prefetch_tune): Likewise.
7318 (thunderx_prefetch_tune): Likewise.
7319 (thunderx2t99_prefetch_tune): Likewise.
7320 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7322 (aarch64_override_options_internal): Update to set
7323 PARAM_PREFETCH_DYNAMIC_STRIDES.
7324 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7325 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7326 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7327 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7328 prefetch-dynamic-strides setting.
7330 2018-05-07 Luis Machado <luis.machado@linaro.org>
7332 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7333 <minimum_stride>: New const int field.
7334 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7335 minimum_stride field.
7336 (exynosm1_prefetch_tune): Likewise.
7337 (thunderxt88_prefetch_tune): Likewise.
7338 (thunderx_prefetch_tune): Likewise.
7339 (thunderx2t99_prefetch_tune): Likewise.
7340 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7341 (aarch64_override_options_internal): Update to set
7342 PARAM_PREFETCH_MINIMUM_STRIDE.
7343 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7344 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7345 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7346 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7347 stride is constant and is below the minimum stride threshold.
7349 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7352 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7353 the type is addressable. Don't force op into register if it has
7356 2018-05-05 Roland McGrath <mcgrathr@google.com>
7359 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7360 any section for which we don't know a specific type it should have,
7361 regardless of name. Previously this was done only for the exact
7362 names ".init_array", ".fini_array", and ".preinit_array".
7363 (default_elf_asm_named_section): Add comment about
7364 relationship with default_section_type_flags and SECTION_NOTYPE.
7365 (get_section): Don't consider it a type conflict if one side has
7366 SECTION_NOTYPE and the other doesn't, as long as neither has the
7367 SECTION_BSS et al used in the default_section_type_flags logic.
7369 2018-05-05 Tom de Vries <tom@codesourcery.com>
7372 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7373 (workaround_barsyncs): New function.
7374 (nvptx_reorg): Use workaround_barsyncs.
7375 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7376 (define_expand "nvptx_membar_cta"): New define_expand.
7377 (define_insn "*nvptx_membar_cta"): New insn.
7379 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7381 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7382 To improve optimization opportunities.
7383 * builtin-types.def: The new needed builtin types for the above.
7385 2018-05-04 Richard Biener <rguenther@suse.de>
7387 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7388 * gimple-ssa-store-merging.c
7389 (imm_store_chain_info::output_merged_store): Remove redundant create,
7390 release split_store vector contents on failure.
7391 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7392 scalar stmt vector on cache hit.
7394 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7396 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7398 * config.gcc (powerpc-xilinx-eabi*): Remove.
7399 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7401 (fusion_addis_mem_combo_load): Ditto.
7402 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7404 (rs6000_cpu_cpp_builtins): Ditto.
7405 * config/rs6000/rs6000-linux.c
7406 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7407 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7408 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7410 (rs6000_setup_reg_addr_masks): Ditto.
7411 (rs6000_init_hard_regno_mode_ok): Ditto.
7412 (rs6000_option_override_internal): Ditto.
7413 (legitimate_lo_sum_address_p): Ditto.
7414 (rs6000_legitimize_address): Ditto.
7415 (rs6000_legitimize_reload_address): Ditto.
7416 (rs6000_legitimate_address_p): Ditto.
7417 (abi_v4_pass_in_fpr): Ditto.
7418 (setup_incoming_varargs): Ditto.
7419 (rs6000_gimplify_va_arg): Ditto.
7420 (rs6000_split_multireg_move): Ditto.
7421 (rs6000_savres_strategy): Ditto.
7422 (rs6000_emit_prologue_components): Ditto.
7423 (rs6000_emit_epilogue_components): Ditto.
7424 (rs6000_emit_prologue): Ditto.
7425 (rs6000_emit_epilogue): Ditto.
7426 (rs6000_elf_file_end): Ditto.
7427 (rs6000_function_value): Ditto.
7428 (rs6000_libcall_value): Ditto.
7429 * config/rs6000/rs6000.h: Ditto.
7430 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7431 (TARGET_MINMAX): ... this. New.
7432 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7433 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7434 (*movsi_internal1_single): Delete.
7435 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7436 mfpu=, mxilinx-fpu): Delete.
7437 * config/rs6000/singlefp.h: Delete.
7438 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7439 * config/rs6000/t-rs6000: Ditto.
7440 * config/rs6000/t-xilinx: Delete.
7441 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7442 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7443 (VStype_simple): Delete.
7444 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7445 * config/rs6000/xfpu.h: Delete.
7446 * config/rs6000/xfpu.md: Delete.
7447 * config/rs6000/xilinx.h: Delete.
7448 * config/rs6000/xilinx.opt: Delete.
7449 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7450 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7452 2018-05-04 Tom de Vries <tom@codesourcery.com>
7455 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7458 2018-05-04 Richard Biener <rguenther@suse.de>
7461 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7462 (expand_complex_div_wide): Likewise.
7463 (expand_complex_operations_1): Likewise.
7464 (expand_complex_libcall): Preserve EH info of the original stmt.
7465 (tree_lower_complex): Handle removed blocks.
7466 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7467 on complex multiplication and division libcall builtins.
7469 2018-05-04 Richard Biener <rguenther@suse.de>
7472 * fold-const.c (negate_expr_p): Restrict negation of operand
7473 zero of a division to when we know that can happen without
7475 (fold_negate_expr_1): Likewise.
7477 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7480 * real.h (real_nextafter): Declare.
7481 * real.c (real_nextafter): New function.
7482 * fold-const-call.c (fold_const_nextafter): New function.
7483 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7484 CASE_CFN_NEXTTOWARD.
7485 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7486 even when arg1_mode is different from arg0_mode.
7488 2018-05-03 Nathan Sidwell <nathan@acm.org>
7490 * doc/extend.texi (Deprecated Features): Remove
7492 (Backwards Compatibility): Likewise.
7493 * doc/invoke.texi (C++ Language Options): Likewise.
7494 (C++ Dialect Options): Likewise.
7496 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7499 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7500 _mm512_mask_mullox_epi64): New intrinsics.
7502 2018-05-03 Tom de Vries <tom@codesourcery.com>
7505 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7506 dump files): Add offload-tree.
7508 2018-05-03 Richard Biener <rguenther@suse.de>
7510 PR tree-optimization/85615
7511 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7512 to loops not nested in BBs loop father to avoid creating multi-entry
7515 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7517 PR tree-optimization/70291
7518 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7519 arguments. Change return type to tree. Emit libcall as a new
7520 statement rather than replacing existing one when inplace_p is true.
7521 (expand_complex_multiplication_components): New function.
7522 (expand_complex_multiplication): Expand floating-point complex
7523 multiplication using the above.
7524 (expand_complex_division): Rename inner_type parameter to type.
7525 Update expand_complex_libcall call-site.
7526 (expand_complex_operations_1): Update expand_complex_multiplication
7527 and expand_complex_division call-sites.
7529 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7532 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7533 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7534 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7535 the highest significant bit of the shift count mask is clear. In
7536 check whether and[sq]i3 is needed verify that all significant bits
7537 of the shift count other than the highest are set.
7539 2018-05-02 Tom de Vries <tom@codesourcery.com>
7542 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7543 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7544 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7545 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7546 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7547 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7548 __builtin_goacc_parlevel_size.
7550 2018-05-02 Richard Biener <rguenther@suse.de>
7552 PR tree-optimization/85597
7553 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7554 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7557 2018-05-02 Tom de Vries <tom@codesourcery.com>
7560 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7561 dump files): Add ltrans-tree.
7563 2018-05-02 Tom de Vries <tom@codesourcery.com>
7566 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7567 dump files): Add wpa-ipa.
7569 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7571 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7572 powerpc*-*-linux*paired* target.
7573 * config/rs6000/750cl.h: Delete.
7574 * config/rs6000/paired.h: Delete.
7575 * config/rs6000/paired.md: Delete.
7576 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7578 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7579 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7580 comment. Remove paired float support.
7581 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7582 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7584 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7585 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7587 * config/rs6000/rs6000.c: Remove paired float support.
7588 (paired_expand_vector_init, paired_expand_vector_move,
7589 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7590 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7591 paired_expand_builtin, paired_expand_predicate_builtin,
7592 paired_init_builtins): Delete.
7593 * config/rs6000/rs6000.h: Remove paired float support.
7594 * config/rs6000/rs6000.md: Remove paired float support.
7595 (move_from_CR_ov_bit): Delete.
7596 * config/rs6000/rs6000.opt (mpaired): Delete.
7597 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7598 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7600 2018-05-02 Richard Biener <rguenther@suse.de>
7603 * gimplify.c (gimplify_save_expr): When in SSA form allow
7604 SAVE_EXPRs to compute to SSA vars.
7606 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7609 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7610 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7611 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7612 clobber operands[2], instead use a new pseudo. Formatting fixes.
7614 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7616 PR tree-optimization/85586
7617 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7618 exit early for statements in the same group if the accesses are
7621 2018-05-02 Tom de Vries <tom@codesourcery.com>
7624 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7627 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7629 PR tree-optimization/85143
7630 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7632 2018-05-01 Tom de Vries <tom@codesourcery.com>
7635 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7636 not found" error message.
7638 2018-05-01 Tom de Vries <tom@codesourcery.com>
7641 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7642 * vec.c (test_ordered_remove_if): New function.
7643 (vec_c_tests): Call test_ordered_remove_if.
7644 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7645 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7646 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7647 VEC_ORDERED_REMOVE_IF.
7649 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7651 PR tree-optimization/82665
7652 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7653 pointer subtraction where arguments come from a memchr call.
7655 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7657 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7658 --push-state --as-needed and --pop-state instead of --as-needed and
7659 --no-as-needed if ld supports it.
7660 * configure: Regenerated.
7663 * doc/install.texi2html: Replace _002d with - and _002a with * in
7664 generated html files using sed.
7666 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7669 * gcc-rich-location.c (blank_line_before_p): New function.
7670 (use_new_line): New function.
7671 (gcc_rich_location::add_fixit_insert_formatted): New function.
7672 * gcc-rich-location.h
7673 (gcc_rich_location::add_fixit_insert_formatted): New function.
7675 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7677 * selftest.c (assert_streq): Rename "expected" and "actual" to
7678 "val1" and "val2". Extend NULL-handling to cover both inputs
7679 symmetrically, while still requiring both to be non-NULL for a pass.
7680 * selftest.h (assert_streq): Rename "expected" and "actual" to
7682 (ASSERT_EQ): Likewise.
7683 (ASSERT_EQ_AT): Likewise.
7684 (ASSERT_KNOWN_EQ): Likewise.
7685 (ASSERT_KNOWN_EQ_AT): Likewise.
7686 (ASSERT_NE): Likewise.
7687 (ASSERT_MAYBE_NE): Likewise.
7688 (ASSERT_MAYBE_NE_AT): Likewise.
7689 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7690 the assertion to pass.
7691 (ASSERT_STREQ_AT): Likewise.
7693 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7695 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7696 interaction with -pie.
7698 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7700 * selftest.h: Fix alphabetization of per-source-file selftest
7703 2018-04-30 Jason Merrill <jason@redhat.com>
7705 PR c++/61982 - dead stores to destroyed objects.
7706 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7709 2018-04-30 Jason Merrill <jason@redhat.com>
7711 * tree.c (build_clobber): New.
7712 * tree.h: Declare it.
7713 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7715 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7717 * diagnostic-show-locus.c (layout::layout): Update for
7718 location_get_source_line returning a char_span.
7719 (struct char_span): Move to input.h.
7720 (struct correction): Update for fields in char_span becoming
7722 (struct source_line): Update for location_get_source_line
7723 returning a char_span.
7724 (layout::print_line): Likewise.
7725 * edit-context.c (edited_file::print_content): Likewise.
7726 (edited_file::print_diff_hunk): Likewise.
7727 (edited_file::print_run_of_changed_lines): Likewise.
7728 (edited_file::get_num_lines): Likewise.
7729 (edited_line::edited_line): Likewise.
7730 * final.c (asm_show_source): Likewise.
7731 * input.c (location_get_source_line): Convert return type
7732 from const char * to char_span, losing the final "line_len"
7734 (dump_location_info): Update for the above.
7735 (get_substring_ranges_for_loc): Likewise. Use a char_span
7736 when handling the literal within the line.
7737 (test_reading_source_line): Update for location_get_source_line
7738 returning a char_span.
7739 * input.h (class char_span): Move here from
7740 diagnostic-show-locus.c, converting from a struct to a class.
7741 Make data members private.
7742 (char_span::operator bool): New.
7743 (char_span::length): New.
7744 (char_span::get_buffer): New.
7745 (char_span::operator[]): New.
7746 (char_span::subspan): Make const.
7747 (char_span::xstrdup): New.
7748 (location_get_source_line): Convert return type from const char *
7749 to char_span, losing the final "line_size" param.
7751 2018-04-30 Jan Hubicka <jh@suse.cz>
7753 * lto-wrapper.c (ltrans_priorities): New static var.
7754 (cmp_priority): New.
7755 (run_gcc): Read priorities and if doing parallel build order
7756 the Makefile by them.
7758 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7760 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7762 2018-04-30 Richard Biener <rguenther@suse.de>
7764 * tree-cfg.c (verify_address): Remove base argument, add
7765 flag whether to check TREE_ADDRESSABLE and do that.
7766 (verify_expr): Remove.
7767 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7768 (verify_gimple_assign_single): Likewise.
7769 (verify_gimple_switch): Likewise.
7770 (verify_expr_location_1): Dereference tp once. Add (disabled)
7771 piece from verify_expr.
7772 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7774 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7776 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7778 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7780 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7781 (small_data_pattern): Likewise.
7782 (arc_rewrite_small_data): Likewise.
7783 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7784 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7785 (get_symbol_alignment): New function.
7786 (legitimate_small_data_address_p): Likewise.
7787 (legitimate_scaled_address): Update, call
7788 legitimate_small_data_address_p.
7789 (output_sdata): New static variable.
7790 (arc_print_operand): Update how we handle small data operands.
7791 (arc_print_operand_address): Likewise.
7792 (arc_legitimate_address_p): Update, use
7793 legitimate_small_data_address_p.
7794 (arc_rewrite_small_data_p): Remove.
7795 (arc_rewrite_small_data_1): Likewise.
7796 (arc_rewrite_small_data): Likewise.
7797 (small_data_pattern): Likewise.
7798 (compact_sda_memory_operand): Update to use
7799 legitimate_small_data_address_p and get_symbol_alignment.
7800 (prepare_move_operands): Don't rewite sdata pattern.
7801 (prepare_extend_operands): Remove.
7802 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7804 (zero_extendqisi2): Likewise.
7805 (zero_extendhisi2): Likewise.
7806 (extendqihi2): Likewise.
7807 (extendqisi2): Likewise.
7808 (extendhisi2): Likewise.
7812 * config/arc/constraints.md (Usd): Change it to memory constraint.
7814 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7816 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7817 as source of std instructions.
7818 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7819 allow 6-bit constants as source for store instructions.
7820 (movdi_insn): Update instruction pattern to allow 6-bit constants
7821 as source for store instructions.
7823 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7825 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7827 2018-04-30 Nathan Sidwell <nathan@acm.org>
7828 Sandra Loosemore <sandra@codesourcery.com>
7830 * dumpfile.c (dump_open): Allow '-' for stdout.
7831 * doc/invoke.texi (Developer Options): Document dump filename
7832 determination early. Document stdin/stdout selection.
7834 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7836 Microblaze Target: PIC data text relative
7838 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7839 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7841 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7842 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7843 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7844 New addressing mode for data-text relative position indepenedent code.
7845 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7846 'ADDRESS_SYMBOLIC_TXT_REL'.
7847 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7848 (microblaze_legitimate_pic_operand): Exclude function calls from
7849 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7850 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7852 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7853 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7854 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7855 for 'address + offset'.
7856 (microblaze_expand_prologue): Add new function prologue call for
7858 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7859 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7860 table in case of TARGET_PIC_DATA_TEXT_REL.
7861 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7862 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7863 Add new macros 'UNSPEC_TEXT',
7864 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7865 + exclude function calls from 'UNSPEC_PLT' in case of data text
7867 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7868 new target hook for generating address diff vector tables in case of
7870 * doc/tm.texi : Regenerate.
7871 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7872 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7873 of addr diff vector generation.
7874 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7875 target hook definition.
7876 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7877 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7878 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7879 option for data text relative.
7881 2018-04-30 Richard Biener <rguenther@suse.de>
7883 * tree-chrec.h (evolution_function_is_constant_p): Remove
7885 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7887 2018-04-30 Richard Biener <rguenther@suse.de>
7890 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7892 2018-04-30 Richard Biener <rguenther@suse.de>
7894 PR tree-optimization/28364
7895 PR tree-optimization/85275
7896 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7897 copying first exit test.
7899 2018-04-28 Mark Wielaard <mark@klomp.org>
7901 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7903 (dwarf_AT): Handle DW_AT_addr_base.
7904 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7906 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7909 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7910 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7911 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7912 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7914 2018-04-28 Richard Biener <rguenther@suse.de>
7916 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7917 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7918 to reflect use. Only add interesting stmts.
7920 2018-04-27 Martin Jambor <mjambor@suse.cz>
7923 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7924 the jump function allows for passing through aggregate values.
7926 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7928 * input.h (in_system_header_at): Convert from macro to inline
7930 (from_macro_expansion_at): Likewise.
7931 (from_macro_definition_at): Likewise.
7933 2018-04-27 Jeff Law <law@redhat.com>
7935 * config.gcc: Mark tile* targets as deprecated/obsolete.
7937 2018-04-27 Richard Biener <rguenther@suse.de>
7939 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7942 2018-04-27 Richard Biener <rguenther@suse.de>
7944 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7946 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7948 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7949 with Yd constraint. Set "preferred_for_speed" attribute from
7950 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7952 (*movdi_internal): Ditto.
7953 (movti_interunit splitters): Remove
7954 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7955 (movdi_interunit splitters): Ditto.
7956 * config/i386/constraints.md (Ye): Remove.
7957 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7959 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7962 * config/aarch64/constraints.md (Usg): Limit to 31.
7965 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7967 PR tree-optimization/85529
7968 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7969 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7970 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7971 zero extension or masking of the MSB bit.
7972 (optimize_range_tests): Add FIRST_BB argument, pass it through
7973 to optimize_range_tests_var_bound.
7974 (maybe_optimize_range_tests, reassociate_bb): Adjust
7975 optimize_range_tests callers.
7977 2018-04-26 Richard Biener <rguenther@suse.de>
7978 Jakub Jelinek <jakub@redhat.com>
7980 * cgraph.h (symbol_table): Just declare debug method here.
7981 * symtab.c (symbol_table::debug): Define.
7983 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7985 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7987 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7989 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7990 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7991 (*movdi_internal): Substitute Yi and Yj constraint with x
7992 and Ym and Yn constraint with y constraint. Update "isa"
7993 attribute and set "preferred_for_speed" attribute from
7994 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7995 (*movsi_internal): Ditto.
7996 (*movdf_internal): Ditto.
7997 (*movsf_internal): Ditto.
7998 (*zero_extendsidi2): Ditto.
7999 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8000 (sse2_loadld): Ditto.
8001 (*vec_extract<ssevecmodelower>_0): Ditto.
8002 (*vec_extractv4si_0_zext_sse4): Ditto.
8003 (vec_concatv2di): Ditto.
8004 (*vec_dup<mode>): Ditto.
8005 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8006 * config/i386/constraints.md (Yi): Remove.
8011 2018-04-26 Nathan Sidwell <nathan@acm.org>
8013 * dumpfile.c (dump_open): New.
8014 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8015 (dump_finish): Detect stdio/stderr by value not name.
8017 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8019 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8021 2018-04-26 Tom de Vries <tom@codesourcery.com>
8024 * config/nvptx/nvptx.c (verify_neutering_jumps)
8025 (verify_neutering_labels): New function
8026 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8028 2018-04-26 Tom de Vries <tom@codesourcery.com>
8031 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8032 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8035 2018-04-26 Richard Biener <rguenther@suse.de>
8036 Tom de Vries <tom@codesourcery.com>
8039 * lto-streamer-out.c (output_function): Fixup loops if required to match
8040 discovery done in the reader.
8042 2018-04-26 Richard Biener <rguenther@suse.de>
8044 PR tree-optimization/85116
8045 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8046 have a loop exit from the single latch predecessor. Remove
8047 case of header with just condition.
8048 (ch_base::copy_headers): Exclude infinite loops from any
8050 (pass_ch::execute): Record exits.
8052 2018-04-26 Richard Biener <rguenther@suse.de>
8054 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8055 prologue cost vector and pass it to vect_get_load_cost.
8056 (vect_get_peeling_costs_all_drs): Likewise.
8057 (vect_peeling_hash_get_lowest_cost): Likewise.
8058 (vect_enhance_data_refs_alignment): Likewise.
8060 2018-04-26 Richard Biener <rguenther@suse.de>
8063 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8064 checking of integer<->pointer conversions.
8065 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8066 sign-/zero-extending pointer types.
8067 (expand_omp_for_static_chunk): Likewise.
8069 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8070 Jean Lee <xiaoyur347@gmail.com>
8072 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8073 (TARGET_ASAN_SHADOW_OFFSET): Define.
8074 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8075 true for -fsanitize=address.
8077 2018-04-25 Mark Wielaard <mark@klomp.org>
8079 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8082 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8084 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8085 than "alu", remove explicit "memory" and "imm_disp" attributes.
8086 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8089 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8090 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8091 gen_lowpart_no_emit.
8093 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8096 * config/i386/i386.c (ix86_expand_builtin): Change memory
8097 operand to XI, extend p0 to Pmode.
8098 * config/i386/i386.md: Change unspec volatile and operand
8099 1 mode to XI, change operand 0 mode to P.
8101 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8103 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8104 GET_MODE_MASK before any checking.
8105 (nds32_can_use_bset_p): Likewise.
8106 (nds32_can_use_btgl_p): Likewise.
8108 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8110 * config/nds32/nds32-doubleword.md: New define_split pattern for
8111 illegal register number.
8113 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8115 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8117 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8119 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8121 2018-04-25 Richard Biener <rguenther@suse.de>
8123 * lto-streamer.h (LTO_major_version): Bump to 8.
8125 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8127 * BASE-VER: Set to 9.0.0.
8129 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8131 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8132 in __abskf2 and __powikf2.
8134 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8137 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8138 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8139 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8140 Use the above on operand 2. Reindent.
8141 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8143 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8146 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8148 * config/i386/i386.opt (mcet): Removed.
8149 * doc/install.texi: Remove -mcet documentation.
8150 * doc/invoke.texi: Likewise.
8152 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8155 * doc/install.texi: Remove -mcet from bootstrap-cet.
8157 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8160 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8161 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8165 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8166 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8167 containing a CONST_VECTOR.
8169 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8171 * doc/install.texi: Update newlib dependency for nvptx.
8173 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8176 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8177 instead of INTVAL when shifting x left.
8179 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8181 PR tree-optimization/85478
8182 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8183 vect_grouped_store_supported for single element vectors.
8185 2018-04-24 Richard Biener <rguenther@suse.de>
8188 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8189 load cost increase to the case of non-constant step.
8191 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8194 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8195 destination if any_malformed_asm.
8197 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8200 * expr.c (store_field): In the bitfield case, if the value comes from
8201 a function call and is returned in registers by means of a PARALLEL,
8202 do not change the mode of the temporary unless BLKmode and VOIDmode.
8204 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8206 PR rtl-optimization/85423
8207 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8208 dependencies to debug insns when the previous insn is non-debug.
8210 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8212 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8213 enums into a single definition.
8214 (fls): Fix predicates and printing.
8217 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8219 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8220 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8221 and short u6 immediate.
8222 (check_if_valid_sleep_operand): Remove.
8223 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8225 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8227 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8228 flag_always_save_lp condition.
8229 * config/nds32/nds32.opt (malways-save-lp): New option.
8231 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8233 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8234 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8235 * config/nds32/nds32.h
8236 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8237 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8239 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8241 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8242 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8244 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8245 Chung-Ju Wu <jasonwucj@gmail.com>
8247 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8249 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8250 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8252 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8254 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8256 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8258 * config/nds32/nds32-protos.h (nds32_data_alignment,
8259 nds32_local_alignment): Declare.
8260 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8261 nds32_local_alignment): New functions.
8262 (TARGET_CONSTANT_ALIGNMENT): Define.
8263 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8265 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8267 * config/nds32/nds32.c
8268 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8269 (TARGET_MODES_TIEABLE_P): Likewise.
8271 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8273 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8276 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8277 Chung-Ju Wu <jasonwucj@gmail.com>
8279 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8280 for unaligned access.
8281 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8283 * config/nds32/nds32-intrinsic.md: Likewise.
8284 * config/nds32/nds32_intrinsic.h: Likewise.
8285 * config/nds32/nds32.h (nds32_builtins): Likewise.
8286 * config/nds32/nds32.opt (munaligned-access): New option.
8287 * config/nds32/nds32.c (nds32_asm_file_start): Display
8288 flag_unaligned_access status.
8290 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8292 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8293 -mno-relax is present.
8294 * config/riscv/linux.h (LINK_SPEC): Ditto.
8296 2018-04-20 Martin Sebor <msebor@redhat.com>
8299 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8301 (gimple_fold_builtin_stxcpy_chk): Same.
8302 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8304 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8307 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8308 __powikf2 when long double is IEEE 128-bit.
8310 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8312 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8313 step to make sure stack always aligned.
8315 2018-04-20 Carl Love <cel@us.ibm.com>
8318 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8319 size check for arg0.
8321 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8322 Tom de Vries <tom@codesourcery.com>
8325 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8326 Emit insns for calls too.
8327 (nvptx_find_par): Always look for worker-level predecessor insn.
8328 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8330 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8331 (nvptx_process_pars): Propagate frames for calls.
8333 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8336 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8338 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8339 (ix86_handle_option): Don't handle OPT_mibt.
8340 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8342 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8344 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8345 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8346 (ix86_target_macros): Define __CET__ with flag_cf_protection
8347 for -fcf-protection.
8348 * config/i386/i386.c (isa2_opts): Remove -mibt.
8349 * config/i386/i386.h (TARGET_IBT): Removed.
8350 (TARGET_IBT_P): Likewise.
8351 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8352 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8353 * config/i386/i386.opt (mcet): Update help message.
8356 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8357 -mcet as an alias for -mshstk.
8359 2018-04-20 Richard Biener <rguenther@suse.de>
8362 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8363 complexity by forcing a single use of the multiply operand.
8365 2018-04-20 Martin Jambor <mjambor@suse.cz>
8368 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8369 recursion dependency to only apply to non-clones.
8371 2018-04-20 Martin Jambor <mjambor@suse.cz>
8374 * ipa-cp.c (create_specialized_node): Check that clones of
8375 self-recursive edges exist during IPA-CP.
8377 2018-04-19 Toon Moene <toon@moene.org>
8379 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8382 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8384 PR tree-optimization/85467
8385 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8386 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8387 VECTOR_CST element to type.
8389 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8392 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8393 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8394 (builtin_longjmp): Likewise.
8395 (save_stack_nonlocal): New pattern.
8396 (restore_stack_nonlocal): Likewise.
8398 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8401 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8402 Replace ASM_OUTPUT_LABEL with fprintf.
8404 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8407 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8408 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8409 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8410 define __IBT__ and __SHSTK__ for -fcf-protection.
8411 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8413 (ix86_trampoline_init): Likewise.
8414 (x86_output_mi_thunk): Likewise.
8415 (ix86_notrack_prefixed_insn_p): Likewise.
8416 (ix86_option_override_internal): Don't disallow -fcf-protection.
8417 * config/i386/i386.md (rdssp<mode>): Also enable for
8419 (incssp<mode>): Likewise.
8420 (nop_endbr): Likewise.
8421 * config/i386/i386.opt (mcet): Change help message to built-in
8425 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8426 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8427 enable CET built-in functions.
8429 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8431 * common/config/i386/i386-common.c
8432 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8433 OPTION_MASK_ISA_MOVDIRI_UNSET,
8434 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8435 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8436 * config.gcc (movdirintrin.h): New header.
8437 * config/i386/cpuid.h (bit_MOVDIRI,
8438 bit_MOVDIR64B): New bits.
8439 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8441 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8442 (VOID, PVOID, PCVOID)): New function types.
8443 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8444 __builtin_ia32_directstoreu_u64,
8445 __builtin_ia32_movdir64b): New builtins.
8446 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8447 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8449 (ix86_valid_target_attribute_inner_p): Ditto.
8450 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8451 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8452 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8453 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8454 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8455 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8456 (movdiri<mode>, movdir64b_<mode>): New.
8457 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8458 * config/i386/immintrin.h: Include movdirintrin.h.
8459 * config/i386/movdirintrin.h: New file.
8460 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8462 2018-04-19 Richard Biener <rguenther@suse.de>
8465 * cfg.c (clear_bb_flags): When loop state says we have
8466 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8468 2018-04-19 Richard Biener <rguenther@suse.de>
8470 PR tree-optimization/84737
8471 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8472 copying restrict info.
8473 (vect_setup_realignment): Use it.
8474 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8475 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8476 the first DR to all generated stores.
8477 (vectorizable_load): Likewise for loads.
8479 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8481 PR tree-optimization/85446
8482 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8483 the integral and pointer types to have the same precision.
8485 * doc/install.texi: Document --disable-cet being the default and
8488 2018-04-18 Martin Liska <mliska@suse.cz>
8490 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8493 2018-04-18 Martin Liska <mliska@suse.cz>
8496 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8499 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8500 arguments if they are comparable.
8502 2018-04-18 Martin Liska <mliska@suse.cz>
8505 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8508 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8511 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8514 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8515 ENDBR after calling __morestack.
8517 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8520 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8521 by using gcc_base_ver to generate a gcc_driver_version, and use
8522 it when generating GCC_DRIVER_NAME.
8523 * configure: Regenerate.
8525 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8528 * config.gcc: Obsolete powerpc*-*-*spe*.
8530 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8533 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8534 (stabstr_D): Change type of unum from unsigned int to
8535 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8538 2018-04-17 Jim Wilson <jimw@sifive.com>
8541 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8542 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8543 Set arg_pointer_offset after using pretend_args_size.
8545 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8547 PR rtl-optimization/85431
8548 * dse.c (record_store): Ignore zero width stores.
8551 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8552 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8553 __builtin_stack_restore rather than after it.
8554 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8555 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8556 argument instead of virtual_dynamic_stack_rtx.
8558 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8560 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8562 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8563 Add note to error message to explain internal mapping of overloaded
8564 built-in function name to non-overloaded built-in function name.
8565 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8568 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8571 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8572 where the inputs overlap with the output.
8574 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8576 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8577 (=v, v) alternative and explicit "memory" attribute.
8578 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8579 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8581 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8582 "sselog1" type instead of "sselog".
8583 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8584 "sselog". Remove explicit "memory" attribute.
8585 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8586 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8588 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8589 "sselog1" type instead of "sselog". Remove explicit "memory"
8591 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8592 use "sselog1" type instead of "sselog". Remove explicit "memory"
8594 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8595 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8597 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8598 "sselog1" type instead of "sselog". Remove explicit "memory"
8600 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8601 use "sselog1" type instead of "sselog". Remove explicit "memory"
8605 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8608 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8611 2018-04-17 Martin Jambor <mjambor@suse.cz>
8614 * ipa-cp.c (create_specialized_node): Call
8615 expand_all_artificial_thunks if necessary.
8617 2018-04-17 Martin Liska <mliska@suse.cz>
8620 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8621 in message, remote space in between '_G' and '('.
8623 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8626 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8627 avx512f_vmcmp<mode>3<round_saeonly_name>,
8628 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8629 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8630 avx512f_rndscale<mode><round_saeonly_name>,
8631 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8632 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8633 Use %<iptr>2 instead of %2 for -masm=intel.
8634 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8635 avx512f_vcvttss2usi<round_saeonly_name>,
8636 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8638 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8639 avx512f_vcvttsd2usi<round_saeonly_name>,
8640 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8641 Use %q1 instead of %1 for -masm=intel.
8642 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8643 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8644 of %3 for -masm=intel.
8645 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8647 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8649 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8651 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8652 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8654 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8656 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8657 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8658 %g1 and one with %0 and %1.
8659 (avx512er_vmrcp28<mode><round_saeonly_name>,
8660 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8662 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8663 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8664 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8665 of %0 and %{%4%} for -masm=intel.
8666 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8667 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8668 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8669 order of %0 and %{%5%}%{z%} for -masm=intel.
8671 2018-04-17 Jan Hubicka <jh@suse.cz>
8674 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8676 2018-04-17 Martin Liska <mliska@suse.cz>
8679 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8680 for target_clone error message. Make default implementation
8681 clone to be a local declaration.
8682 (separate_attrs): Add new argument and check for an empty
8684 (expand_target_clones): Handle it.
8685 (ipa_target_clone): Make redirection just for target_clones
8688 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8689 Tom de Vries <tom@codesourcery.com>
8692 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8693 tiled basic blocks without omp continue statements.
8695 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8698 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8699 vec_extract expression as having side effects to make sure it gets
8702 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8705 * config/i386/i386.c (get_builtin_code_for_version): Check
8708 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8711 * gcc/config.gcc: Support "skylake".
8712 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8714 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8715 (processor_target_table): Add "skylake".
8716 (ix86_option_override_internal): Add "skylake".
8717 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8718 PROCESSOR_CANNONLAKE.
8719 (get_builtin_code_for_version): Fix priority for
8720 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8721 PROCESSOR_SKYLAKE-AVX512.
8722 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8723 (processor_type): Add PROCESSOR_SKYLAKE.
8725 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8726 Jason Merrill <jason@redhat.com>
8729 * convert.c (convert_to_integer_1): Use direct recursion for
8730 enumeral types and types with a precision less than the number
8731 of bits in their mode.
8733 2018-04-16 Julia Koval <julia.koval@intel.com>
8736 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8737 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8739 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8742 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8743 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8744 and -mno-direct-move.
8746 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8749 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8750 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8751 Ensure negative shifts result in {0}.
8753 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8755 PR rtl-optimization/79916
8756 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8757 regs (if any) to define how to gnerate SD moves when LRA is in
8760 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8762 PR rtl-optimization/85393
8763 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8764 * except.c (expand_dw2_landing_pad_for_region): Make static.
8765 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8766 a label and unconditional jump to old_bb, rather than
8767 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8770 PR rtl-optimization/85376
8771 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8772 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8773 instead of a specific value.
8775 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8776 Bin Cheng <bin.cheng@arm.com>
8778 PR tree-optimization/82965
8779 PR tree-optimization/83991
8780 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8781 by_profile_only parameter.
8782 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8783 information if the loop was predicted to iterate too many times.
8784 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8786 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8789 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8792 2018-04-13 Martin Liska <mliska@suse.cz>
8793 Jakub Jelinek <jakub@redhat.com>
8796 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8797 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8798 * builtins.c (expand_builtin_memory_copy_args): Use
8799 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8800 handle dest_addr == pc_rtx.
8802 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8805 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8806 asked to not generate direct moves.
8807 (fix_trunc<mode>si2_stfiwx): Similar.
8808 (fix_trunc<mode>si2_internal): Similar.
8810 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8813 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8814 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8815 lookup if dest in some wider mode is known to be const0_rtx and
8816 if so, record permanent equivalence for it to be ZERO_EXTEND of
8817 the narrower mode destination.
8819 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8821 * lto-streamer-out.c (output_function): Revert 259346.
8822 * omp-expand.c (expand_oacc_for): Likewise.
8824 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8826 PR rtl-optimization/85354
8827 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8828 * sel-sched.c (sel_global_init): ... here.
8830 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8833 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8834 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8835 mode for PE-COFF targets.
8836 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8837 (i386_pe_asm_lto_end): Likewise.
8838 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8839 (TARGET_ASM_LTO_END): Likewise.
8840 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8841 (i386_pe_asm_lto_start): New function.
8842 (i386_pe_asm_lto_end): Likewise.
8844 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8845 Richard Biener <rguenther@suse.de>
8848 * lto-streamer-out.c (output_function): Fix CFG loop state before
8850 * omp-expand.c (expand_oacc_for): Handle calls to internal
8851 functions like regular functions.
8853 2018-04-12 Richard Biener <rguenther@suse.de>
8856 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8857 for the early LTO debug to properly generate references to it
8858 during DIE emission. Do not re-use that for the skeleton for
8860 (dwarf2out_early_finish): Likewise.
8862 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8865 * config/i386/sse.md
8866 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8867 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8868 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8869 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8870 and output is a reg, avoid creating invalid lowpart subreg, but
8871 instead split into a 512-bit move. Don't split if not AVX512VL,
8872 input is xmm16+ reg and output is a mem.
8873 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8874 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8875 xmm16+ reg and output is a mem.
8877 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8879 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8880 also for flag_dwarf2_cfi_asm.
8882 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8884 PR rtl-optimization/85342
8885 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8886 a bool scalar var inside of the loop instead. Don't try to update
8887 recog_data.operand after failed apply_change_group.
8889 2018-04-12 Tom de Vries <tom@codesourcery.com>
8892 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8893 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8894 array with flexible array member as array without given dimension.
8895 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8896 argument for undefined param to true.
8898 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8901 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8902 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8903 from PowerPC section.
8904 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8905 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8906 help text that is too long.
8907 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8908 help text that is too long.
8909 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8910 help text that is too long.
8912 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8914 * config/alpha/alpha.md (stack_probe_internal): Rename
8915 from "probe_stack". Update all callers.
8917 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8919 PR rtl-optimization/84566
8920 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8921 sched_macro_fuse_insns.
8923 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8926 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8927 (compute_block_dependences): ... from here.
8929 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8931 PR tree-optimization/85331
8932 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8933 from int to HOST_WIDE_INT.
8935 2018-04-11 Martin Jambor <mjambor@suse.cz>
8938 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8939 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8940 not the same as the source val.
8941 (cgraph_edge_brings_value_p): New parameter.
8942 (gather_edges_for_value): Pass destination value to
8943 cgraph_edge_brings_value_p.
8944 (perhaps_add_new_callers): Likewise.
8945 (get_info_about_necessary_edges): Likewise and exclude values brought
8946 only by self-recursive edges.
8947 (create_specialized_node): Redirect only clones of self-calling edges.
8948 (+self_recursive_pass_through_p): New function.
8949 (find_more_scalar_values_for_callers_subset): Use it.
8950 (find_aggregate_values_for_callers_subset): Likewise.
8951 (known_aggs_to_agg_replacement_list): Removed.
8952 (decide_whether_version_node): Re-calculate known constants for all
8953 remaining context clones.
8955 2018-04-11 Richard Biener <rguenther@suse.de>
8958 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8959 from early DWARF output.
8960 (dwarf2out_early_finish): Output line info unconditionally into
8961 early DWARF and add reference to it.
8963 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8966 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8967 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8968 other than V2DFmode using iptr mode attribute.
8969 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8971 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8973 PR rtl-optimization/84659
8974 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8976 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8979 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8981 (output_loc_list): Pass address of a dummy size variable even in the
8982 locview handling loop.
8983 (index_location_lists): Add comment on why skip_loc_list_entry can't
8986 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8989 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8992 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8995 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8996 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8997 and -mstring-compare-inline-limit.
8999 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9002 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9003 for stack clash protection in a register whenever we need it to be in
9006 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9008 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9009 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9011 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9014 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9016 (mlong-double-): Ditto.
9017 * config/rs6000/sysv4.opt (msdata=): Ditto.
9018 (mtls-size=): Ditto.
9020 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9022 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9023 erroneous entries for
9024 "vector int vec_ldl (int, long int *)", and
9025 "vector unsigned int vec_ldl (int, unsigned long int *)".
9026 Add comments and entries for
9027 "vector bool char vec_ldl (int, bool char *)",
9028 "vector bool short vec_ldl (int, bool short *)",
9029 "vector bool int vec_ldl (int, bool int *)",
9030 "vector bool long long vec_ldl (int, bool long long *)",
9031 "vector pixel vec_ldl (int, pixel *)",
9032 "vector long long vec_ldl (int, long long *)",
9033 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9034 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9035 type tree bool_long_long_type_node and correct definition of
9036 bool_V2DI_type_node to make reference to this new type tree.
9037 (rs6000_mangle_type): Replace erroneous reference to
9038 bool_long_type_node with bool_long_long_type_node.
9039 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9040 comments to emphasize sign distinctions for char and int types and
9041 replace RS6000_BTI_bool_long constant with
9042 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9043 use of RS6000_BTI_pixel.
9044 (bool_long_type_node): Remove this macro definition.
9045 (bool_long_long_type_node): New macro definition
9047 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9049 PR rtl-optimization/85300
9050 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9051 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9052 simplify_unary_operation fails.
9054 2018-04-10 Martin Liska <mliska@suse.cz>
9056 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9057 cgraph_edge and ipa_ref.
9059 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9063 * config/i386/sse.md
9064 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9065 computation of the VEC_MERGE selector from mask.
9066 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9067 Fix decoding of the VEC_MERGE selector into mask.
9069 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9071 PR tree-optimization/85286
9072 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9074 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9076 * final.c (final_1): Set insn_last_address as well as
9077 insn_current_address.
9079 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9082 * explow.c (emit_stack_probe): Call validize_mem on memory location
9083 before passing it to gen_probe_stack. Create address operand and
9084 legitimize it for the probe_stack_address case.
9086 2018-04-09 Jan Hubicka <jh@suse.cz>
9089 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9090 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9091 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9093 (free_lang_data): Rebuild type inheritance graph.
9095 2018-04-09 Martin Sebor <msebor@redhat.com>
9097 * invoke.texi (-finline-small-functions): Mention other optimization
9099 (-findirect-inlining, -fpartial-inlining): Same.
9100 (-finline-functions-called-once): Same.
9101 (-freorder-blocks-and-partition): Same.
9103 2018-04-09 Jan Hubicka <jh@suse.cz>
9106 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9107 jumps; choose last target that matches the criteria (i.e.
9108 no partition changes for non-crossing jumps).
9109 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9110 support for redirecting crossing jumps to non-crossing.
9112 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9114 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9115 also for naked functions.
9117 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9119 * config/arc/arc.md (add_shift): New pattern.
9120 (add_shift2): Likewise.
9121 (sub_shift): Likewise.
9122 (sub_shift_cmp0_noout): Likewise.
9123 (compare_si_ashiftsi): Likewise.
9124 (xbfu_cmp0_noout): New combine pattern.
9125 (xbfu_cmp0"): Likewise.
9126 (movsi_set_cc_insn): Place the predicable variant first.
9127 (commutative_binary_cmp0_noout): Remove clobber.
9128 (commutative_binary_cmp0): New pattern.
9129 (noncommutative_binary_cmp0): Likewise.
9130 (noncommutative_binary_cmp0_noout): Likewise.
9131 (noncommutative_binary_comparison_result_used): Removed.
9132 (rsub_cmp0): New pattern.
9133 (rsub_cmp0_noout): Likewise.
9134 (extzvsi): Changed, keep only meaningful variants.
9135 (SQH, SEZ): New iterators.
9136 (SQH_postfix): New mode attribute.
9137 (SEZ_prefix): New code attribute.
9138 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9139 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9140 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9142 (noncommutative_operator): Check the availability of barrel
9145 2018-04-09 Richard Biener <rguenther@suse.de>
9147 PR tree-optimization/85284
9148 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9149 Only use the niter constraining form of simple_iv when the exit
9152 2018-04-09 Tom de Vries <tom@codesourcery.com>
9155 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9156 (define_expand "*memory_barrier"): New define_expand.
9157 (define_insn "memory_barrier"): New insn.
9159 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9161 PR rtl-optimization/80463
9162 PR rtl-optimization/83972
9163 PR rtl-optimization/83480
9165 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9166 correct producer for the insn.
9167 (tidy_control_flow): Fixup seqnos in case of debug insns.
9169 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9171 PR rtl-optimization/83913
9173 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9174 different sched-times when merging exprs.
9176 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9178 PR rtl-optimization/83962
9180 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9181 tidy_fallthru_edge and tidy_control_flow.
9183 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9185 PR rtl-optimization/83530
9187 * sel-sched.c (force_next_insn): New global variable.
9188 (remove_insn_for_debug): When force_next_insn is true, also leave only
9189 next insn in the ready list.
9190 (sel_sched_region): When the region wasn't scheduled, make another pass
9191 over it with force_next_insn set to 1.
9193 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9195 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9197 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9198 for interrupt control.
9199 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9200 functions for interrupt control.
9201 * config/nds32/nds32-intrinsic.md: Likewise.
9202 * config/nds32/nds32_intrinsic.h: Likewise.
9203 * config/nds32/nds32.h (nds32_builtins): Likewise.
9205 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9207 * config/nds32/nds32.c (nds32_init_machine_status,
9208 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9209 strict_aligned_p field.
9210 (nds32_expand_to_rtl_hook): New function.
9211 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9212 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9214 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9215 Chung-Ju Wu <jasonwucj@gmail.com>
9217 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9218 * config/nds32/nds32-n7.md: New file.
9219 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9220 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9222 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9223 * config/nds32/nds32.md (pipeline_model): Add n7.
9224 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9225 * config/nds32/pipelines.md: Include n7 settings.
9227 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9228 Chung-Ju Wu <jasonwucj@gmail.com>
9230 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9231 * config/nds32/nds32-e8.md: New file.
9232 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9233 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9235 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9236 * config/nds32/nds32.md (pipeline_model): Add e8.
9237 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9238 * config/nds32/pipelines.md: Include e8 settings.
9240 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9241 Chung-Ju Wu <jasonwucj@gmail.com>
9243 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9244 * config/nds32/nds32-n8.md: New file.
9245 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9246 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9248 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9249 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9250 * config/nds32/nds32.md (pipeline_model): Add n8.
9251 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9252 * config/nds32/pipelines.md: Include n8 settings.
9254 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9255 Chung-Ju Wu <jasonwucj@gmail.com>
9257 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9258 * config/nds32/nds32-n9-2r1w.md: New file.
9259 * config/nds32/nds32-n9-3r2w.md: New file.
9260 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9261 nds32_register_ports): New or modify for cpu n9.
9262 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9264 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9265 * config/nds32/nds32-utils.c: New file.
9266 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9267 TARGET_MUL_SLOW): Define.
9268 * config/nds32/nds32.md (pipeline_model): New attribute.
9269 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9270 New options that support cpu n9.
9271 * config/nds32/pipelines.md: Include n9 settings.
9272 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9274 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9276 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9277 information if necessary.
9278 (output_cond_branch_compare_zero): Likewise.
9279 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9280 (nds32_target_alignment): Refine for alignment.
9281 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9282 (FUNCTION_BOUNDARY): Modify.
9283 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9285 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9287 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9289 * config/nds32/constants.md (unspec_volatile_element): Add values for
9290 TLB operation and data prefetch.
9291 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9292 functions for TLB operation and data prefetch.
9293 * config/nds32/nds32-intrinsic.md: Likewise.
9294 * config/nds32/nds32_intrinsic.h: Likewise.
9295 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9296 (nds32_print_operand): Likewise.
9297 * config/nds32/nds32.h (nds32_builtins): Likewise.
9299 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9300 Andrew Pinski <pinsika@gcc.gnu.org>
9303 * match.pd: Use constant_boolean_node of correct type instead of
9304 boolean_true_node or boolean_false_node for simplifying
9305 pointer comparisons to zero.
9307 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9309 PR tree-optimization/80021
9310 * tree.c (verify_type_variant): Make error call in verify_variant_match
9311 translatable and remove final full stop.
9313 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9315 * config/nds32/constants.md (unspec_volatile_element): Add
9316 UNSPEC_VOLATILE_EH_RETURN.
9317 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9318 nds32_output_stack_pop): Support dwarf exception handling process.
9319 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9320 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9321 exception handling process.
9322 (nds32_compute_stack_frame): Likewise.
9323 (nds32_return_addr_rtx): Likewise.
9324 (nds32_initial_elimination_offset): Likewise.
9325 (nds32_expand_prologue): Likewise.
9326 (nds32_expand_epilogue): Likewise.
9327 (nds32_dynamic_chain_address): New function.
9328 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9330 (DYNAMIC_CHAIN_ADDRESS): Define.
9331 (EH_RETURN_DATA_REGNO): Define.
9332 (EH_RETURN_STACKADJ_RTX): Define.
9333 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9334 patterns for dwarf exception handling.
9336 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9338 * config/nds32/nds32.h: Clean up obsolete macros.
9340 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9342 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9343 Add enum values for particular instructions.
9344 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9345 particular intrinsic functions.
9346 * config/nds32/nds32-intrinsic.md: Likewise.
9347 * config/nds32/nds32_intrinsic.h: Likewise.
9348 * config/nds32/nds32.h (nds32_builtins): Likewise.
9349 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9350 (btst, ave): New patterns for particular instructions.
9352 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9354 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9355 Add enum values for atomic load/store and memory sync.
9356 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9358 * config/nds32/nds32-intrinsic.md: Likewise.
9359 * config/nds32/nds32_intrinsic.h: Likewise.
9360 * config/nds32/nds32.h (nds32_builtins): Likewise.
9362 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9364 PR tree-optimization/85257
9365 * fold-const.c (native_encode_vector): If not all elts could fit
9366 and off is -1, return 0 rather than offset.
9367 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9368 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9369 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9370 adjust buffer in native_interpret_expr call.
9372 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9374 * config/nds32/constants.md (unspec_volatile_element): Add cache
9375 control enum values.
9376 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9377 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9378 * config/nds32/nds32.c (nds32_cctl_names): New.
9379 (nds32_print_operand): Handle cache control register names.
9380 * config/nds32/nds32.h (nds32_builtins): New enum values.
9381 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9383 * config/nds32/nds32.md (type): Add mmu.
9384 * config/nds32/pipelines.md (simple_insn): Add mmu.
9386 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9388 * config/nds32/nds32.md (type): Remove call.
9389 * config/nds32/pipelines.md (simple_insn): Likewise.
9391 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9393 * config/nds32/constants.md (unspec_volatile_element): Add
9394 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9395 UNSPEC_VOLATILE_FMFCFG.
9396 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9397 description for fmfcfg and fmfcsr.
9398 (bdesc_1arg): Add fmtcsr.
9399 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9400 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9401 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9402 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9403 unspec_fmfcfg): New patterns.
9404 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9405 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9406 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9407 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9408 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9409 __nds32__fmfcfg): Define.
9411 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9413 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9414 intrinsic register names.
9415 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9416 intrinsic register enum values and macros.
9418 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9420 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9421 for load/store addressing form.
9422 (nds32_print_operand_address): Likewise.
9424 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9427 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9428 based on LABEL_REF. Remove useless assertion.
9429 (pic_address_needs_scratch): Fix formatting.
9430 (sparc_legitimize_pic_address): Minor tweaks.
9431 (sparc_delegitimize_address): Adjust assertion accordingly.
9432 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9433 into symbolic_operand.
9434 (movsi_high_pic_label_ref): Likewise.
9435 (movsi_lo_sum_pic_label_ref): Likewise.
9436 (movdi_pic_label_ref): Likewise.
9437 (movdi_high_pic_label_ref): Likewise.
9438 (movdi_lo_sum_pic_label_ref): Likewise.
9440 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9442 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9443 custom LIB_SPEC setup.
9445 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9446 Kito Cheng <kito.cheng@gmail.com>
9448 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9449 * config/riscv/freebsd.h: New.
9451 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9453 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9454 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9457 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9458 Kito Cheng <kito.cheng@gmail.com>
9460 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9461 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9462 * config/nds32/nds32-protos.h (nds32_output_call,
9463 nds32_output_return): Declare.
9464 * config/nds32/nds32.md: Refine all the call and return patterns.
9466 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9469 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9470 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9472 PR rtl-optimization/84872
9473 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9474 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9477 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9479 * expr.c (copy_blkmode_to_reg): Revert 254862.
9480 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9482 2018-04-06 Richard Biener <rguenther@suse.de>
9485 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9486 after seeing a component reference with an adjacent field. Treat
9487 refs to arrays at struct end of external decls similar to
9488 refs to unconstrained commons.
9490 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9493 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9494 look through SAVE_EXPRs with non-side-effects argument. Adjust
9496 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9499 2018-04-06 Richard Biener <rguenther@suse.de>
9502 * alias.c (find_base_term): New wrapper around find_base_term
9503 unwinding CSELIB_VAL_PTR changes.
9504 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9507 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9509 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9511 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9512 constant definitions.
9513 ("nop"): lr 0,0 -> nopr r0
9514 ("nop_lr0", "nop_lr1"): New insn definitions.
9516 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9518 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9519 NDS32_V3PUSH_AVAILABLE_P macro.
9521 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9522 Chung-Ju Wu <jasonwucj@gmail.com>
9524 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9525 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9526 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9527 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9528 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9529 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9530 * config/nds32/constraints.md: New constraints and checking for hard
9531 float configuration.
9532 * config/nds32/iterators.md: New mode iterator and attribute for hard
9533 float configuration.
9534 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9536 * config/nds32/nds32-fpu.md: New file.
9537 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9538 deal with hard float code generation.
9539 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9541 (abi_type, float_reg_number): New enum type.
9542 * config/nds32/nds32-predicates.c: New predicates for hard float.
9543 * config/nds32/nds32-protos.h: Declare functions for hard float.
9544 * config/nds32/nds32.c: Implementation for hard float configuration.
9545 * config/nds32/nds32.h: Definitions for hard float configuration.
9546 * config/nds32/nds32.md: Include hard float machine description and
9547 modify patterns for hard float configuration.
9548 * config/nds32/nds32.opt: New options for hard float configuration.
9549 * config/nds32/predicates.md: New predicates for hard float
9552 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9554 * common/config/nds32/nds32-common.c
9555 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9557 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9560 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9561 CONSTRUCTOR_ELT (ctor, ...)->value.
9563 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9566 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9568 2018-04-05 Tom de Vries <tom@codesourcery.com>
9571 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9574 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9575 Kito Cheng <kito.cheng@gmail.com>
9577 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9578 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9579 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9580 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9582 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9583 Kito Cheng <kito.cheng@gmail.com>
9585 * config/nds32/constraints.md (Ufe): New memory constraint.
9586 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9587 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9588 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9590 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9591 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9593 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9595 * config/nds32/nds32.md: Use optimize_size in the condition for
9596 alu-shift instructions.
9598 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9600 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9602 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9604 * config/nds32/nds32.md (negsi2): Refine pattern.
9606 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9607 Chung-Ju Wu <jasonwucj@gmail.com>
9609 * config/nds32/iterators.md (shift_rotate): New code iterator.
9610 (shift): New code attribute.
9611 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9612 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9613 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9614 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9615 bit-wise operations.
9616 (andsi3, *andsi3): Ditto.
9617 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9618 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9619 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9620 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9621 nds32_ior_operand, nds32_xor_operand): New predicates.
9623 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9625 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9626 (addsi3, subsi3): ... this.
9628 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9630 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9632 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9634 * config/nds32/nds32.md: Adjust indention.
9636 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9638 * config/nds32/nds32.md (feature): New attribute.
9640 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9642 * config/nds32/nds32.md (subtype): New attribute.
9644 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9647 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9648 expansion to perform a bitwise AND of the argument followed by a
9649 boolean negation of the result.
9651 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9653 PR rtl-optimization/84878
9654 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9655 the basic block. Assert the use reference is not artificial and that
9656 it has an associated insn.
9658 2018-04-04 Michael Matz <matz@suse.de>
9660 * builtins.c (compute_objsize): Pass correct operand
9661 to array_at_struct_end_p.
9663 2018-04-04 Richard Biener <rguenther@suse.de>
9666 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9667 from contexts for DINFO_LEVEL_TERSE and below.
9669 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9671 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9672 resiter_operand condition.
9673 * config/nds32/nds32.md (*move<mode>): Ditto.
9675 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9676 Monk Chiang <sh.chiang04@gmail.com>
9678 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9680 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9682 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9684 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9685 Kito Cheng <kito.cheng@gmail.com>
9687 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9688 nds32_cond_code_str, output_cond_branch,
9689 output_cond_branch_compare_zero, nds32_expand_cbranch,
9690 nds32_expand_cstore, nds32_expand_movcc,
9691 nds32_output_cbranchsi4_equality_zero,
9692 nds32_output_cbranchsi4_equality_reg,
9693 nds32_output_cbranchsi4_equality_reg_or_const_int,
9694 nds32_output_cbranchsi4_greater_less_zero: New functions.
9695 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9696 nds32_expand_cstore, nds32_expand_movcc,
9697 nds32_output_cbranchsi4_equality_zero,
9698 nds32_output_cbranchsi4_equality_reg,
9699 nds32_output_cbranchsi4_equality_reg_or_const_int,
9700 nds32_output_cbranchsi4_greater_less_zero): Declare.
9701 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9702 nds32_rimm11s_operand): New predicates.
9703 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9704 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9707 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9709 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9710 * config/nds32/nds32.md: Ditto.
9711 * config/nds32/pipelines.md: Ditto.
9713 2018-04-04 Richard Biener <rguenther@suse.de>
9715 PR tree-optimization/85168
9716 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9717 propagating abnormals.
9719 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9721 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9723 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9724 Kito Cheng <kito.cheng@gmail.com>
9726 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9727 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9728 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9729 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9730 * config/nds32/nds32.md (sibcall_internal): New.
9731 (sibcall_register): Remove.
9732 (sibcall_immediate): Remove.
9733 (sibcall_value_internal): New.
9734 (sibcall_value_register): Remove.
9735 (sibcall_value_immediate): Remove.
9736 * config/nds32/predicates.md (nds32_general_register_operand): New.
9737 (nds32_call_address_operand): New.
9739 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9741 PR rtl-optimization/85167
9742 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9743 bb_defs if *split_p, instead preinitialize it to NULL.
9745 PR tree-optimization/85156
9746 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9747 evaluating the argument multiple times.
9749 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9751 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9753 (_mm_cvtpd_ps): Likewise.
9754 (_mm_cvttpd_epi32): Likewise.
9755 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9756 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9757 vector, pixel, and bool following altivec.h include.
9759 2018-04-03 Martin Sebor <msebor@redhat.com>
9761 * doc/extend.texi (Common Function Attributes): Clarify.
9762 (const attribute): Likewise.
9763 (pure attribute): Likewise.
9765 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9768 * config/i386/i386.c (ix86_expand_vector_set): Use
9769 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9771 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9773 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9774 instructions when changing rounding bits to preserve precision bits
9775 in the x87 control word.
9777 2018-04-03 Martin Liska <mliska@suse.cz>
9779 PR tree-optimization/82491
9780 * rtl.h (strip_offset_and_add): Replace += suboffset with
9781 poly_uint64 () + suboffset.
9783 2018-03-29 Martin Liska <mliska@suse.cz>
9784 Martin Jambor <mjambor@suse.cz>
9787 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9788 param_type is not an integral or pointer type.
9790 2018-04-03 Richard Biener <rguenther@suse.de>
9792 * sese.h (recompute_all_dominators): Remove.
9794 2018-04-02 Martin Sebor <msebor@redhat.com>
9796 * doc/invoke.texi (-Wrestrict): Fix typos.
9798 2018-04-02 Jim Wilson <jimw@sifive.com>
9800 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9801 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9802 (<optab>di3, <optab>si3_extend): Likewise.
9803 (<optab>si3_mask, <optab>si3_mask_1): New.
9804 (<optab>di3_mask, <optab>di3_mask_1): New.
9805 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9806 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9807 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9809 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9811 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9814 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9816 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9817 (nds32_canonicalize_comparison): New function.
9819 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9820 Kito Cheng <kito.cheng@gmail.com>
9821 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9823 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9824 * config/nds32/constants.md (unspec_volatile_element): Add
9825 UNSPEC_VOLATILE_RELAX_GROUP.
9826 * config/nds32/nds32-relax-opt.c: New file.
9827 * config/nds32/nds32-predicates.c
9828 (nds32_symbol_load_store_p): New function.
9829 * config/nds32/nds32-protos.h
9830 (nds32_symbol_load_store_p): Declare function.
9831 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9832 * config/nds32/nds32.c
9833 (nds32_register_pass): New function to register pass.
9834 (nds32_register_passes): New function to register passes.
9835 * config/nds32/nds32.md (relax_group): New pattern.
9836 * config/nds32/nds32.opt (mrelax-hint): New option.
9837 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9839 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9841 * config/nds32/t-nds32: Modify files dependency.
9843 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9845 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9846 (PROFILE_HOOK): Define its implementation.
9848 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9850 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9851 type and 32-bit size.
9853 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9856 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9857 (V_128_256): New mode iterator.
9858 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9859 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9860 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9862 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9863 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9865 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9868 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9869 NaN inputs correctly.
9871 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9874 * config/rs6000/vsx.md (??r): New mode attribute.
9875 (*vsx_mov<mode>_64bit): Use it.
9876 (*vsx_mov<mode>_32bit): Likewise.
9878 2018-03-30 Martin Sebor <msebor@redhat.com>
9880 PR tree-optimization/84818
9881 * builtins.c (check_access): Use warning_n.
9883 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9886 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9888 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9891 2018-03-30 Julia Koval <julia.koval@intel.com>
9894 * x86-tune.def (movx, partial_reg_dependency): Enable for
9897 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9900 * lra-constraints.c (process_alt_operands): Move setting
9901 this_alternative_matches below.
9903 2018-03-29 Martin Liska <mliska@suse.cz>
9906 * doc/invoke.texi: Document how LTO works with debug info.
9907 Describe auto-load support of binutils. Mention 'x86-64'
9908 as valid option value of -march option.
9910 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9912 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9915 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9916 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9917 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9920 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9923 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9924 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9925 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9926 for RS6000_BTM_POWERPC64.
9927 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9928 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9929 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9934 2018-03-28 Carl Love <cel@us.ibm.com>
9937 2017-09-27 Carl Love <cel@us.ibm.com>
9939 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9940 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9941 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9944 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9946 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9947 instead of __vector bool.
9948 (_mm_max_pu8): Likewise.
9949 (_mm_min_pi16): Likewise.
9951 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9954 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9955 (DIVWEUO): Likewise.
9957 (DIVDEUO): Likewise.
9958 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9959 DIVWEUO and DIVDEUO.
9960 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9961 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9962 (div_extend): Likewise.
9963 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9965 (__builtin_divweuo): Likewise.
9966 (__builtin_divdeo): Likewise.
9967 (__builtin_divdeuo): Likewise.
9969 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9972 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9973 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9975 PR tree-optimization/82004
9976 * gimple-match-head.c (optimize_pow_to_exp): New function.
9977 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9978 Don't fold to exp if optimize_pow_to_exp is false.
9980 2018-03-28 Martin Liska <mliska@suse.cz>
9983 * calls.c (initialize_argument_information): Fix trailing space.
9984 * common.opt: Fix typo and provide better explanation for
9985 -fsanitize-coverage option.
9986 * config/i386/i386.opt: Fix typo.
9988 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9989 Martin Liska <mliska@suse.cz>
9992 * gimplify.c (asan_poison_variable): Don't do the check for
9993 gimplify_omp_ctxp here.
9994 (gimplify_decl_expr): Do it here.
9995 (gimplify_target_expr): Likewise.
9997 2018-03-28 Martin Liska <mliska@suse.cz>
10000 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10001 chkp_type_bounds_count if MPX is not enabled.
10003 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10005 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10007 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10010 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10011 function to create the function decl for complex long double
10012 multiply and divide for -mabi=ieeelongdouble.
10013 (init_float128_ieee): Call it.
10015 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10018 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10019 -fcf-protection=branch -mibt.
10020 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10022 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10025 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10026 arm_word_relocations.
10028 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10031 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10032 extern array declarations.
10034 2018-03-27 Richard Biener <rguenther@suse.de>
10036 PR middle-end/84067
10037 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10038 explicit single_use checks.
10040 2018-03-27 Richard Biener <rguenther@suse.de>
10042 PR tree-optimization/85082
10043 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10046 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10048 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10049 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10050 Turn on fasynchronous-unwind-tables and funwind-tables.
10052 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10055 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10056 (*bmi_blsr_<mode>_ccz): Ditto.
10058 2018-03-26 Tom de Vries <tom@codesourcery.com>
10060 PR tree-optimization/85063
10061 * omp-general.c (offloading_function_p): New function. Factor out
10063 * omp-offload.c (pass_omp_target_link::gate): ... here.
10064 * omp-general.h (offloading_function_p): Declare.
10065 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10066 with attribute omp declare target for offloading functions.
10068 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10070 PR tree-optimization/84005
10071 * tree-data-ref.h (get_base_for_alignment): Declare.
10072 * tree-data-ref.c (get_base_for_alignment_1): New function.
10073 (get_base_for_alignment): Likewise.
10074 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10075 get_base_for_alignment to find a suitable base object, instead
10076 of always using drb->base_address.
10078 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10080 PR inline-asm/85022
10081 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10082 known size by default.
10084 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10086 PR inline-asm/85030
10087 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10088 and non BLKmode operands.
10090 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10093 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10094 Clean up attributes.
10096 2018-03-23 Richard Biener <rguenther@suse.de>
10099 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10100 we are going to emit early debug for LTO.
10102 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10104 PR inline-asm/85034
10105 * function.c (match_asm_constraints_1): Don't optimize if input
10106 doesn't satisfy general_operand predicate for output's mode.
10108 PR inline-asm/85022
10109 * alias.c (write_dependence_p): Don't require for x_canonicalized
10110 non-VOIDmode if x has VOIDmode.
10113 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10114 just don't try to optimize it rather than assert it never happens.
10116 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10118 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10119 macro expansions for definition of ST_INTERNAL_<mode> and
10120 LD_INTERNAL_<mode> builtins.
10121 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10123 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10125 (altivec_expand_st_builtin): Likewise.
10126 (altivec_expand_builtin): Remove calls to deleted functions.
10127 (rs6000_address_for_altivec): Delete this function.
10128 * config/rs6000/vector.md: Remove expands for
10129 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10131 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10134 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10135 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10136 re-computing once computed.
10137 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10138 (arm_init_machine_status): Initialize
10139 machine->static_chain_stack_bytes.
10141 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10144 * doc/extend.texi: Add four new prototypes for vec_ld.
10145 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10146 definitions for more logical presentation.
10147 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10148 entries for V1TI variants of __builtin_altivec_ld builtin.
10149 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10150 handling of V1TI variant of LVX icode pattern.
10151 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10152 (rs6000_gimple_fold_builtin): Likewise.
10153 (altivec_init_builtins): Add code to define
10154 __builtin_altivec_lvx_v1ti function.
10156 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10158 PR inline-asm/84941
10159 * function.c (match_asm_constraints_1): Don't do the optimization
10160 if input isn't a REG, SUBREG, MEM or constant.
10162 2018-03-22 Tom de Vries <tom@codesourcery.com>
10164 PR tree-optimization/84956
10165 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10166 bb_has_abnormal_pred.
10168 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10171 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10172 DECL_INITIAL (decl) to decl at the end.
10173 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10174 adjust the comment.
10176 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10178 * doc/extend.texi (__builtin_tgmath): Document when complex
10179 integer types are treated as _Complex _Float64.
10181 2018-03-21 Tom de Vries <tom@codesourcery.com>
10183 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10185 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10187 PR tree-optimization/84960
10188 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10189 if it is ENTRY block, move them into single succ of ENTRY in that case.
10191 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10193 PR tree-optimization/84811
10194 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10195 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10196 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10197 (known_subrange_p): Update accordingly. Cast each value involved
10198 in the size comparison, rather than casting the result of the
10201 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10203 PR tree-optimization/84982
10204 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10205 by flipping the least significant bit rather than all bits from
10206 bitpos to bitpos + bitsize - 1.
10208 2018-03-21 Nathan Sidwell <nathan@acm.org>
10210 * doc/extend.texi (Deprecated Features): Remove mention of
10211 long-deleted deprecations.
10213 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10216 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10217 * configure: Regenerate.
10219 2018-03-21 Tom de Vries <tom@codesourcery.com>
10221 PR tree-optimization/83126
10222 * tree-parloops.c (num_phis): New function.
10223 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10225 2018-03-21 Nathan Sidwell <nathan@acm.org>
10227 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10228 mention anon-struct/union members and trailing attributes.
10230 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10232 PR tree-optimization/84969
10233 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10234 builtin memset partitions if they set different rhs values.
10236 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10238 PR rtl-optimization/84989
10239 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10240 VEC_DUPLICATE with scalar result mode.
10242 2018-03-21 Martin Liska <mliska@suse.cz>
10245 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10246 not intended return statement.
10248 2018-03-21 Martin Liska <mliska@suse.cz>
10251 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10252 (chkp_find_bound_slots_1): Limit number of iterations.
10254 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10257 * Minor grammar fixes for x86 options.
10259 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10262 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10263 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10265 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10268 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10269 (altivec_lvx_<mode>_1op): Likewise.
10270 (altivec_stvx_<mode>_2op): Likewise.
10271 (altivec_stvx_<mode>_1op): Likewise.
10272 (altivec_lvx_<VM2:mode>): New define_expand.
10273 (altivec_stvx_<VM2:mode>): Likewise.
10274 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10275 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10276 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10277 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10278 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10279 (rs6000_gen_lvx): Likewise.
10280 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10281 (altivec_expand_stv_builtin): Likewise.
10282 (altivec_expand_builtin): Likewise.
10283 * config/rs6000/vector.md: Likewise.
10285 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10288 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10291 2018-03-20 Richard Biener <rguenther@suse.de>
10294 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10295 sign-conversions as zero, fall back to standard scalar_stmt
10298 2018-03-20 Martin Liska <mliska@suse.cz>
10301 * predict.c (rebuild_frequencies): Handle case when we have
10302 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10304 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10307 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10308 flag_section_anchors.
10309 * varasm.c (use_blocks_for_decl_p): Remove hack for
10310 dw2_force_const_mem.
10313 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10315 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10316 be created, use lowpart_subreg of operands[0] rather than operands[0]
10318 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10319 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10320 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10321 and n constraint instead of aarch64_shift_imm_di and Usd.
10322 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10323 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10325 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10328 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10329 to favor GPR over NEON registers.
10330 (<shift>di3_neon): Likewise.
10332 2018-03-20 Tom de Vries <tom@codesourcery.com>
10335 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10336 (nvptx_process_pars): Emit bar.sync asap and alap.
10338 2018-03-20 Tom de Vries <tom@codesourcery.com>
10341 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10342 seen_label if seen_label is already set.
10344 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10347 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10348 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10349 Use 1U instead of 1. Formatting fixes.
10352 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10353 instead of TREE_TYPE (s1) for the return value.
10355 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10357 PR tree-optimization/84946
10358 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10359 bitsize + bitsize in poly_uint64 rather than poly_int64.
10362 * dwarf2asm.c: Include fold-const.c.
10363 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10364 of decl rather than decl itself.
10366 PR rtl-optimization/84643
10367 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10369 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10372 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10373 calling assemble_variable.
10375 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10378 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10379 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10381 2018-03-19 Jim Wilson <jimw@sifive.com>
10384 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10385 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10386 (riscv_first_stack_step): Likewise.
10387 (riscv_option_override): Use STACK_BOUNDARY instead of
10388 MIN_STACK_BOUNDARY.
10389 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10390 MIN_STACK_BOUNDARY.
10391 (BIGGEST_ALIGNMENT): Set to 128.
10392 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10393 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10396 2018-03-19 Richard Biener <rguenther@suse.de>
10398 PR tree-optimization/84933
10399 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10400 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10402 2018-03-19 Richard Biener <rguenther@suse.de>
10404 PR tree-optimization/84859
10405 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10406 (cond_if_else_store_replacement): Perform sinking operation on
10407 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10408 Generalize what a BB with a single eligible store is.
10410 2018-03-19 Richard Biener <rguenther@suse.de>
10412 PR tree-optimization/84929
10413 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10414 chrec_is_positive against non-chrec arg.
10416 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10419 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10421 2018-03-18 Martin Liska <mliska@suse.cz>
10423 PR rtl-optimization/84635
10424 * regrename.c (build_def_use): Use matches_mode only when
10427 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10429 PR tree-optimization/84913
10430 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10431 vectorize chains of COND_EXPRs.
10433 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10435 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10437 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10439 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10441 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10443 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10445 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10446 Kito Cheng <kito.cheng@gmail.com>
10448 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10449 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10450 (nds32_adjust_reg_alloc_order): New function.
10451 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10453 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10455 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10456 nds32_print_operand, nds32_print_operand_address): Use
10457 HOST_WIDE_INT_PRINT_DEC instead.
10459 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10461 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10463 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10466 * config/i386/i386.c (initial_ix86_tune_features,
10467 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10468 unsigned long long.
10469 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10470 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10471 rather than 1u << ix86_tune. Formatting fix.
10472 (ix86_option_override_internal): Change ix86_arch_mask from
10473 unsigned int to unsigned HOST_WIDE_INT, initialize to
10474 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10475 (ix86_function_specific_restore): Likewise.
10477 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10480 * postreload.c (reload_combine_recognize_pattern): Perform
10481 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10482 truncate_int_for_mode the result for the destination's mode.
10485 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10486 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10488 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10491 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10494 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10495 regno_allocno_class_array and sorted_pseudos.
10496 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10497 insns where regno is used.
10499 2018-03-16 Martin Liska <mliska@suse.cz>
10502 * multiple_target.c (create_dispatcher_calls): Redirect
10503 reference in the symbol table.
10505 2018-03-16 Martin Liska <mliska@suse.cz>
10508 * multiple_target.c (create_dispatcher_calls): Redirect also
10511 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10515 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10516 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10517 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10519 2018-03-16 Julia Koval <julia.koval@intel.com>
10521 * doc/invoke.texi (Skylake Server): Add CLWB.
10522 Cannonlake): Remove CLWB.
10524 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10526 PR tree-optimization/84841
10527 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10529 (FLOAT_ONE_CONST_TYPE): Define.
10530 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10531 (sort_by_operand_rank): Put entries with higher constant_type last
10532 rather than first to match comments.
10534 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10536 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10539 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10542 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10543 check and use address of target always.
10545 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10548 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10549 (indirect_thunk_bnd_needed): Likewise.
10550 (indirect_thunks_used): Likewise.
10551 (indirect_thunks_bnd_used): Likewise.
10552 (indirect_return_needed): New.
10553 (indirect_return_bnd_needed): Likewise.
10554 (output_indirect_thunk_function): Add a bool argument for
10556 (output_indirect_thunk_function): Don't generate alias for
10557 function return thunk.
10558 (ix86_code_end): Call output_indirect_thunk_function to generate
10559 function return thunks.
10560 (ix86_output_function_return): Set indirect_return_bnd_needed
10561 and indirect_return_needed instead of indirect_thunk_bnd_needed
10562 and indirect_thunk_needed.
10564 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10566 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10567 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10568 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10570 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10571 Paul Hua <paul.hua.gm@gmail.com>
10574 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10576 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10578 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10579 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10580 resp. SFmode cases.
10582 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10585 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10586 instead of GET_MODE_SIZE when comparing Units.
10588 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10591 * varasm.c (hash_section): Return an unchangeble hash value
10592 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10593 Return !aarch64_can_use_per_function_literal_pools_p ().
10595 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10598 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10599 rather than address of cmode as last argument to prepare_cmp_insn.
10601 2018-03-15 Julia Koval <julia.koval@intel.com>
10603 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10604 F_AVX512VNNI, F_AVX512BITALG): New.
10606 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10609 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10610 insn for floating-point loads and stores.
10612 2018-03-14 Carl Love <cel@us.ibm.com>
10614 * config/rs6000/rs6000-c.c: Add macro definitions for
10615 ALTIVEC_BUILTIN_VEC_PERMXOR.
10616 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10617 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10618 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10619 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10621 * config/doc/extend.texi: Add prototypes for vec_permxor.
10623 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10626 * diagnostic-show-locus.c (class layout_point): Convert m_line
10627 from int to linenum_type.
10628 (line_span::comparator): Use linenum "compare" function when
10629 comparing line numbers.
10630 (test_line_span): New function.
10631 (layout_range::contains_point): Convert param "row" from int to
10633 (layout_range::intersects_line_p): Likewise.
10634 (layout::will_show_line_p): Likewise.
10635 (layout::print_source_line): Likewise.
10636 (layout::should_print_annotation_line_p): Likewise.
10637 (layout::print_annotation_line): Likewise.
10638 (layout::print_leading_fixits): Likewise.
10639 (layout::annotation_line_showed_range_p): Likewise.
10640 (struct line_corrections): Likewise for field m_row.
10641 (line_corrections::line_corrections): Likewise for param "row".
10642 (layout::print_trailing_fixits): Likewise.
10643 (layout::get_state_at_point): Likewise.
10644 (layout::get_x_bound_for_row): Likewise.
10645 (layout::print_line): Likewise.
10646 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10647 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10648 * input.c (selftest::test_linenum_comparisons): New function.
10649 (selftest::input_c_tests): Call it.
10650 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10651 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10652 * selftest.h (ASSERT_GT): New macro.
10653 (ASSERT_GT_AT): New macro.
10654 (ASSERT_LT): New macro.
10655 (ASSERT_LT_AT): New macro.
10657 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10659 PR rtl-optimization/84780
10660 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10662 2018-03-14 Martin Liska <mliska@suse.cz>
10664 * tree.c (record_node_allocation_statistics): Use
10665 get_stats_node_kind.
10666 (get_stats_node_kind): New function extracted from
10667 record_node_allocation_statistics.
10668 (free_node): Use get_stats_node_kind.
10670 2018-03-14 Richard Biener <rguenther@suse.de>
10672 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10673 that the value-set of ANTIC_IN doesn't grow.
10676 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10678 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10679 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10681 2018-03-14 Julia Koval <julia.koval@intel.com>
10683 * config.gcc (icelake-client, icelake-server): New.
10685 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10686 (initial_ix86_arch_features): Ditto.
10687 (PTA_SKYLAKE): Add SGX.
10688 (PTA_ICELAKE): Remove.
10689 (PTA_ICELAKE_CLIENT): New.
10690 (PTA_ICELAKE_SERVER): New.
10691 (ix86_option_override_internal): Split up icelake on icelake client and
10693 (get_builtin_code_for_version): Ditto.
10694 (fold_builtin_cpu): Ditto.
10695 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10696 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10697 * config/i386/i386.h (processor_type): Ditto.
10698 * doc/invoke.texi: Ditto.
10700 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10703 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10704 INTEGER_CST offset, add it together with bitpos / 8 and
10705 sign extend based on POINTER_SIZE.
10709 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10712 * config/i386/constraints.md (Yc): New register constraint.
10713 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10714 Use Yc constraint for alternative 2 of operand 0. Remove
10715 preferred_for_speed attribute.
10717 2018-03-14 Richard Biener <rguenther@suse.de>
10719 PR tree-optimization/84830
10720 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10721 with the old one to avoid oscillations.
10723 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10726 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10728 (assign_by_spills): Return a flag of reload assignment failure.
10729 Do not process the reload assignment failures. Do not spill other
10730 reload pseudos if they has the same reg class. Update n if
10732 (lra_assign): Add a return arg. Set up from the result of
10733 assign_by_spills call.
10734 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10735 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10736 usage_insns if it is not NULL.
10737 (spill_hard_reg_in_range): New function.
10738 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10739 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10740 function prototypes.
10741 (lra_assign): Change prototype.
10742 * lra.c (lra): Add code to deal with fails by splitting hard reg
10745 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10747 * config/riscv/riscv.opt (mrelax): New option.
10748 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10749 "norelax" when riscv_mrelax is disabled.
10750 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10752 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10755 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10756 reassociation for int modes.
10758 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10760 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10761 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10763 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10764 * config/aarch64/aarch64-sve.md
10765 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10766 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10767 (*extend<mode><Vwide>2): Rename to...
10768 (aarch64_sve_extend<mode><Vwide>2): ...this.
10769 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10770 renaming the old pattern to...
10771 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10773 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10774 define_expand, renaming the old pattern to...
10775 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10776 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10777 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10778 account when deciding which SVE instruction the optab should use.
10779 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10781 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10783 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10784 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10785 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10786 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10787 (tlsdesc_small_<mode>): Turn a define_expand and use
10788 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10789 (tlsdesc_small_advsimd_<mode>): ...this.
10790 (tlsdesc_small_sve_<mode>): New pattern.
10792 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10794 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10795 (UNSPEC_UMUL_HIGHPART): New constants.
10796 (MUL_HIGHPART): New int iteraor.
10797 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10798 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10800 (*<su>mul<mode>3_highpart): New define_insn.
10802 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10805 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10808 2018-03-13 Martin Liska <mliska@suse.cz>
10811 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10813 (sem_item_optimizer::~sem_item_optimizer): Release it.
10814 (sem_item_optimizer::merge_classes): Register variable aliases.
10815 (sem_item_optimizer::fixup_pt_set): New function.
10816 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10817 * ipa-icf.h: Declare new variables and functions.
10819 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10821 PR middle-end/84834
10822 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10823 integer_pow2p@2 and test integer_pow2p in condition.
10824 (A < 0 ? C : 0): Similarly for @1.
10826 PR middle-end/84831
10827 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10828 characters starting at p contain '\0' character, don't look beyond
10832 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10833 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10836 * reg-stack.c (change_stack): Change update_end var from int to
10837 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10838 also call set_block_for_insn on the newly added insns and rescan.
10841 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10842 on the last operand.
10845 * tree.c (stabilize_reference_1): Return save_expr (e) for
10846 STATEMENT_LIST even if it doesn't have side-effects.
10848 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10850 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10852 2018-03-12 Renlin Li <renlin.li@arm.com>
10854 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10855 aarch64_output_scalar_simd_mov_immediate.
10857 2018-03-12 Martin Sebor <msebor@redhat.com>
10859 PR tree-optimization/83456
10860 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10861 for perfectly overlapping calls to memcpy.
10862 (gimple_fold_builtin_memory_chk): Same.
10863 (gimple_fold_builtin_strcpy): Handle no-warning.
10864 (gimple_fold_builtin_stxcpy_chk): Same.
10865 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10867 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10869 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10870 parameter. Use it for SFmode.
10871 (rs6000_function_arg_advance_1): Adjust.
10872 (rs6000_function_arg): Adjust.
10873 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10875 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10877 PR rtl-optimization/84169
10878 PR rtl-optimization/84780
10879 * combine.c (can_combine_p): Check for a 2-insn combination whether
10880 the destination register is used between the two insns, too.
10882 2018-03-12 Richard Biener <rguenther@suse.de>
10884 PR tree-optimization/84803
10885 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10886 for refs DR analysis didn't process.
10888 2018-03-12 Richard Biener <rguenther@suse.de>
10890 PR tree-optimization/84777
10891 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10892 force-vectorize loops ignore whether we are optimizing for size.
10894 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10896 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10897 (TARGET_MD_ASM_ADJUST): Define.
10899 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10900 Kito Cheng <kito.cheng@gmail.com>
10901 Chung-Ju Wu <jasonwucj@gmail.com>
10903 * config/nds32/nds32.c (nds32_compute_stack_frame,
10904 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10905 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10906 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10907 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10908 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10909 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10910 * config/nds32/nds32.md (prologue, epilogue): Use macro
10911 NDS32_V3PUSH_AVAILABLE_P to do checking.
10913 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10916 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10917 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10918 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10919 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10920 addition of most attributes on !orig_type_die or the attribute not
10921 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10923 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10924 Chung-Ju Wu <jasonwucj@gmail.com>
10926 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10927 __NDS32_VH__ macro.
10928 * config/nds32/nds32.opt (mvh): New option.
10930 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10931 Chung-Ju Wu <jasonwucj@gmail.com>
10933 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10935 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10936 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10939 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10940 Chung-Ju Wu <jasonwucj@gmail.com>
10942 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10944 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10945 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10947 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10948 Kito Cheng <kito.cheng@gmail.com>
10949 Chung-Ju Wu <jasonwucj@gmail.com>
10951 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10952 UNSPEC_FFMISM and UNSPEC_FLMISM.
10953 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10954 for ffb, ffmism and flmism.
10955 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10956 (unspec_ffmism): Ditto.
10957 (unspec_flmism): Ditto.
10958 (nds32_expand_builtin_impl): Check if string extension is available.
10959 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10960 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10962 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10965 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10968 * lra-assigns.c (assign_by_spills): Return a flag of reload
10969 assignment failure. Do not process the reload assignment
10970 failures. Do not spill other reload pseudos if they has the same
10972 (lra_assign): Add a return arg. Set up from the result of
10973 assign_by_spills call.
10974 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10975 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10976 usage_insns if it is not NULL.
10977 (spill_hard_reg_in_range): New function.
10978 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10979 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10980 function prototypes.
10981 (lra_assign): Change prototype.
10982 * lra.c (lra): Add code to deal with fails by splitting hard reg
10985 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10988 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10990 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10993 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10994 (dw_val_node): Add val_symbolic_view.
10995 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10996 (symview_upper_bound): New.
10997 (new_line_info_table): Initialize symviews_since_reset.
10998 (dwarf2out_source_line): Count symviews_since_reset and set
10999 symview_upper_bound.
11000 (dw_val_equal_p): Handle symview.
11001 (add_AT_symview): New.
11002 (print_dw_val): Handle symview.
11003 (attr_checksum, attr_checksum_ordered): Likewise.
11004 (same_dw_val_p, size_of_die): Likewise.
11005 (value_format, output_die): Likewise.
11006 (add_high_low_attributes): Use add_AT_symview for entry_view.
11007 (dwarf2out_finish): Reset symview_upper_bound, clear
11010 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11013 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11014 Add strict argument and use it.
11015 (rs6000_split_multireg_move): Update for new strict argument.
11016 (mem_operand_gpr): Disallow all non-offsettable addresses.
11017 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11019 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11022 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11023 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11024 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11027 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11028 decl, use remap_type if we want to use the type.
11030 2018-03-09 Martin Sebor <msebor@redhat.com>
11032 PR tree-optimization/84526
11033 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11035 (builtin_access::generic_overlap): Be prepared to handle non-array
11038 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11040 PR rtl-optimization/84682
11041 * lra-constraints.c (process_address_1): Check is_address flag
11042 for address constraints.
11043 (process_alt_operands): Likewise.
11044 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11045 preprocess_constraints.
11046 * recog.h (preprocess_constraints): Add oploc parameter.
11048 * recog.c (preprocess_constraints): Test address_operand for
11049 CT_ADDRESS constraints.
11051 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11054 * lra-assigns.c (assign_by_spills): Return a flag of reload
11055 assignment failure. Do not process the reload assignment
11056 failures. Do not spill other reload pseudos if they has the same
11058 (lra_assign): Add a return arg. Set up from the result of
11059 assign_by_spills call.
11060 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11061 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11062 usage_insns if it is not NULL.
11063 (spill_hard_reg_in_range): New function.
11064 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11065 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11066 function prototypes.
11067 (lra_assign): Change prototype.
11068 * lra.c (lra): Add code to deal with fails by splitting hard reg
11071 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11074 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11075 Accept complain bool parameter. Only emit errors if it is true.
11076 (arm_parse_cpu_option_name): Likewise.
11077 (arm_target_thumb_only): Adjust callers of the above.
11078 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11079 prototype to take a default true bool parameter.
11080 (arm_parse_arch_option_name): Likewise.
11082 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11083 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11087 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11088 * configure: Regenerate.
11089 * configure.ac ("linker --version-script option"): New.
11090 ("linker soname option"): New.
11092 2018-03-09 Richard Biener <rguenther@suse.de>
11094 PR tree-optimization/84775
11095 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11096 immediate uses of predicate stmts and mark them modified.
11099 PR tree-optimization/84178
11100 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11102 (version_loop_for_if_conversion): Delay update_ssa call.
11103 (tree_if_conversion): Delay update_ssa until after predicate
11106 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11109 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11110 when the function accesses prior frames.
11112 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11115 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11116 gen_llsym, otherwise call maybe_gen_llsym.
11118 PR inline-asm/84742
11119 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11120 has ',' character inside of it.
11122 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11125 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11126 as clobbering CC_REGNUM.
11128 2018-03-08 Richard Biener <rguenther@suse.de>
11130 PR middle-end/84552
11131 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11132 (follow_copies_to_constant): Do not follow SSA names registered
11135 2018-03-08 Richard Biener <rguenther@suse.de>
11137 PR tree-optimization/84178
11138 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11140 (version_loop_for_if_conversion): Delay update_ssa call.
11141 (tree_if_conversion): Delay update_ssa until after predicate
11144 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11146 PR tree-optimization/84178
11147 * tree-if-conv.c (release_bb_predicate): Remove the
11148 the assertion that the stmts have NULL use_ops.
11149 Discard the statements, asserting that they haven't
11150 yet been added to a BB.
11152 2018-03-08 Richard Biener <rguenther@suse.de>
11154 PR tree-optimization/84746
11155 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11156 (phi_translate): Pass in destination ANTIC_OUT set.
11157 (phi_translate_1): Likewise. For a simplified result lookup
11158 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11159 (phi_translate_set): Adjust.
11160 (do_pre_regular_insertion): Likewise.
11161 (do_pre_partial_partial_insertion): Likewise.
11163 2018-03-08 Martin Liska <mliska@suse.cz>
11165 PR gcov-profile/84735
11166 * doc/gcov.texi: Document usage of profile files.
11167 * gcov-io.h: Document changes in the format.
11169 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11173 * dwarf2out.c (struct dw_line_info_table): Update comments for
11175 (FORCE_RESET_NEXT_VIEW): New.
11176 (FORCE_RESETTING_VIEW_P): New.
11177 (RESETTING_VIEW_P): Check for -1 too.
11178 (ZERO_VIEW_P): Likewise.
11179 (new_line_info_table): Force-reset next view.
11180 (dwarf2out_begin_function): Likewise.
11181 (dwarf2out_source_line): Simplify zero_view_p initialization.
11182 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11183 view directly. Omit view when omitting .loc at line 0.
11185 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11187 PR tree-optimization/84740
11188 * tree-switch-conversion.c (process_switch): Call build_constructors
11189 only if info.phi_count is non-zero.
11191 PR tree-optimization/84739
11192 * tree-tailcall.c (find_tail_calls): Check call arguments against
11193 DECL_ARGUMENTS (current_function_decl) rather than
11194 DECL_ARGUMENTS (func) when checking for tail recursion.
11196 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11198 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11199 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11200 Volker Reichelt's entry and add entries for people that perform
11201 GCC fuzzy testing and report numerous bugs.
11203 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11206 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11207 readonly data in sdata, if that is disabled.
11208 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11209 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11210 -mreadonly-in-sdata option.
11212 2018-03-07 Martin Sebor <msebor@redhat.com>
11214 PR tree-optimization/84468
11215 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11216 basic block when looking for nul assignment.
11218 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11221 * except.h (output_function_exception_table): Adjust prototype.
11222 * except.c (output_function_exception_table): Remove FNNAME parameter
11223 and add SECTION parameter. Ouput one part of the table at a time.
11224 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11225 the first part of the exception table and emit unwind directives.
11226 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11227 (i386_pe_seh_cold_init): Likewise.
11228 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11229 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11230 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11231 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11232 * config/i386/winnt.c: Include except.h.
11233 (struct seh_frame_state): Add reg_offset, after_prologue and
11234 in_cold_section fields.
11235 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11236 (i386_pe_seh_cold_init): New function.
11237 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11238 to seh->in_cold_section.
11239 (seh_emit_push): Record the offset of the push.
11240 (seh_emit_save): Record the offet of the save.
11241 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11242 Test seh->after_prologue to disregard the epilogue.
11243 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11244 (i386_pe_end_cold_function): New function.
11246 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11249 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11250 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11253 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11255 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11256 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11258 PR middle-end/84723
11259 * multiple_target.c: Include tree-inline.h and intl.h.
11260 (expand_target_clones): Diagnose and fail if node->definition and
11261 !tree_versionable_function_p (node->decl).
11263 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11265 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11267 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11268 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11269 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11271 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11274 * combine.c (try_combine): Use reg_or_subregno instead of handling
11275 just paradoxical SUBREGs and REGs.
11277 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11279 * config/arc/arc.c (arc_finalize_pic): Remove function.
11280 (arc_must_save_register): We use single base PIC register, remove
11281 checks to save/restore the PIC register.
11282 (arc_expand_prologue): Likewise.
11283 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11285 (arc_verify_short): Likewise.
11286 (arc_attr_type): Likewise.
11287 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11288 (walk_stores): Likewise.
11289 (arc_address_cost): Make it static.
11290 (arc_verify_short): Likewise.
11291 (branch_dest): Likewise.
11292 (arc_attr_type): Likewise.
11293 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11294 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11295 (arc_final_prescan_insn): Remove inserting the nops due to
11296 hardware hazards. It is done in reorg step.
11297 (insn_length_variant_t): Remove.
11298 (insn_length_parameters_t): Likewise.
11299 (arc_insn_length_parameters): Likewise.
11300 (arc_get_insn_variants): Likewise.
11301 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11303 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11305 PR inline-asm/84683
11306 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11309 PR tree-optimization/84687
11310 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11312 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11314 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11316 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11317 Rename to ppc_speculation_barrier.
11318 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11319 __builtin_ppc_speculation_barrier.
11321 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11324 * combine.c (combine_simplify_rtx): Don't try to simplify if
11325 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11328 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11330 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11331 to 32 bytes when compiling for POWER9.
11333 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11336 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11337 regparm >= 3 with no arg reg available also for calls with
11338 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11341 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11343 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11345 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11348 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11350 2018-03-05 Richard Biener <rguenther@suse.de>
11352 PR tree-optimization/84486
11353 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11354 When inserting a __builtin_assume_aligned call set the LHS
11355 SSA name alignment info accordingly.
11357 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11359 PR tree-optimization/84114
11360 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11361 Avoid reassociation of FLOAT_MODE addition.
11363 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11365 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11366 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11367 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11368 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11369 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11370 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11371 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11373 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11374 __builtin_ia32_wbinvd): New builtins.
11375 (SPECIAL_ARGS2): New.
11376 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11377 (SPECIAL_ARGS2): New.
11378 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11379 (ix86_valid_target_attribute_inner_p): Ditto.
11380 (ix86_init_mmx_sse_builtins): Add special_args2.
11381 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11382 TARGET_WBNOINVD_P): New.
11383 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11384 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11385 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11386 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11387 * config/i386/pconfigintrin.h: New file.
11388 * config/i386/wbnoinvdintrin.h: Ditto.
11389 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11391 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11393 2018-03-05 Richard Biener <rguenther@suse.de>
11395 PR tree-optimization/84670
11396 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11398 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11399 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11400 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11401 doesn't grow if all successors (recursively) were visited at least
11404 2018-03-05 Richard Biener <rguenther@suse.de>
11406 PR tree-optimization/84650
11407 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11408 if executed in the loop pipeline.
11410 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11412 * doc/configfiles.texi (Configuration Files): Move info about
11413 conditionalizing $target-protos.h to...
11414 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11415 differs from $target-protos.h.
11417 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11418 Chung-Ju Wu <jasonwucj@gmail.com>
11420 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11421 * config/nds32/nds32-multiple.md (setmemsi): Define.
11422 * config/nds32/nds32-memory-manipulation.c
11423 (nds32_gen_dup_4_byte_to_word_value): New.
11424 (emit_setmem_word_loop): New.
11425 (emit_setmem_byte_loop): New.
11426 (nds32_expand_setmem_loop): New.
11427 (nds32_expand_setmem_loop_v3m): New.
11428 (nds32_expand_setmem_unroll): New.
11429 (nds32_expand_setmem): New.
11431 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11432 Chung-Ju Wu <jasonwucj@gmail.com>
11434 * config/nds32/nds32-memory-manipulation.c
11435 (nds32_emit_load_store): New.
11436 (nds32_emit_post_inc_load_store): New.
11437 (nds32_emit_mem_move): New.
11438 (nds32_emit_mem_move_block): New.
11439 (nds32_expand_movmemsi_loop_unknown_size): New.
11440 (nds32_expand_movmemsi_loop_known_size): New.
11441 (nds32_expand_movmemsi_loop): New.
11442 (nds32_expand_movmemsi_unroll): New.
11443 (nds32_expand_movmemqi): Rename ...
11444 (nds32_expand_movmemsi): ... to this.
11445 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11446 (movmemsi): ... to this.
11447 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11448 (nds32_expand_movmemsi): ... to this.
11450 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11451 Monk Chiang <sh.chiang04@gmail.com>
11452 Chung-Ju Wu <jasonwucj@gmail.com>
11454 * config/nds32/nds32-protos.h
11455 (nds32_expand_load_multiple): New arguments.
11456 (nds32_expand_store_multiple): Ditto.
11457 (nds32_valid_multiple_load_store): Rename ...
11458 (nds32_valid_multiple_load_store_p): ... to this.
11459 * config/nds32/nds32-memory-manipulation.c
11460 (nds32_expand_load_multiple): Refine implementation.
11461 (nds32_expand_store_multiple): Ditto.
11462 * config/nds32/nds32-multiple.md
11463 (load_multiple): Update nds32_expand_load_multiple interface.
11464 (store_multiple): Update nds32_expand_store_multiple interface.
11465 * config/nds32/nds32-predicates.c
11466 (nds32_valid_multiple_load_store): Rename ...
11467 (nds32_valid_multiple_load_store_p): ... to this and refine
11469 * config/nds32/predicates.md
11470 (nds32_load_multiple_and_update_address_operation): New predicate.
11471 (nds32_store_multiple_and_update_address_operation): New predicate.
11473 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11474 Chung-Ju Wu <jasonwucj@gmail.com>
11476 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11477 (combo): New attribute.
11478 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11480 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11482 * config/nds32/nds32.opt: Change -mcmodel= default value.
11484 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11485 Monk Chiang <sh.chiang04@gmail.com>
11486 Chung-Ju Wu <jasonwucj@gmail.com>
11488 * config/nds32/constants.md (unspec_element): New enum.
11489 * config/nds32/constraints.md (Umw): New constraint.
11490 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11491 * config/nds32/nds32-intrinsic.md: Likewise.
11492 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11493 (nds32_valid_smw_lwm_base_p): New.
11494 (nds32_output_smw_single_word): New.
11495 (nds32_output_lmw_single_word): New.
11496 (nds32_expand_unaligned_load): New.
11497 (nds32_expand_unaligned_store): New.
11498 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11499 (nds32_output_smw_single_word): Declare.
11500 (nds32_output_lmw_single_word): Declare.
11501 (nds32_expand_unaligned_load): Declare.
11502 (nds32_expand_unaligned_store): Declare.
11503 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11504 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11505 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11506 NDS32_BUILTIN_UASTORE_DW.
11507 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11510 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11511 Kito Cheng <kito.cheng@gmail.com>
11512 Chung-Ju Wu <jasonwucj@gmail.com>
11514 * config/nds32/nds32-intrinsic.c
11515 (nds32_expand_builtin_null_ftype_reg): Delete.
11516 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11517 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11518 (nds32_read_argument): New.
11519 (nds32_legitimize_target): Ditto.
11520 (nds32_legitimize_argument): Ditto.
11521 (nds32_check_constant_argument): Ditto.
11522 (nds32_expand_unop_builtin): Ditto.
11523 (nds32_expand_unopimm_builtin): Ditto.
11524 (nds32_expand_binop_builtin): Ditto.
11525 (nds32_builtin_decl_impl): Ditto.
11526 (builtin_description): Ditto.
11527 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11528 (nds32_init_builtins_impl): Ditto.
11529 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11530 (nds32_builtin_decl): New.
11531 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11532 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11534 2018-03-02 Jeff Law <law@redhat.com>
11536 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11537 (redundant_insn, fill_simple_delay_slots): Likewise.
11538 (fill_slots_from_thread): Likewise.
11539 * resource.c (mark_referenced_resources): Likewise.
11540 (mark_set_resources, find_dead_or_set_registers): Likewise.
11542 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11544 * substring-locations.h (format_warning_va): Formatting fix for
11545 ATTRIBUTE_GCC_DIAG.
11546 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11548 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11549 * substring-locations.c: Include intl.h.
11550 (format_warning_va): Turned into small wrapper around
11551 format_warning_n_va, renamed to ...
11552 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11553 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11555 (format_warning_at_substring_n): New function.
11556 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11557 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11558 format_warning_at_substring with just a shorter name instead of
11559 const function pointer.
11560 (fmtwarn_n): New function.
11561 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11562 appropriate, get rid of all the fmtstr temporaries, move conditionals
11563 with G_() wrapped string literals directly into fmtwarn arguments,
11564 cast dir.len to (int), formatting fixes.
11566 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11568 * doc/invoke.texi: Remove "Cilk Plus" references.
11570 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11571 Richard Biener <rguenther@suse.de>
11574 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11575 for error or warning attributes if CALL_FROM_THUNK_P is set.
11578 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11581 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11582 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11585 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11586 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11588 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11589 instead of -1U in last predictors element's probability member.
11591 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11594 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11595 arguments if they are comparable.
11597 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11599 PR tree-optimization/84634
11600 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11601 masks and masked_loop_p with a single loop_masks, making sure it's
11602 null for bb vectorization.
11604 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11606 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11607 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11608 loop->force_vectorize to check whether there is no alias.
11610 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11613 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11615 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11617 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11618 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11619 instead of a loop around prev_real_insn.
11620 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11623 PR inline-asm/84625
11624 * config/i386/i386.c (ix86_print_operand): Use conditional
11625 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11628 2018-03-02 Richard Biener <rguenther@suse.de>
11630 PR tree-optimization/84427
11631 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11632 (bitmap_set_subtract_values): Rewrite to handle multiple
11635 (prune_clobbered_mems): Likewise.
11636 (phi_translate): Take edge instead of pred/phiblock.
11637 (phi_translate_1): Likewise.
11638 (phi_translate_set): Likewise. Insert all translated
11639 exprs for a value into the set, keeping possibly multiple
11640 expressions per value.
11641 (compute_antic_aux): Adjust for phi_translate changes.
11642 When intersecting union the expressions and prune those
11643 not in the final value set, keeping possibly multiple
11644 expressions per value. Do not use value-insertion
11645 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11646 all expressions. Add verification that the value-sets
11647 only shrink during iteration.
11648 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11649 (do_pre_regular_insertion): Likewise.
11650 (do_pre_partial_partial_insertion): Likewise.
11652 2018-03-02 Richard Biener <rguenther@suse.de>
11655 * config/darwin.c (saved_debug_info_level): New static global.
11656 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11657 (darwin_asm_lto_end): Restore debug info generation settings.
11659 2018-03-01 Martin Liska <mliska@suse.cz>
11662 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11663 volatile arguments.
11665 2018-03-01 Richard Biener <rguenther@suse.de>
11668 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11669 type annotation with LTO when debug was disabled at compile-time.
11671 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11673 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11675 (mips_final_postscan_insn): Likewise.
11677 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11679 PR rtl-optimization/84528
11680 * alias.c (init_alias_target): Add commentary.
11681 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11682 a unique base value if the frame pointer is not eliminated
11683 to the stack pointer.
11685 2018-03-01 Tom de Vries <tom@codesourcery.com>
11687 PR rtl-optimization/83327
11688 * lra-int.h (hard_regs_spilled_into): Declare.
11689 * lra.c (hard_regs_spilled_into): Define.
11690 (init_reg_info): Init hard_regs_spilled_into.
11691 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11692 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11693 (process_bb_lives): Handle hard_regs_spilled_into.
11694 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11695 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11697 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11699 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11700 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11701 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11702 * config/rs6000/aix72.h: New file.
11704 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11706 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11707 instead of warning_at with conditional singular and plural messages
11711 * stor-layout.c (update_alignment_for_field): For
11712 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11713 && !DECL_PACKED (field), do the alignment update, just use
11714 only desired_align instead of MAX (type_align, desired_align)
11716 (place_field): Don't do known_align < desired_align handling
11717 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11718 is non-NULL, instead do it after rli->prev_field handling and
11719 only if not within a bitfield word. For DECL_PACKED (field)
11720 use type_align of BITS_PER_UNIT.
11722 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11724 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11725 superfluous parentheses and trailing spaces.
11727 2018-02-28 Richard Biener <rguenther@suse.de>
11729 PR tree-optimization/84584
11730 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11731 SCoPs with fake exit edge.
11733 2018-02-28 Martin Liska <mliska@suse.cz>
11736 * timevar.c (timer::print): Fix format to properly print 100%
11739 2018-02-28 Richard Biener <rguenther@suse.de>
11741 PR middle-end/84607
11742 * genmatch.c (capture_info::walk_match): Do not mark
11743 captured expressions without operands as expr_p given
11744 they act more like predicates and should be subject to
11745 "lost tail" side-effect preserving.
11747 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11749 PR rtl-optimization/81611
11750 * auto-inc-dec.c (attempt_change): Move dead note from
11751 mem_insn if it's the next use of regno
11752 (find_address): Take address use of reg holding
11753 non-incremented value. Add parm to limit search to the named
11755 (merge_in_block): Attempt to use a mem insn that is the next
11756 use of the original regno.
11758 2018-02-27 Martin Sebor <msebor@redhat.com>
11761 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11762 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11764 2018-02-27 Martin Sebor <msebor@redhat.com>
11766 PR translation/84207
11767 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11768 n argument to unsigned HOST_WIDE_INT.
11769 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11770 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11771 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11772 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11774 2018-02-27 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/84512
11777 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11778 Do not use the estimate returned from record_stmt_cost for
11779 the scalar iteration cost but sum properly using add_stmt_cost.
11781 2018-02-27 Richard Biener <rguenther@suse.de>
11783 PR tree-optimization/84466
11784 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11785 Adjust last change to less strictly validate use operands.
11787 2018-02-27 Martin Liska <mliska@suse.cz>
11789 PR gcov-profile/84548
11790 * gcov.c (process_file): Allow partial overlap and consider it
11791 also as group functions.
11792 (output_lines): Properly calculate range of lines for a group.
11794 2018-02-27 Martin Liska <mliska@suse.cz>
11796 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11797 'ggc' suffixes. Change first column width.
11798 (timer::print): Fix formatting of the column.
11800 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11802 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11803 preserve inline entry blocks for the sake of debug inline
11804 entry point markers alone.
11805 (remove_unused_locals): Suggest in comments a better place to
11806 force the preservation of inline entry blocks that are
11807 otherwise unused, but do not preserve them.
11809 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11811 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11813 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11816 * config/i386/constraints.md (Bs): Replace
11817 ix86_indirect_branch_register with
11818 TARGET_INDIRECT_BRANCH_REGISTER.
11820 * config/i386/i386.md (indirect_jump): Likewise.
11821 (tablejump): Likewise.
11822 (*sibcall_memory): Likewise.
11823 (*sibcall_value_memory): Likewise.
11824 Peepholes of indirect call and jump via memory: Likewise.
11825 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11826 (*sibcall_value_GOT_32): Likewise.
11827 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11828 (GOT_memory_operand): Likewise.
11829 (call_insn_operand): Likewise.
11830 (sibcall_insn_operand): Likewise.
11831 (GOT32_symbol_operand): Likewise.
11832 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11834 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11836 PR rtl-optimization/83496
11837 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11838 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11839 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11840 redundant insn, if any.
11841 (relax_delay_slots): Likewise.
11842 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11844 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11846 PR tree-optimization/83965
11847 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11848 that grouped statements are part of a reduction chain. Return
11849 true if the statement is not marked as a reduction itself but
11850 is part of a group.
11851 (vect_recog_dot_prod_pattern): Don't check whether the statement
11852 is part of a group here.
11853 (vect_recog_sad_pattern): Likewise.
11854 (vect_recog_widen_sum_pattern): Likewise.
11856 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11859 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11861 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11864 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11866 (ix86_output_indirect_function_return): New prototype.
11867 (ix86_split_simple_return_pop_internal): Likewise.
11868 * config/i386/i386.c (indirect_return_via_cx): New.
11869 (indirect_return_via_cx_bnd): Likewise.
11870 (indirect_thunk_name): Handle return va CX_REG.
11871 (output_indirect_thunk_function): Create alias for
11872 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11873 (ix86_output_indirect_jmp): Remove the bool argument.
11874 (ix86_output_indirect_function_return): New function.
11875 (ix86_split_simple_return_pop_internal): Likewise.
11876 * config/i386/i386.md (*indirect_jump): Don't pass false
11877 to ix86_output_indirect_jmp.
11878 (*tablejump_1): Likewise.
11879 (simple_return_pop_internal): Change it to define_insn_and_split.
11880 Call ix86_split_simple_return_pop_internal to split it for
11881 -mfunction-return=.
11882 (simple_return_indirect_internal): Call
11883 ix86_output_indirect_function_return instead of
11884 ix86_output_indirect_jmp.
11886 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11889 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11890 memset and value initialization afterwards.
11892 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11894 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11896 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11899 * common/config/aarch64/aarch64-common.c
11900 (aarch_option_optimization_table[]): Switch
11901 off fomit-frame-pointer
11903 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11904 Chung-Ju Wu <jasonwucj@gmail.com>
11906 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11908 (store_multiple): Ditto.
11910 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11912 * config.gcc: Add --with-cpu support for nds32 target.
11913 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11914 * config/nds32/nds32.opt: Add -mcpu= option.
11916 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11918 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11919 isel=yes): Warn for these deprecated options.
11921 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11923 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11924 ISA_2_5_MASKS_EMBEDDED.
11926 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11928 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11929 p->max as pointers rather than using iterative_hash_expr.
11931 2018-02-23 Carl Love <cel@us.ibm.com>
11933 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11934 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11936 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11937 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11938 P8V_BUILTIN_VEC_VUNSIGNED2.
11940 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11943 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11944 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11946 * lra-constraints.c (curr_insn_transform): Set up
11947 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11949 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11950 * lra-eliminations.c (spill_pseudos): Ditto.
11951 (process_insn_for_elimination): Ditto.
11952 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11953 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11954 LRA_NON_CLOBBERED_ALT.
11956 2018-02-22 Martin Sebor <msebor@redhat.com>
11958 PR tree-optimization/84480
11959 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11960 to maybe_diag_stxncpy_trunc. Call it.
11961 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11962 from gimple_fold_builtin_strcpy. Print inlining stack.
11963 (handle_builtin_stxncpy): Print inlining stack.
11964 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11966 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11969 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11970 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11971 and -fcheck-pointer-bounds are used together.
11972 (indirect_thunk_prefix): New enum.
11973 (indirect_thunk_need_prefix): New function.
11974 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11975 "_nt" instead of "_bnd" for NOTRACK prefix.
11976 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11977 (output_indirect_thunk_function): Likewise.
11979 (ix86_code_end): Update output_indirect_thunk_function calls.
11980 (ix86_output_indirect_branch_via_reg): Replace
11981 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11982 (ix86_output_indirect_branch_via_push): Likewise.
11983 (ix86_output_function_return): Likewise.
11984 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11985 incompatible with -fcf-protection=branch and
11986 -fcheck-pointer-bounds.
11988 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11991 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11992 Change gcc_assert call to output_operand_lossage.
11994 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11996 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11998 2018-02-22 DJ Delorie <dj@redhat.com>
11999 Sebastian Perta <sebastian.perta@renesas.com>
12000 Oleg Endo <olegendo@gcc.gnu.org>
12002 * config/rx/rx.c (rx_rtx_costs): New function.
12003 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12005 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12007 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12009 2018-02-22 Martin Liska <mliska@suse.cz>
12012 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12013 Add "native" as a possible value.
12015 2018-02-22 Martin Liska <mliska@suse.cz>
12018 * config/i386/i386.c (ix86_option_override_internal):
12019 Add "native" as a possible value for -march and -mtune.
12021 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12024 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12025 to all type variants.
12027 PR tree-optimization/84503
12028 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12029 width as info->bitpos + info->bitsize - start.
12030 (merged_store_group::merge_overlapping): Simplify width computation.
12031 (check_no_overlap): New function.
12032 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12033 start + width and last_order of the group, fail if check_no_overlap
12035 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12036 to group if check_no_overlap fails.
12038 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12040 * config/rs6000/altivec.md: Delete contraint arguments to
12041 define_expand, define_split, and define_peephole2, and in
12042 define_insn_and_split if always unused.
12043 * config/rs6000/darwin.md: Ditto.
12044 * config/rs6000/dfp.md: Ditto.
12045 * config/rs6000/rs6000.md: Ditto.
12046 * config/rs6000/sync.md: Ditto.
12047 * config/rs6000/vector.md: Ditto.
12048 * config/rs6000/vsx.md: Ditto.
12050 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12052 * config/rs6000/altivec.md: Write output control strings as braced
12053 blocks instead of double-quoted strings.
12054 * config/rs6000/darwin.md: Ditto.
12055 * config/rs6000/rs6000.md: Ditto.
12056 * config/rs6000/vector.md: Ditto.
12057 * config/rs6000/vsx.md: Ditto.
12059 2018-02-21 Jason Merrill <jason@redhat.com>
12061 PR c++/84314 - ICE with templates and fastcall attribute.
12062 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12064 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12066 * ipa-cp.c (determine_versionability): Fix comment typos.
12068 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12071 * ipa-cp.c (determine_versionability): Do not version functions caling
12074 2018-02-21 Martin Liska <mliska@suse.cz>
12077 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12078 Add "native" as a possible value.
12079 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12080 the macro when native cpu detection is available.
12082 2018-02-21 Martin Liska <mliska@suse.cz>
12085 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12086 Add "native" as a possible value.
12087 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12088 when native cpu detection is available.
12090 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12091 Martin Sebor <msebor@redhat.com>
12093 PR tree-optimization/84478
12094 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12096 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12097 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12098 support which is conservatively correct, for 2 only stay conservative
12099 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12100 argument to the 2 argument get_range_strlen, adjust 6 arg
12101 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12103 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12104 (gimple_fold_builtin_strlen): Pass true as last argument to
12107 2018-02-20 Martin Sebor <msebor@redhat.com>
12109 PR middle-end/84095
12110 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12111 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12112 (builtin_memref::builtin_memref): Factor out parts into
12113 set_base_and_offset and call it.
12115 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12117 PR middle-end/84406
12118 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12119 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12120 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12121 search at the associated MODE_INT.
12123 2018-02-20 Jeff Law <law@redhat.com>
12125 PR middle-end/82123
12126 PR tree-optimization/81592
12127 PR middle-end/79257
12128 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12129 for range data rather than using global data.
12130 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12131 range data rather than using global data.
12132 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12133 pass it to children as needed.
12134 (struct directive::fmtresult): Similarly.
12135 (struct directive::set_width): Similarly.
12136 (struct directive::set_precision): Similarly.
12137 (format_integer, format_directive, parse_directive): Similarly.
12138 (format_none): Accept unnamed vr_values parameter.
12139 (format_percent, format_floating, format_character): Similarly.
12140 (format_string, format_plain): Similarly.
12141 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12142 the EVRP range analyzer for range data rather than using global data.
12143 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12144 gimple-ssa-evrp-analyze.h
12145 (class sprintf_dom_walker): Add after_dom_children member function.
12146 Add evrp_range_analyzer member.
12147 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12148 range analyzer as needed.
12149 (sprintf_dom_walker::after_dom_children): New member function.
12150 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12152 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12153 (evrp_range_analyzer::pop_to_marker): Likewise.
12155 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12157 PR tree-optimization/84419
12158 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12159 with the required type if its current type is compatible but
12162 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12164 PR middle-end/82004
12165 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12166 after vectorization.
12168 2018-02-20 Martin Liska <mliska@suse.cz>
12171 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12172 possible values if we don't have a hint.
12174 2018-02-20 Martin Liska <mliska@suse.cz>
12178 * final.c (shorten_branches): Build align_tab array with one
12180 * opts.c (finish_options): Add alignment option limit check.
12181 (MAX_CODE_ALIGN): Likewise.
12182 (MAX_CODE_ALIGN_VALUE): Likewise.
12183 * doc/invoke.texi: Document maximum allowed option value for
12184 all -falign-* options.
12186 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12189 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12190 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12191 * var-tracking.c (emit_note_insn_var_location): Remove all references
12192 to NOTE_INSN_CALL_ARG_LOCATION.
12193 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12194 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12195 Use copy_rtx_if_shared.
12196 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12197 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12198 (dwarf2out_var_location): Remove handling of
12199 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12201 * final.c (final_scan_insn): Remove all references to
12202 NOTE_INSN_CALL_ARG_LOCATION.
12203 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12204 before dumping final insns.
12205 * except.c (emit_note_eh_region_end): Remove all references to
12206 NOTE_INSN_CALL_ARG_LOCATION.
12207 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12208 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12209 * config/arc/arc.c (hwloop_optimize): Likewise.
12210 * config/arm/arm.c (create_fix_barrier): Likewise.
12211 * config/s390/s390.c (s390_chunkify_start): Likewise.
12212 * config/sh/sh.c (find_barrier): Likewise.
12213 * config/i386/i386.c (rest_of_insert_endbranch,
12214 ix86_seh_fixup_eh_fallthru): Likewise.
12215 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12216 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12217 * config/frv/frv.c (frv_function_prologue): Likewise.
12218 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12220 (note_outside_basic_block_p): Remove all references to
12221 NOTE_INSN_CALL_ARG_LOCATION.
12222 * gengtype.c (adjust_field_rtx_def): Likewise.
12223 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12225 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12226 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12229 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12232 PR tree-optimization/84452
12233 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12234 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12237 2018-02-19 Martin Liska <mliska@suse.cz>
12240 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12242 2018-02-19 Martin Liska <mliska@suse.cz>
12243 Richard Sandiford <richard.sandiford@linaro.org>
12245 PR tree-optimization/82491
12246 * gimple-fold.c (get_base_constructor): Make earlier bail out
12249 2018-02-19 Carl Love <cel@us.ibm.com>
12251 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12252 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12254 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12255 P8V_BUILTIN_VEC_NEG.
12257 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12259 * config/rl78/rl78.md (movdf): New define expand.
12261 2018-02-19 Martin Liska <mliska@suse.cz>
12264 * doc/invoke.texi: Fix typo.
12265 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12267 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12269 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12270 handle rs6000_single_float and rs6000_double_float specially for
12273 2018-02-16 Jeff Law <law@redhat.com>
12275 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12276 (pop_regs): Use it.
12278 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12281 * ipa-inline.c (inline_small_functions): Fix a typo.
12283 2018-02-16 Nathan Sidwell <nathan@acm.org>
12285 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12287 2018-02-16 Carl Love <cel@us.ibm.com>
12289 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12290 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12291 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12292 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12293 expansion to P8V_BUILTIN_VEC_FLOAT2.
12295 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12297 PR rtl-optimization/70023
12298 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12299 src_regno into account.
12301 2018-02-16 Carl Love <cel@us.ibm.com>
12303 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12304 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12305 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12306 * config/rs6000/rs6000.c: Remove case statements for
12307 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12308 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12309 and P9V_BUILTIN_VEC_VINSERT4B.
12310 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12311 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12312 * config/rs6000/vsx.md:
12313 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12316 2018-02-16 Carl Love <cel@us.ibm.com>
12318 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12320 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12322 * config/rs6000/rs6000-c.c: Add the definitions for
12323 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12324 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12325 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12326 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12327 definition for insert4b and define insn *insert3b_internal.
12328 * doc/extend.texi: Add documentation for vec_extract4b.
12330 2018-02-16 Nathan Sidwell <nathan@acm.org>
12332 * doc/extend.texi (Backwards Compatibility): Mention friend
12333 injection. Note for-scope is deprecated.
12334 * doc/invoke.texi (-ffriend-injection): Deprecate.
12336 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12338 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12339 that moved to I2, also allow destinations that are a paradoxical
12340 subreg (instead of a normal reg).
12342 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12345 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12348 2018-02-16 Richard Biener <rguenther@suse.de>
12350 PR tree-optimization/84037
12351 PR tree-optimization/84016
12353 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12354 Adjust vec_construct for the fact we need additional higher latency
12355 128bit inserts for AVX256 and AVX512 vector builds.
12356 (ix86_add_stmt_cost): Scale vector construction cost for
12359 2018-02-16 Richard Biener <rguenther@suse.de>
12361 PR tree-optimization/84417
12362 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12363 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12364 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12366 2018-02-16 Martin Liska <mliska@suse.cz>
12369 * internal-fn.def (ASAN_CHECK): Set proper flags.
12370 (ASAN_MARK): Likewise.
12372 2018-02-16 Julia Koval <julia.koval@intel.com>
12374 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12375 from PTA_CANNONLAKE.
12377 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12380 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12381 Use ++iter rather than iter++ for std::list iterators.
12382 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12383 defer deleting them until all nodes in the forest are processed. Do
12384 free even leaf nodes. Change to_process into auto_vec.
12387 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12388 * vec.h (vec_default_construct): Use memset instead of placement new
12389 if BROKEN_VALUE_INITIALIZATION is defined.
12390 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12391 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12394 PR rtl-optimization/83723
12395 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12396 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12397 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12399 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12401 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12403 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12405 PR rtl-optimization/81443
12406 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12407 from inner REGs to paradoxical SUBREGs.
12409 2018-02-16 Richard Biener <rguenther@suse.de>
12411 PR tree-optimization/84399
12412 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12413 For operands we can analyze at their definition make sure we can
12414 analyze them at each use as well.
12416 2018-02-16 Richard Biener <rguenther@suse.de>
12418 PR tree-optimization/84190
12419 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12420 volatile accesses if the decl isn't volatile.
12422 2018-02-15 Jason Merrill <jason@redhat.com>
12424 PR c++/84314 - ICE with templates and fastcall attribute.
12425 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12426 TYPE_CANONICAL on an existing type.
12428 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12430 PR tree-optimization/84383
12431 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12432 dstoff nor call operand_equal_p if dstbase is NULL.
12434 PR tree-optimization/84334
12435 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12436 also a CONSTANT_CLASS_P, punt.
12438 2018-02-14 Jim Wilson <jimw@sifive.com>
12440 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12441 first SMALL_OPERAND check. New local min_second_step. Move assert
12442 to where locals are set. Add TARGET_RVC support.
12443 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12445 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12447 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12449 2018-02-14 Martin Sebor <msebor@redhat.com>
12451 PR tree-optimization/83698
12452 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12453 arrays constrain the offset range to their bounds.
12454 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12455 (builtin_access::overlap): Avoid setting the size of overlap if it's
12457 (maybe_diag_overlap): Also consider arrays when deciding what values
12458 of offsets to include in diagnostics.
12460 2018-02-14 Martin Sebor <msebor@redhat.com>
12463 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12464 that correspond to the kind of a declaration.
12466 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12469 * config/pa/pa.md: Load address of PIC label using the linkage table
12470 if the label is nonlocal.
12472 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12474 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12475 warning message if user requests -maltivec=be.
12476 * doc/invoke.texi: Document deprecation of -maltivec=be.
12478 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12481 * config/rs6000/rs6000-c.c: Update definitions for
12482 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12483 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12485 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12488 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12489 add _get_ssp intrinsics. Remove argument from
12490 __builtin_ia32_rdssp[d|q].
12491 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12492 * config/i386/i386-builtin.def: Remove argument from
12493 __builtin_ia32_rdssp[d|q].
12494 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12495 ix86_expand_special_args_builtin for _rdssp[d|q].
12496 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12497 Clear register before usage.
12498 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12499 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12501 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12503 PR tree-optimization/84357
12504 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12505 operand 1 of an ARRAY_REF too.
12507 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12510 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12511 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12513 (set_of_reg): New struct.
12514 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12515 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12516 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12518 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12519 Split into bitclr, bitset, bitinvert patterns if appropriate.
12520 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12521 use rx_fuse_in_memory_bitop.
12522 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12523 to named insn, correct maximum insn length.
12525 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12528 * machmode.def: Define a complex mode for PARTIAL_INT.
12529 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12531 * doc/rtl.texi: Document CSPImode.
12532 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12534 (msp430_hard_regno_nregs_with_padding): Likewise.
12536 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12539 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12541 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12543 PR rtl-optimization/84169
12544 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12545 we generated a parallel as new i3 and we split that to new i2 and i3
12546 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12547 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12548 those to i2, not i1. Partially rewrite this scan code.
12550 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12553 * stor-layout.c (place_field): For variable length fields, adjust
12554 offset_align afterwards not just based on the field's alignment,
12555 but also on the size.
12557 PR middle-end/84309
12558 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12559 of exps and logs in the use_exp2 case.
12561 2018-02-13 Jeff Law <law@redhat.com>
12563 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12564 entry for "vector".
12566 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12569 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12573 * common.opt (gas-loc-support, gas-locview-support): New.
12574 (ginline-points, ginternal-reset-location-views): New.
12575 * doc/invoke.texi: Document them. Use @itemx where intended.
12576 (gvariable-location-views): Adjust.
12577 * target.def (reset_location_view): New.
12578 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12579 (TARGET_RESET_LOCATION_VIEW): New.
12580 * doc/tm.texi: Rebuilt.
12581 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12582 (dwarf2out_default_as_locview_support): New.
12583 (output_asm_line_debug_info): Use option variables.
12584 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12585 (output_loc_list): Likewise.
12586 (add_high_low_attributes): Check option variables.
12587 Don't output entry view attribute in strict mode.
12588 (gen_inlined_subroutine_die): Check option variables.
12589 (dwarf2out_inline_entry): Likewise.
12590 (init_sections_and_labels): Likewise.
12591 (dwarf2out_early_finish): Likewise.
12592 (maybe_reset_location_view): New, from...
12593 (dwarf2out_var_location): ... here. Call it.
12594 * debug.h (dwarf2out_default_as_loc_support): Declare.
12595 (dwarf2out_default_as_locview_support): Declare.
12596 * hooks.c (hook_int_rtx_insn_0): New.
12597 * hooks.h (hook_int_rtx_insn_0): Declare.
12598 * toplev.c (process_options): Take -gas-loc-support and
12599 -gas-locview-support from dwarf2out. Enable
12600 -gvariable-location-views by default only with locview
12601 assembler support. Enable -ginternal-reset-location-views by
12602 default only if the target defines the corresponding hook.
12603 Enable -ginline-points by default if location views are
12604 enabled; force it disabled if statement frontiers are
12606 * tree-inline.c (expand_call_inline): Check option variables.
12607 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12609 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12611 PR tree-optimization/84321
12612 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12613 handling. Also check whether the anti-range contains any values
12614 that satisfy the mask; switch to a VR_RANGE if not.
12616 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12619 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12621 2018-02-13 Martin Jambor <mjambor@suse.cz>
12624 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12625 of call statements, also set location of a load to a temporary.
12627 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12629 * config/rl78/rl78.c (add_vector_labels): New function.
12630 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12631 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12632 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12633 which checks that no arguments are passed.
12634 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12635 * doc/extend.texi: Documentation for the new attribute.
12637 2018-02-13 Andreas Schwab <schwab@suse.de>
12639 * config/riscv/linux.h (CPP_SPEC): Define.
12641 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12644 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12645 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12646 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12647 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12648 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12649 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12650 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12651 temporarily for AES and PCLMUL builtins.
12653 PR tree-optimization/84339
12654 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12655 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12658 PR middle-end/84309
12659 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12660 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12661 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12663 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12665 * omp-simd-clone.h: New file.
12666 * omp-simd-clone.c: Include omp-simd-clone.h.
12667 (expand_simd_clones): No longer static.
12668 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12669 cgraph.h and omp-simd-clone.h.
12670 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12671 (vect_recog_widen_shift_pattern): Formatting fix.
12672 (vect_pattern_recog_1): Don't check optab for calls.
12675 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12676 operands[2] into a REG before using gen_lowpart on it.
12678 2018-02-12 Jeff Law <law@redhat.com>
12681 * config/sh/sh.c (find_barrier): Consider a sibling call
12684 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12685 successfully back substituting a reg.
12687 2018-02-12 Richard Biener <rguenther@suse.de>
12689 PR tree-optimization/84037
12690 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12691 parameter, move visited init to caller.
12692 (vect_slp_analyze_operations): Separate cost from validity
12693 check, initialize visited once for all instances.
12694 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12696 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12698 (vect_model_promotion_demotion_cost): Likewise.
12699 (vectorizable_bswap): Guard cost modeling with !slp_node
12700 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12702 (vectorizable_call): Likewise.
12703 (vectorizable_conversion): Likewise.
12704 (vectorizable_assignment): Likewise.
12705 (vectorizable_shift): Likewise.
12706 (vectorizable_operation): Likewise.
12707 (vectorizable_store): Likewise.
12708 (vectorizable_load): Likewise.
12709 (vectorizable_condition): Likewise.
12710 (vectorizable_comparison): Likewise.
12712 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12715 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12716 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12717 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12719 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12721 PR middle-end/83665
12722 * params.def (inline-min-speedup): Increase from 8 to 15.
12723 (max-inline-insns-auto): Decrease from 40 to 30.
12724 * ipa-split.c (consider_split): Add some buffer for function to
12725 be considered inlining candidate.
12726 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12729 2018-02-12 Richard Biener <rguenther@suse.de>
12731 PR tree-optimization/84037
12732 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12733 matched stmts if we cannot swap the non-matched ones.
12735 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12737 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12738 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12739 _mm_maskz_scalef_round_ss): New intrinsics.
12740 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12741 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12742 __builtin_ia32_scalefss_round): Remove.
12743 (__builtin_ia32_scalefsd_mask_round,
12744 __builtin_ia32_scalefss_mask_round): New intrinsics.
12745 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12746 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12747 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12748 "<round_constraint>")): Changed to ...
12749 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12750 "<round_scalar_constraint>")): ... this.
12751 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12752 %0, %1, %2<round_op3>}"): Changed to ...
12753 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12754 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12755 %2<round_scalar_mask_op3>}"): ... this.
12756 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12758 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12760 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12761 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12762 (_mm_maskz_sqrt_round_ss): New intrinsics.
12763 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12764 (__builtin_ia32_sqrtsd_mask_round)
12765 (__builtin_ia32_sqrtss_mask_round): New builtins.
12766 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12767 (__builtin_ia32_sqrtss_round): Remove.
12768 (__builtin_ia32_sqrtsd_mask_round)
12769 (__builtin_ia32_sqrtss_mask_round): New builtins.
12770 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12771 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12772 ((match_operand:VF_128 1 "vector_operand"
12773 "xBm,<round_constraint>")): Changed to ...
12774 ((match_operand:VF_128 1 "vector_operand"
12775 "xBm,<round_scalar_constraint>")): ... this.
12776 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12777 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12778 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12779 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12780 %<iptr>1<round_scalar_mask_op3>}): ... this.
12781 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12782 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12784 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12787 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12788 Cast vec_cmpeq result to correct type.
12789 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12790 Cast vec_cmpgt result to correct type.
12792 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12794 * final.c (final_scan_insn_1): Renamed from...
12795 (final_scan_insn): ... this. New wrapper, to recover
12796 seen from the outermost call in recursive ones.
12797 * config/sparc/sparc.c (output_return): Drop seen from call.
12798 (output_sibcall): Likewise.
12799 * config/visium/visium.c (output_branch): Likewise.
12801 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12803 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12806 2018-02-10 Alan Modra <amodra@gmail.com>
12809 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12810 Specify LR as an input.
12812 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12815 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12816 remove_member_access_dummy_vars): New functions.
12817 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12818 lower_omp_1, execute_lower_omp): Use them.
12820 PR rtl-optimization/84308
12821 * shrink-wrap.c (spread_components): Release todo vector.
12823 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12825 PR rtl-optimization/57193
12826 * ira-color.c (struct allocno_color_data): Add member
12827 conflict_allocno_hard_prefs.
12828 (update_conflict_allocno_hard_prefs): New.
12829 (bucket_allocno_compare_func): Add a preference based on
12830 conflict_allocno_hard_prefs.
12831 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12832 (color_allocnos): Remove a dead code. Initiate
12833 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12835 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12838 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12839 constraint from =wa to wa. Avoid a subreg on the output operand,
12840 instead use a pseudo and subreg it in a move.
12841 (p9_xxbrd_<mode>): Changed to ...
12842 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12843 (p9_xxbrd_v2df): New expander.
12844 (p9_xxbrw_<mode>): Changed to ...
12845 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12846 (p9_xxbrw_v4sf): New expander.
12848 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12850 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12852 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12855 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12856 multiply in 32-bit mode.
12857 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12858 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12861 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12863 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12864 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12865 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12866 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12868 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12871 * dwarf2out.c (is_trivial_indirect_ref): New function.
12872 (dwarf2out_late_global_decl): Do not generate a location
12873 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12874 and that are not defined in the current unit.
12876 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12878 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12879 instead of a libcall for UNORDERED.
12881 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12884 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12885 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12887 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12890 * config/s390/s390.c (s390_set_current_function): Invoke
12891 s390_indirect_branch_settings also if fndecl didn't change.
12893 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12895 * config/rs6000/rs6000.md (blockage): Set length to zero.
12897 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12899 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12901 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12904 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12905 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12909 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12910 PARALLEL incoming that failed vt_get_decl_and_offset check.
12912 PR middle-end/84237
12913 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12914 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12916 (get_variable_section): For decls in named .bss* sections pass true as
12917 second argument to bss_initializer_p.
12919 2018-02-09 Marek Polacek <polacek@redhat.com>
12920 Jakub Jelinek <jakub@redhat.com>
12923 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12924 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12925 Sync some changes from cxx_fold_indirect_ref.
12927 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12929 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12931 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12932 (BLOCK_INLINE_ENTRY_LABEL): New.
12933 (dwarf2out_var_location): Disregard inline entry markers.
12934 (inline_entry_data): New struct.
12935 (inline_entry_data_hasher): New hashtable type.
12936 (inline_entry_data_hasher::hash): New.
12937 (inline_entry_data_hasher::equal): New.
12938 (inline_entry_data_table): New variable.
12939 (add_high_low_attributes): Add DW_AT_entry_pc and
12940 DW_AT_GNU_entry_view attributes if a pending entry is found
12941 in inline_entry_data_table. Add old entry_pc attribute only
12942 if debug nonbinding markers are disabled.
12943 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12944 markers are enabled.
12945 (block_within_block_p, dwarf2out_inline_entry): New.
12946 (dwarf2out_finish): Check that no entries remained in
12947 inline_entry_data_table.
12948 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12949 (final_scan_insn, notice_source_line): Likewise.
12950 (rest_of_clean_state): Skip inline entry markers.
12951 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12953 * gimple.c (gimple_build_debug_inline_entry): New.
12954 * gimple.h (enum gimple_debug_subcode): Add
12955 GIMPLE_DEBUG_INLINE_ENTRY.
12956 (gimple_build_debug_inline_entry): Declare.
12957 (gimple_debug_inline_entry_p): New.
12958 (gimple_debug_nonbind_marker_p): Adjust.
12959 * insn-notes.def (INLINE_ENTRY): New.
12960 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12961 inline entry marker notes.
12962 (print_insn): Likewise.
12963 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12964 (INSN_DEBUG_MARKER_KIND): Likewise.
12965 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12966 * tree-inline.c (expand_call_inline): Build and insert
12967 debug_inline_entry stmt.
12968 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12969 inline entry blocks early, if nonbind markers are enabled.
12970 (dump_scope_block): Dump fragment info.
12971 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12972 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12973 (gimple_build_debug_inline_entry): New.
12974 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12975 Enable/disable inline entry points too.
12976 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12977 (DEBUG_INSN): Describe inline entry markers.
12979 * common.opt (gvariable-location-views): New.
12980 (gvariable-location-views=incompat5): New.
12981 * config.in: Rebuilt.
12982 * configure: Rebuilt.
12983 * configure.ac: Test assembler for view support.
12984 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12985 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12986 * dwarf2out.c (var_loc_view): New typedef.
12987 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12988 (dwarf2out_locviews_in_attribute): New.
12989 (dwarf2out_locviews_in_loclist): New.
12990 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12991 (enum dw_line_info_opcode): Add LI_adv_address.
12992 (struct dw_line_info_table): Add view.
12993 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12994 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12995 (zero_view_p): New variable.
12996 (ZERO_VIEW_P): New macro.
12997 (output_asm_line_debug_info): New.
12998 (struct var_loc_node): Add view.
12999 (add_AT_view_list, AT_loc_list): New.
13000 (add_var_loc_to_decl): Add view param. Test it against last.
13001 (new_loc_list): Add view params. Record them.
13002 (AT_loc_list_ptr): Handle loc and view lists.
13003 (view_list_to_loc_list_val_node): New.
13004 (print_dw_val): Handle dw_val_class_view_list.
13005 (size_of_die): Likewise.
13006 (value_format): Likewise.
13007 (loc_list_has_views): New.
13008 (gen_llsym): Set vl_symbol too.
13009 (maybe_gen_llsym, skip_loc_list_entry): New.
13010 (dwarf2out_maybe_output_loclist_view_pair): New.
13011 (output_loc_list): Output view list or entries too.
13012 (output_view_list_offset): New.
13013 (output_die): Handle dw_val_class_view_list.
13014 (output_dwarf_version): New.
13015 (output_compilation_unit_header): Use it.
13016 (output_skeleton_debug_sections): Likewise.
13017 (output_rnglists, output_line_info): Likewise.
13018 (output_pubnames, output_aranges): Update version comments.
13019 (output_one_line_info_table): Output view numbers in asm comments.
13020 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13021 Call maybe_gen_llsym.
13022 (loc_list_from_tree_1): Adjust.
13023 (add_AT_location_description): Create view list attribute if
13024 needed, check it's absent otherwise.
13025 (convert_cfa_to_fb_loc_list): Adjust.
13026 (maybe_emit_file): Call output_asm_line_debug_info for test.
13027 (dwarf2out_var_location): Reset views as needed. Precompute
13028 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13029 attribute. Set view.
13030 (new_line_info_table): Reset next view.
13031 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13032 (dwarf2out_source_line): Likewise. Output view resets and labels to
13033 the assembler, or select appropriate line info opcodes.
13034 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13035 (optimize_string_length): Catch it. Adjust.
13036 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13037 dw_val_class_view_list, and remove it if no longer needed.
13038 (hash_loc_list): Hash view numbers.
13039 (loc_list_hasher::equal): Compare them.
13040 (optimize_location_lists): Check whether a view list symbol is
13041 needed, and whether the locview attribute is present, and
13042 whether they match. Remove the locview attribute if no longer
13044 (index_location_lists): Call skip_loc_list_entry for test.
13045 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13046 Use output_dwarf_version.
13047 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13048 (struct dw_val_node): Add val_view_list.
13049 * final.c (SEEN_NEXT_VIEW): New.
13050 (set_next_view_needed): New.
13051 (clear_next_view_needed): New.
13052 (maybe_output_next_view): New.
13053 (final_start_function): Rename to...
13054 (final_start_function_1): ... this. Take pointer to FIRST,
13055 add SEEN parameter. Emit param bindings in the initial view.
13056 (final_start_function): Reintroduce SEEN-less interface.
13057 (final): Rename to...
13058 (final_1): ... this. Take SEEN parameter. Output final pending
13059 next view at the end.
13060 (final): Reintroduce seen-less interface.
13061 (final_scan_insn): Output pending next view before switching
13062 sections or ending a block. Mark the next view as needed when
13063 outputting variable locations. Notify debug backend of section
13064 changes, and of location view changes.
13065 (rest_of_handle_final): Adjust.
13066 * toplev.c (process_options): Autodetect value for debug variable
13067 location views option. Warn on incompat5 without -gdwarf-5.
13068 * doc/invoke.texi (gvariable-location-views): New.
13069 (gvariable-location-views=incompat5): New.
13070 (gno-variable-location-views): New.
13072 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13074 PR tree-optimization/84136
13075 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13076 that the result of find_edge is non-NULL.
13078 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13081 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13082 storing integer register in SImode. Fix cost of 256 and 512
13083 byte aligned SSE register store.
13085 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13087 * config/i386/i386.c (ix86_multiplication_cost): Fix
13088 multiplication cost for TARGET_AVX512DQ.
13090 2018-02-08 Marek Polacek <polacek@redhat.com>
13092 PR tree-optimization/84238
13093 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13096 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13098 PR tree-optimization/84265
13099 * tree-vect-stmts.c (vectorizable_store): Don't treat
13100 VMAT_CONTIGUOUS accesses as grouped.
13101 (vectorizable_load): Likewise.
13103 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13105 PR tree-optimization/81635
13106 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13107 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13108 (test_round_for_mask): New functions.
13109 (wide_int_cc_tests): Call test_round_for_mask.
13110 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13111 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13112 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13113 range returned by get_range_info.
13115 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13118 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13119 * symtab.c: Include builtins.h
13120 (symtab_node::output_to_lto_symbol_table_p): Move here
13121 from lto-streamer-out.c:output_symbol_p.
13122 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13123 (output_symbol_p): Move all logic to symtab.c
13124 (produce_symtab): Update.
13126 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13128 * config/s390/s390-opts.h (enum indirect_branch): Define.
13129 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13130 (s390_indirect_branch_via_thunk)
13131 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13132 (enum s390_indirect_branch_type): Define.
13133 * config/s390/s390.c (struct s390_frame_layout, struct
13134 machine_function): Remove.
13135 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13136 (indirect_branch_table_label_no, indirect_branch_table_name):
13138 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13139 (enum s390_indirect_branch_option): Define.
13140 (s390_return_addr_from_memory): New function.
13141 (s390_handle_string_attribute): New function.
13142 (s390_attribute_table): Add new attribute handler.
13143 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13144 (s390_indirect_branch_via_thunk): New function.
13145 (s390_indirect_branch_via_inline_thunk): New function.
13146 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13147 sibling call optimization for non z10 compiles.
13148 (s390_emit_call): Force indirect branch target to be a single
13149 register. Add r1 clobber for non-z10 compiles.
13150 (s390_emit_epilogue): Emit return jump via return_use expander.
13151 (s390_reorg): Handle JUMP_INSNs as execute targets.
13152 (s390_option_override_internal): Perform validity checks for the
13153 new command line options.
13154 (s390_indirect_branch_attrvalue): New function.
13155 (s390_indirect_branch_settings): New function.
13156 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13157 (s390_output_indirect_thunk_function): New function.
13158 (s390_code_end): Implement target hook.
13159 (s390_case_values_threshold): Implement target hook.
13160 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13162 * config/s390/s390.h (struct s390_frame_layout)
13163 (struct machine_function): Move here from s390.c.
13164 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13165 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13166 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13167 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13168 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13169 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13170 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13171 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13172 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13173 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13174 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13175 (mnemonic attribute): Add values which aren't recognized
13177 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13178 pattern for branch conversion. Fix mnemonic attribute.
13179 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13180 indirect branch via thunk if requested.
13181 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13182 ("*indirect_jump"): Disable for branch conversion using out of
13184 ("indirect_jump_via_thunk<mode>_z10")
13185 ("indirect_jump_via_thunk<mode>")
13186 ("indirect_jump_via_inlinethunk<mode>_z10")
13187 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13188 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13189 ("casesi_jump_via_inlinethunk<mode>_z10")
13190 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13191 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13192 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13193 ("*indirect2_jump"): Disable for branch conversion.
13194 ("casesi_jump"): Turn into expander and expand patterns for branch
13196 ("return_use"): New expander.
13197 ("*return"): Emit return via thunk and rename it to ...
13198 ("*return<mode>"): ... this one.
13199 * config/s390/s390.opt: Add new options and and enum for the
13202 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13204 * lra-constraints.c (match_reload): Unconditionally use
13205 gen_lowpart_SUBREG, rather than selecting between that
13206 and equivalent gen_rtx_SUBREG code.
13208 2018-02-08 Richard Biener <rguenther@suse.de>
13210 PR tree-optimization/84233
13211 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13212 changed flag instead of boguously re-using phi_inserted.
13214 2018-02-08 Martin Jambor <mjambor@suse.cz>
13216 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13217 static local variables.
13219 2018-02-08 Richard Biener <rguenther@suse.de>
13221 PR tree-optimization/84278
13222 * tree-vect-stmts.c (vectorizable_store): When looking for
13223 smaller vector types to perform grouped strided loads/stores
13224 make sure the mode is supported by the target.
13225 (vectorizable_load): Likewise.
13227 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13229 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13230 Increase LDP/STP opportunities by adding adjacent callee-saves.
13232 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13234 PR rtl-optimization/84068
13235 PR rtl-optimization/83459
13236 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13238 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13240 PR tree-optimization/84224
13241 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13242 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13243 non-zero arguments.
13245 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13248 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13249 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13250 count, remove one USE.
13252 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13254 * doc/install.texi (Configuration): Document the
13255 --with-long-double-format={ibm,ieee} PowerPC configuration
13259 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13260 Convert from define_expand to be define_insn_and_split. Rework
13261 float/double/_Float128 conversions to QI/HI/SImode to work with
13262 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13263 conversions to QI/HImode types did a store and then a load to
13264 truncate the value. For conversions to VSX registers, don't split
13265 the insn, instead emit the code directly. Use the code iterator
13266 any_fix to combine signed and unsigned conversions.
13267 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13268 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13269 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13270 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13271 (fix_<mode>di2_hw): Likewise.
13272 (fixuns_<mode>di2_hw): Likewise.
13273 (fix_<mode>si2_hw): Likewise.
13274 (fixuns_<mode>si2_hw): Likewise.
13275 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13276 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13277 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13278 fix<uns>_trunc<SFDF:mode>si2_p8.
13279 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13280 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13281 (fix<uns>_<mode>_mem): Likewise.
13282 (fctiw<u>z_<mode>_mem): Likewise.
13283 (fix<uns>_<mode>_mem): Likewise.
13284 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13285 the register allocator from doing a direct move to the GPRs to do
13286 a store, and instead use the ISA 3.0 store byte/half-word from
13287 vector register instruction. For IEEE 128-bit floating point,
13288 also optimize stores of 32-bit ints.
13289 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13291 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13293 * genextract.c (push_pathstr_operand): New function to support
13295 (walk_rtx): Call push_pathstr_operand.
13296 (print_path): Support [a-zA-Z].
13298 2018-02-07 Richard Biener <rguenther@suse.de>
13300 PR tree-optimization/84037
13301 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13302 (cse_and_gimplify_to_preheader): Declare.
13303 (vect_get_place_in_interleaving_chain): Likewise.
13304 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13306 (_loop_vec_info::~_loop_vec_info): Delete it.
13307 (cse_and_gimplify_to_preheader): New function.
13308 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13309 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13310 (vectorizable_load): Likewise. For grouped stores always base
13311 the IV on the first element.
13312 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13313 condition before gimplifying.
13315 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13317 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13318 *DIV_EXPR and *MOD_EXPR.
13320 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13323 * config/i386/i386.c (ix86_option_override_internal): Mask out
13324 the CF_SET bit when checking -fcf-protection.
13326 2018-02-07 Tom de Vries <tom@codesourcery.com>
13329 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13332 2018-02-07 Richard Biener <rguenther@suse.de>
13334 PR tree-optimization/84204
13335 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13338 PR tree-optimization/84205
13339 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13340 special-case isl_ast_op_zdiv_r.
13342 PR tree-optimization/84223
13343 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13344 Only add conditions from within the region.
13345 (gather_bbs::after_dom_children): Adjust.
13347 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13350 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13351 * config/avr/avr.md: Only post-reload split REG-REG moves if
13352 either register is GENERAL_REG_P.
13354 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13356 PR tree-optimization/84235
13357 * tree-ssa-scopedtables.c
13358 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13359 if the subtraction is performed in floating point type where NaNs are
13360 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13361 build 1. Formatting fix.
13363 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13366 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13367 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13368 and skip it regardless of bb boundaries. Use CALL_P macro,
13369 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13370 unnecessarily, formatting fix.
13372 2018-02-06 Michael Collison <michael.collison@arm.com>
13374 * config/arm/thumb2.md:
13375 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13376 (*thumb_mov_notscc): Ditto.
13378 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13381 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13382 unsigned_fix, not "s".
13384 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13386 * configure.ac (gcc_fn_eh_frame_ro): New function.
13387 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13388 correct .eh_frame permissions.
13389 * configure: Regenerate.
13391 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13393 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13394 irrelevant options.
13396 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13398 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13399 Display warning message for -mno-speculate-indirect-jumps.
13401 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13403 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13405 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13407 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13409 PR tree-optimization/84225
13410 * tree-eh.c (find_trapping_overflow): Only call
13411 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13413 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13416 * config/i386/i386.c: Reimplement the check of possible options
13417 -mibt/-mshstk conbination. Change error messages.
13418 * doc/invoke.texi: Fix a typo: remove extra '='.
13420 2018-02-06 Marek Polacek <polacek@redhat.com>
13422 PR tree-optimization/84228
13423 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13425 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13428 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13429 emitted arch directives.
13430 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13431 __ARM_FEATURE_COPROC before changing architectures.
13433 2018-02-06 Richard Biener <rguenther@suse.de>
13435 * config/i386/i386.c (print_reg): Fix typo.
13436 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13438 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13440 * configure: Regenerate.
13442 2018-02-05 Martin Sebor <msebor@redhat.com>
13444 PR tree-optimization/83369
13445 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13448 2018-02-05 Martin Liska <mliska@suse.cz>
13450 * doc/invoke.texi: Cherry-pick upstream r323995.
13452 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13454 * ira.c (ira_init_register_move_cost): Adjust comment.
13456 2018-02-05 Martin Liska <mliska@suse.cz>
13458 PR gcov-profile/84137
13459 * doc/gcov.texi: Fix typo in documentation.
13461 2018-02-05 Martin Liska <mliska@suse.cz>
13463 PR gcov-profile/83879
13464 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13465 using dlopen functionality.
13467 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13469 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13470 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13471 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13472 _mm_maskz_range_round_ss): New intrinsics.
13473 (__builtin_ia32_rangesd128_round)
13474 (__builtin_ia32_rangess128_round): Remove.
13475 (__builtin_ia32_rangesd128_mask_round,
13476 __builtin_ia32_rangess128_mask_round): New builtins.
13477 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13478 __builtin_ia32_rangess128_round): Remove.
13479 (__builtin_ia32_rangesd128_mask_round,
13480 __builtin_ia32_rangess128_mask_round): New builtins.
13481 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13482 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13483 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13484 "<round_saeonly_constraint>")): Changed to ...
13485 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13486 "<round_saeonly_scalar_constraint>")): ... this.
13487 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13488 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13489 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13490 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13491 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13493 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13495 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13497 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13498 Remove all values except native, 8540 and 8548.
13500 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13502 * config/i386/i386.c (ix86_output_function_return): Pass
13503 INVALID_REGNUM, instead of -1, as invalid register number to
13504 indirect_thunk_name and output_indirect_thunk.
13506 2018-02-02 Julia Koval <julia.koval@intel.com>
13508 * config.gcc: Add -march=icelake.
13509 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13510 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13511 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13512 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13513 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13514 (processor_target_table): Add icelake.
13515 (ix86_option_override_internal): Handle new PTAs.
13516 (get_builtin_code_for_version): Handle icelake.
13517 (M_INTEL_COREI7_ICELAKE): New.
13518 (fold_builtin_cpu): Handle icelake.
13519 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13520 * doc/invoke.texi: Add -march=icelake.
13522 2018-02-02 Julia Koval <julia.koval@intel.com>
13524 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13525 to wide_int_bitmask.
13526 * wide-int-bitmask.h: New.
13528 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13531 * config/i386/i386.md: Replace Pmode with word_mode in
13532 builtin_setjmp_setup and builtin_longjmp to support x32.
13534 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13538 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13540 (rs6000_supported_cpu_names): New static variable.
13541 (linux_cpu_translation_table): Likewise.
13542 (elf_platform) <cpu>: Define new static variable and use it.
13543 Translate kernel AT_PLATFORM name to canonical name if needed.
13544 Error if platform name is unknown.
13546 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13549 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13551 2018-02-01 Jeff Law <law@redhat.com>
13554 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13555 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13556 the scratch if RELEASE_VIA_POP is false.
13557 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13558 If we have to save a temporary register, decrement SIZE appropriately.
13559 Pass new arguments to release_scratch_register_on_entry.
13560 (ix86_adjust_stack_and_probe): Likewise.
13561 (ix86_emit_probe_stack_range): Pass new arguments to
13562 release_scratch_register_on_entry.
13564 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13566 PR rtl-optimization/84157
13567 * combine.c (change_zero_ext): Use REG_P predicate in
13568 front of HARD_REGISTER_P predicate.
13570 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13572 * config/avr/avr.c (avr_option_override): Move disabling of
13573 -fdelete-null-pointer-checks to...
13574 * common/config/avr/avr-common.c (avr_option_optimization_table):
13577 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13579 PR tree-optimization/81635
13580 * tree-data-ref.c (split_constant_offset_1): For types that
13581 wrap on overflow, try to use range info to prove that wrapping
13584 2018-02-01 Renlin Li <renlin.li@arm.com>
13587 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13588 TAILCALL_ADDR_REGS.
13589 (aarch64_register_move_cost): Likewise.
13590 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13591 TAILCALL_ADDR_REGS.
13592 (REG_CLASS_NAMES): Likewise.
13593 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13594 TAILCALL_ADDR_REGS. Remove IP registers.
13595 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13597 2018-02-01 Richard Biener <rguenther@suse.de>
13599 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13600 for specifying RPO order and allow NULL for that.
13601 * domwalk.c (dom_walker::dom_walker): Likewise.
13602 (dom_walker::walk): Handle NULL RPO order.
13603 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13605 (rewrite_update_dom_walker): Likewise.
13606 (mark_def_dom_walker): Likewise.
13608 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13610 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13611 (aarch64_maybe_expand_sve_subreg_move): Declare.
13612 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13613 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13615 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13616 that are semantically a reverse operation.
13617 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13618 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13619 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13621 (aarch64_can_change_mode_class): For big-endian, forbid changes
13622 between two SVE modes if they have different element sizes.
13624 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13626 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13627 the TImode handling for big-endian targets.
13629 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13631 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13632 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13634 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13635 Remove BSWAP handing for big-endian targets and use the form of
13636 LD1RQ appropariate for the mode.
13638 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13640 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13641 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13642 duplicated element.
13644 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13647 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13648 check for operands that need to go through aarch64_sve_reload_be.
13650 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13652 PR tree-optimization/81661
13653 PR tree-optimization/84117
13654 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13655 * tree-eh.c: Include gimplify.h.
13656 (find_trapping_overflow, replace_trapping_overflow,
13657 rewrite_to_non_trapping_overflow): New functions.
13658 * tree-vect-loop.c: Include tree-eh.h.
13659 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13660 * tree-data-ref.c: Include tree-eh.h.
13661 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13663 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13665 PR rtl-optimization/84123
13666 * combine.c (change_zero_ext): Check if hard register satisfies
13667 can_change_dest_mode before calling gen_lowpart_SUBREG.
13669 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13672 * ira.c (ira_init_register_move_cost): Remove assert.
13674 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13676 PR rtl-optimization/84071
13677 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13678 * doc/tm.texi: Regenerate.
13680 2018-01-31 Richard Biener <rguenther@suse.de>
13682 PR tree-optimization/84132
13683 * tree-data-ref.c (analyze_miv_subscript): Properly
13684 check whether evolution_function_is_affine_multivariate_p
13685 before calling gcd_of_steps_may_divide_p.
13687 2018-01-31 Julia Koval <julia.koval@intel.com>
13690 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13691 * config/i386/i386.md (rdpid_rex64) New.
13692 (rdpid): Make 32bit only.
13694 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13697 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13698 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13700 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13703 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13705 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13707 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13709 PR rtl-optimization/84071
13710 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13711 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13713 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13715 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13716 (arc_attribute_table): Add 'aux' attribute.
13717 (arc_in_small_data_p): Consider aux like variables.
13718 (arc_is_aux_reg_p): New function.
13719 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13720 (arc_get_aux_arg): New function.
13721 (prepare_move_operands): Handle aux-register access.
13722 (arc_handle_aux_attribute): New function.
13723 * doc/extend.texi (ARC Variable attributes): Add subsection.
13725 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13727 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13728 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13729 (arc_attribute_table): Add 'uncached' attribute.
13730 (arc_print_operand): Print '.di' flag for uncached memory
13732 (arc_in_small_data_p): Do not consider for small data the uncached
13734 (arc_is_uncached_mem_p): New function.
13735 * config/arc/predicates.md (compact_store_memory_operand): Check
13736 for uncached memory accesses.
13737 (nonvol_nonimm_operand): Likewise.
13738 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13740 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13743 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13744 falign-loops=): Add Optimization flag.
13746 2018-01-30 Jeff Law <law@redhat.com>
13749 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13750 INT_REGISTERS_SAVED. Check it prior to calling
13751 get_scratch_register_on_entry.
13752 (ix86_adjust_stack_and_probe): Similarly.
13753 (ix86_emit_probe_stack_range): Similarly.
13754 (ix86_expand_prologue): Corresponding changes.
13756 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13759 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13760 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13762 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13765 * lra-constraints.c (curr_insn_transform): Process AND in the
13768 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13770 PR rtl-optimization/83986
13771 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13772 dependence against last_pending_memory_flush in addition to
13773 pending_jump_insns.
13775 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13777 PR tree-optimization/81611
13778 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13781 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13784 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13787 2018-01-30 Richard Biener <rguenther@suse.de>
13788 Jakub Jelinek <jakub@redhat.com>
13790 PR tree-optimization/84111
13791 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13792 inner loops added during recursion, as they don't have up-to-date
13795 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13798 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13799 (can_inline_edge_by_limits_p): ... here.
13800 (can_early_inline_edge_p, check_callers,
13801 update_caller_keys, update_callee_keys, recursive_inlining,
13802 add_new_edges_to_heap, speculation_useful_p,
13803 inline_small_functions,
13804 inline_small_functions, flatten_function,
13805 inline_to_all_callers_1): Update.
13807 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13809 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13812 2018-01-30 Richard Biener <rguenther@suse.de>
13814 PR tree-optimization/83008
13815 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13816 invariant and constant vector uses in stmts when they need
13817 more than one stmt.
13819 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13822 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13823 * configure: Regenerate.
13825 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13827 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13829 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13830 Use gen_rtx_REG rather than gen_lowpart.
13832 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13834 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13835 rather than 0 when creating partial subregs.
13837 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13839 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13842 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13845 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13846 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13847 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13848 flags. This restores the settings used before the 2017-07-24.
13849 Turning off pre increment/decrement/modify allows IVOPTS to
13850 optimize DF/SF loops where the index is an int.
13852 2018-01-29 Richard Biener <rguenther@suse.de>
13853 Kelvin Nilsen <kelvin@gcc.gnu.org>
13856 * tree-vect-stmts.c (vectorizable_call): Don't call
13857 targetm.vectorize_builtin_md_vectorized_function if callee is
13860 2018-01-22 Carl Love <cel@us.ibm.com>
13862 * doc/extend.tex: Fix typo in second arg in
13863 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13865 2018-01-29 Richard Biener <rguenther@suse.de>
13867 PR tree-optimization/84086
13868 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13869 (flush_ssaname_freelist): When SSA names were released reset
13870 the SCEV hash table.
13872 2018-01-29 Richard Biener <rguenther@suse.de>
13874 PR tree-optimization/84057
13875 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13876 removed paths when removing edges.
13878 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13880 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13881 -mfunction-return=@var{choice}.
13883 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13885 PR diagnostic/84034
13886 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13887 Handle CR like TAB.
13888 (layout::print_source_line): Likewise.
13889 (test_get_line_width_without_trailing_whitespace): Add test cases.
13891 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13893 PR middle-end/84040
13894 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13897 2018-01-26 Jim Wilson <jimw@sifive.com>
13899 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13901 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13904 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13906 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13907 and CMP + SUB-immediate -> SUBS.
13909 2018-01-26 Martin Sebor <msebor@redhat.com>
13911 PR tree-optimization/83896
13912 * tree-ssa-strlen.c (get_string_len): Rename...
13913 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13914 Avoid assuming length is constant.
13915 (handle_char_store): Use HOST_WIDE_INT for string length.
13917 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13920 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13921 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13923 2018-01-26 Richard Biener <rguenther@suse.de>
13925 PR rtl-optimization/84003
13926 * dse.c (record_store): Only record redundant stores when
13927 the earlier store aliases at least all accesses the later one does.
13929 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13931 PR rtl-optimization/83985
13932 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13933 REG_CFA_RESTORE insns.
13934 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13935 REG_CFA_RESTORE insns here.
13938 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13939 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13941 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13943 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13944 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13945 (arc_init): Likewise.
13946 (arc_override_options): Likewise.
13947 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13949 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13951 * config/arc/arc.h (TARGET_DBNZ): Define.
13952 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13953 properly set the tune attribute.
13954 (dbnz): Use TARGET_DBNZ guard.
13955 * config/arc/arc.opt (mtune): Add core3 option.
13957 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13959 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13960 recognize new pic like addresses.
13961 (arc_delegitimize_address): Clean up.
13963 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13965 * config/arc/arc-arches.def: Option mrf16 valid for all
13967 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13968 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13969 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13970 * config/arc/arc-tables.opt: Regenerate.
13971 * config/arc/arc.c (arc_conditional_register_usage): Handle
13972 reduced register file case.
13973 (arc_file_start): Set must have build attributes.
13974 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13975 mrf16 option value.
13976 * config/arc/arc.opt (mrf16): Add new option.
13977 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13978 * config/arc/genmultilib.awk: Handle new mrf16 option.
13979 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13980 * config/arc/t-multilib: Regenerate.
13981 * doc/invoke.texi (ARC Options): Document mrf16 option.
13983 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13985 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13986 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13987 (arc_attribute_table): Add 'secure_call' attribute.
13988 (arc_print_operand): Print secure call operand.
13989 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13991 (arc_is_secure_call_p): New function. * config/arc/arc.md
13992 (call_i): Add support for sjli instruction.
13993 (call_value_i): Likewise.
13994 * config/arc/constraints.md (Csc): New constraint.
13996 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13997 John Eric Martin <John.Martin@emmicro-us.com>
13999 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14000 * config/arc/arc.c (_arc_jli_section): New struct.
14001 (arc_jli_section): New type.
14002 (rc_jli_sections): New static variable.
14003 (arc_handle_jli_attribute): New function.
14004 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14005 (arc_file_end): New function.
14006 (TARGET_ASM_FILE_END): Define.
14007 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14008 (arc_add_jli_section): New function.
14009 (jli_call_scan): Likewise.
14010 (arc_reorg): Call jli_call_scan.
14011 (arc_output_addsi): Remove 'S' from printing asm operand.
14012 (arc_is_jli_call_p): New function.
14013 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14015 (movhi_insn): Likewise.
14016 (movsi_insn): Likewise.
14017 (movsi_set_cc_insn): Likewise.
14018 (loadqi_update): Likewise.
14019 (load_zeroextendqisi_update): Likewise.
14020 (load_signextendqisi_update): Likewise.
14021 (loadhi_update): Likewise.
14022 (load_zeroextendhisi_update): Likewise.
14023 (load_signextendhisi_update): Likewise.
14024 (loadsi_update): Likewise.
14025 (loadsf_update): Likewise.
14026 (movsicc_insn): Likewise.
14027 (bset_insn): Likewise.
14028 (bxor_insn): Likewise.
14029 (bclr_insn): Likewise.
14030 (bmsk_insn): Likewise.
14031 (bicsi3_insn): Likewise.
14032 (cmpsi_cc_c_insn): Likewise.
14033 (movsi_ne): Likewise.
14034 (movsi_cond_exec): Likewise.
14035 (clrsbsi2): Likewise.
14036 (norm_f): Likewise.
14046 (call_i): Remove 'S' asm letter, add jli instruction.
14047 (call_value_i): Likewise.
14048 * config/arc/arc.op (mjli-always): New option.
14049 * config/arc/constraints.md (Cji): New constraint.
14050 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14052 (subsf3_fpx): Likewise.
14053 (mulsf3_fpx): Likewise.
14054 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14056 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14057 function attrbutes.
14058 * doc/invoke.texi (ARC): Document mjli-always option.
14060 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14062 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14063 avoid addition with 0 and use incw and decw where possible.
14065 2018-01-26 Richard Biener <rguenther@suse.de>
14067 PR tree-optimization/81082
14068 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14069 association if it requires casting to unsigned.
14070 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14071 from fold_plusminus_mult_expr to catch important cases late when
14072 range info is available.
14074 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14076 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14077 * configure.ac (hidden_linkonce): New test.
14078 * configure: Regenerate.
14079 * config.in: Regenerate.
14081 2018-01-26 Julia Koval <julia.koval@intel.com>
14083 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14084 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14085 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14086 _mm_mask_bitshuffle_epi64_mask): Fix type.
14087 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14088 USI_FTYPE_V4DI_V4DI_USI): Remove.
14089 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14090 __builtin_ia32_vpshufbitqmb256_mask,
14091 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14092 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14093 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14095 2018-01-26 Alan Modra <amodra@gmail.com>
14098 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14099 UNSPEC_VBPERMQ. Sort other unspecs.
14101 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14103 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14105 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14107 PR middle-end/83055
14108 * predict.c (drop_profile): Do not push/pop cfun; update also
14110 (handle_missing_profiles): Fix logic looking for zero profiles.
14112 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14114 PR middle-end/83977
14115 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14116 on functions with #pragma omp declare simd or functions with simd
14118 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14119 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14120 Remove trailing \n from warning_at calls.
14122 2018-01-25 Tom de Vries <tom@codesourcery.com>
14125 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14126 for neutered workers.
14128 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14131 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14132 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14134 2018-01-24 Jeff Law <law@redhat.com>
14137 * i386.c (get_probe_interval): Move to earlier point.
14138 (ix86_compute_frame_layout): If -fstack-clash-protection and
14139 the frame is larger than the probe interval, then use pushes
14140 to save registers rather than reg->mem moves.
14141 (ix86_expand_prologue): Remove conditional for int_registers_saved
14144 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14147 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14148 min/max for never referenced object.
14150 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14152 PR middle-end/83977
14153 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14155 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14156 attributes from DECL_ATTRIBUTES (decl) without affecting
14157 DECL_ATTRIBUTES (current_function_decl).
14158 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14159 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14161 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14163 PR tree-optimization/83979
14164 * fold-const.c (fold_comparison): Use constant_boolean_node
14165 instead of boolean_{true,false}_node.
14167 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14169 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14172 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14174 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14175 Simplify the clause that sets the length attribute.
14176 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14177 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14178 clause that sets the length attribute.
14179 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14181 2018-01-24 Tom de Vries <tom@codesourcery.com>
14184 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14185 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14186 Add strict parameter.
14187 (prevent_branch_around_nothing): Insert dummy insn between branch to
14188 label and label with no ptx insn inbetween.
14189 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14191 2018-01-24 Tom de Vries <tom@codesourcery.com>
14194 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14195 for neutered threads in warp.
14196 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14198 2018-01-24 Richard Biener <rguenther@suse.de>
14200 PR tree-optimization/83176
14201 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14204 2018-01-24 Richard Biener <rguenther@suse.de>
14206 PR tree-optimization/82819
14207 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14208 code generating pluses that are no-ops in the target precision.
14210 2018-01-24 Richard Biener <rguenther@suse.de>
14212 PR middle-end/84000
14213 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14215 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14217 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14218 to merge probabilities.
14219 * predict.c (probably_never_executed): Also mark as cold functions
14220 with global 0 profile and guessed local profile.
14221 * profile-count.c (profile_probability::combine_with_count): New
14223 * profile-count.h (profile_probability::operator*,
14224 profile_probability::operator*=, profile_probability::operator/,
14225 profile_probability::operator/=): Reduce precision to adjusted
14226 and set value to guessed on contradictory divisions.
14227 (profile_probability::combine_with_freq): Remove.
14228 (profile_probability::combine_wiht_count): Declare.
14229 (profile_count::force_nonzero):: Set to adjusted.
14230 (profile_count::probability_in):: Set quality to adjusted.
14231 * tree-ssa-tail-merge.c (replace_block_by): Use
14232 combine_with_count.
14234 2018-01-23 Andrew Waterman <andrew@sifive.com>
14235 Jim Wilson <jimw@sifive.com>
14237 * config/riscv/riscv.c (riscv_stack_boundary): New.
14238 (riscv_option_override): Set riscv_stack_boundary. Handle
14239 riscv_preferred_stack_boundary_arg.
14240 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14241 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14242 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14243 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14244 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14245 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14247 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14250 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14251 of struct ix86_frame.
14252 (ix86_expand_epilogue): Likewise. Add a local variable for
14253 the reg_save_offset field in struct ix86_frame.
14255 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14257 PR tree-optimization/82604
14258 * tree-loop-distribution.c (enum partition_kind): New enum item
14259 PKIND_PARTIAL_MEMSET.
14260 (partition_builtin_p): Support above new enum item.
14261 (generate_code_for_partition): Ditto.
14262 (compute_access_range): Differentiate cases that equality can be
14263 proven at all loops, the innermost loops or no loops.
14264 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14265 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14266 (finalize_partitions, distribute_loop): Don't fuse partition of
14267 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14268 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14269 parloop is enabled.
14271 2018-01-23 Martin Liska <mliska@suse.cz>
14273 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14274 order to ignore the predictor.
14275 (PRED_POLYMORPHIC_CALL): Likewise.
14276 (PRED_RECURSIVE_CALL): Likewise.
14278 2018-01-23 Martin Liska <mliska@suse.cz>
14280 * tree-profile.c (tree_profiling): Print function header to
14281 aware reader which function we are working on.
14282 * value-prof.c (gimple_find_values_to_profile): Do not print
14283 not interesting value histograms.
14285 2018-01-23 Martin Liska <mliska@suse.cz>
14287 * profile-count.h (enum profile_quality): Add
14288 profile_uninitialized as the first value. Do not number values
14289 as they are zero based.
14290 (profile_count::verify): Update sanity check.
14291 (profile_probability::verify): Likewise.
14293 2018-01-23 Nathan Sidwell <nathan@acm.org>
14295 * doc/invoke.texi (ffor-scope): Deprecate.
14297 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14299 PR tree-optimization/83510
14300 * domwalk.c (set_all_edges_as_executable): New function.
14301 (dom_walker::dom_walker): Convert bool param
14302 "skip_unreachable_blocks" to enum reachability. Move setup of
14303 edge flags to set_all_edges_as_executable and only do it when
14304 reachability is REACHABLE_BLOCKS.
14305 * domwalk.h (enum dom_walker::reachability): New enum.
14306 (dom_walker::dom_walker): Convert bool param
14307 "skip_unreachable_blocks" to enum reachability.
14308 (set_all_edges_as_executable): New decl.
14309 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14310 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14312 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14313 but converting true to REACHABLE_BLOCKS.
14314 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14316 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14317 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14318 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14320 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14321 if check_all_array_refs will be called.
14323 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14325 * tree.c (selftest::test_location_wrappers): Add more test
14328 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14330 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14331 (selftest::test_bit_in_range): Likewise.
14333 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14336 * doc/sourcebuild.texi (vect_float): Say that the selector
14337 only describes the situation when -funsafe-math-optimizations is on.
14338 (vect_float_strict): Document.
14340 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14342 PR tree-optimization/83965
14343 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14344 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14345 instead of checking only for a reduction.
14346 (vect_recog_widen_sum_pattern): Likewise.
14348 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14350 * predict.c (probably_never_executed): Only use precise profile info.
14351 (compute_function_frequency): Skip after inlining hack since we now
14352 have quality checking.
14354 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14356 * profile-count.h (profile_probability::very_unlikely,
14357 profile_probability::unlikely, profile_probability::even): Set
14358 precision to guessed.
14360 2018-01-23 Richard Biener <rguenther@suse.de>
14362 PR tree-optimization/83963
14363 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14364 Properly terminate dominator walk when crossing the exit edge not
14365 when visiting its source block.
14367 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14370 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14371 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14373 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14375 PR tree-optimization/83957
14376 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14377 semicolon after for body surrounded by braces.
14379 PR tree-optimization/83081
14380 * profile-count.h (profile_probability::split): New method.
14381 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14382 Use profile_probability::split.
14383 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14384 when splitting a single conditional jump into 2.
14386 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14388 PR tree-optimization/69452
14389 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14392 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14394 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14395 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14396 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14398 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14400 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14402 * config/rl78/rl78.md (movdi): New define_expand.
14403 * config/rl78/rl78.c (rl78_split_movdi): New function.
14405 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14408 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14410 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14411 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14412 128-bit to produce an UNSPEC move to get the double word with the
14413 signbit and then a shift directly to do signbit.
14414 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14415 implementation with a new version that just does either a direct
14416 move or a regular move. Move memory interface to separate insns.
14417 Move insns so they are next to the expander.
14418 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14419 with signbit move. Split big and little endian case.
14420 (signbit<mode>2_dm_mem_le): Likewise.
14421 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14422 (signbit<mode>2_dm2): Likewise.
14424 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14426 * config/rl78/rl78.md (anddi3): New define_expand.
14428 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14430 * config/rl78/rl78.md (umindi3): New define_expand.
14432 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14434 * config/rl78/rl78.md (smindi3): New define_expand.
14436 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14438 * config/rl78/rl78.md (smaxdi3): New define_expand.
14440 2018-01-22 Carl Love <cel@us.ibm.com>
14442 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14443 LVX_V1TI): Add macro expansion.
14444 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14445 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14446 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14447 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14448 Change check to determine if the instruction is a byte reversing
14449 entry. Fix typo in comment.
14450 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14451 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14452 Add def_builtin calls for new builtins.
14453 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14454 Add define_insn expansion.
14456 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14458 * config/rl78/rl78.md (umaxdi3): New define_expand.
14460 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14462 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14463 for non-QImode registers.
14465 2018-01-22 Richard Biener <rguenther@suse.de>
14467 PR tree-optimization/83963
14468 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14469 including the loop exit block.
14470 (scop_detection::merge_sese): Likewise.
14471 (scop_detection::add_scop): Do it here instead.
14473 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14475 * doc/sourcebuild.texi (arm_softfloat): Document.
14477 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14480 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14481 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14482 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14484 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14485 David Edelsohn <dje.gcc@gmail.com>
14488 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14489 Change "crset eq" to "crset 2".
14490 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14491 (*call_indirect_aix<mode>_nospec): Likewise.
14492 (*call_value_indirect_aix<mode>_nospec): Likewise.
14493 (*call_indirect_elfv2<mode>_nospec): Likewise.
14494 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14495 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14496 change assembly output from . to $.
14497 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14498 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14499 (*tablejump<mode>_internal1_nospec): Likewise.
14501 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14504 * config/sh/sh_optimize_sett_clrt.cc:
14505 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14507 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14509 PR tree-optimization/83940
14510 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14511 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14512 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14513 use it to pass back the definition type.
14514 (vect_check_store_rhs): Likewise rhs_dt_out.
14515 (vect_build_gather_load_calls): Add a mask_dt argument and use
14516 it instead of a call to vect_is_simple_use.
14517 (vectorizable_store): Update calls to vect_check_load_store_mask
14518 and vect_check_store_rhs. Use the dt returned by the latter instead
14519 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14520 instead of calls to vect_is_simple_use. Pass the scalar rather
14521 than the vector operand to vect_is_simple_use when handling
14522 second and subsequent copies of an rhs value.
14523 (vectorizable_load): Update calls to vect_check_load_store_mask
14524 and vect_build_gather_load_calls. Use the cached mask_dt and
14525 gs_info.offset_dt instead of calls to vect_is_simple_use.
14527 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14529 PR middle-end/83945
14530 * tree-emutls.c: Include gimplify.h.
14531 (lower_emutls_2): New function.
14532 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14533 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14534 it before further processing.
14537 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14538 UINTVAL (trueop1) instead of INTVAL (op1).
14540 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14544 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14545 INCOMING_FRAME_SP_OFFSET if not defined.
14546 (scan_trace): Add ENTRY argument. If true and
14547 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14548 emit a note to adjust the CFA offset.
14549 (create_cfi_notes): Adjust scan_trace callers.
14550 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14551 INCOMING_FRAME_SP_OFFSET in the CIE.
14552 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14553 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14555 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14556 * doc/tm.texi: Regenerated.
14558 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14560 PR rtl-optimization/83147
14561 * lra-constraints.c (remove_inheritance_pseudos): Use
14562 lra_substitute_pseudo_within_insn.
14564 2018-01-19 Tom de Vries <tom@codesourcery.com>
14565 Cesar Philippidis <cesar@codesourcery.com>
14568 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14570 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14573 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14574 spaces for function labels.
14576 2018-01-19 Martin Liska <mliska@suse.cz>
14578 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14579 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14580 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14581 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14582 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14583 (PRED_CONST_RETURN): Change from 69 to 65.
14584 (PRED_NULL_RETURN): Change from 91 to 71.
14585 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14586 (PRED_LOOP_GUARD): Change from 66 to 73.
14588 2018-01-19 Martin Liska <mliska@suse.cz>
14590 * predict.c (predict_insn_def): Add new assert.
14591 (struct branch_predictor): Change type to signed integer.
14592 (test_prediction_value_range): Amend test to cover
14593 PROB_UNINITIALIZED.
14594 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14595 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14596 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14597 (PRED_LOOP_IV_COMPARE): Likewise.
14598 * predict.h (PROB_UNINITIALIZED): Define new constant.
14600 2018-01-19 Martin Liska <mliska@suse.cz>
14602 * predict.c (dump_prediction): Add new format for
14603 analyze_brprob.py script which is enabled with -details
14605 * profile-count.h (precise_p): New function.
14607 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14609 PR tree-optimization/83922
14610 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14611 there are no statements that need masking.
14612 (vect_active_double_reduction_p): New function.
14613 (vect_analyze_loop_operations): Use it when handling phis that
14614 are not in the loop header.
14616 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14618 PR tree-optimization/83914
14619 * tree-vect-loop.c (vectorizable_induction): Don't convert
14620 init_expr or apply the peeling adjustment for inductions
14621 that are nested within the vectorized loop.
14623 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14625 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14628 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14632 * function.h (gimplify_parameters): Add gimple_seq * argument.
14633 * function.c: Include gimple.h and options.h.
14634 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14635 for the added local temporaries if needed.
14636 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14637 if there are any parameter cleanups, wrap whole body into a
14638 try/finally with the cleanups.
14640 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14643 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14644 Use GET_MODE_CLASS for scalar floating point.
14646 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14650 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14651 Fix call of call_cgraph_insertion_hooks.
14653 2018-01-18 Martin Sebor <msebor@redhat.com>
14655 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14657 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14660 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14663 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14666 * common.opt: (-ffile-prefix-map): New option.
14667 * opts.c (common_handle_option): Defer it.
14668 * opts-global.c (handle_common_deferred_options): Handle it.
14669 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14670 * file-prefix-map.h: New file.
14671 (remap_debug_filename, add_debug_prefix_map): ...here.
14672 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14673 * final.c (debug_prefix_map, add_debug_prefix_map
14674 remap_debug_filename): Move to...
14675 * file-prefix-map.c: New file.
14676 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14677 generalize, get rid of alloca(), use strrchr() instead of strchr().
14678 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14679 Implement in terms of add_prefix_map().
14680 (remap_macro_filename, remap_debug_filename): Implement in term of
14682 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14683 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14684 * dbxout.c: Include file-prefix-map.h.
14685 * varasm.c: Likewise.
14686 * vmsdbgout.c: Likewise.
14687 * xcoffout.c: Likewise.
14688 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14689 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14690 * doc/invoke.texi (-ffile-prefix-map): Document.
14691 (-fdebug-prefix-map): Update description.
14693 2018-01-18 Martin Liska <mliska@suse.cz>
14695 * config/i386/i386.c (indirect_thunk_name): Document that also
14697 (output_indirect_thunk): Document why both instructions
14698 (pause and lfence) are generated.
14700 2018-01-18 Richard Biener <rguenther@suse.de>
14702 PR tree-optimization/83887
14703 * graphite-scop-detection.c
14704 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14705 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14706 (scop_detection::merge_sese): Re-implement with a flood-fill
14707 algorithm that properly finds a SESE region if it exists.
14709 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14712 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14713 pointer_diff optimizations use view_convert instead of convert.
14715 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14717 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14718 Generate different code for -mno-speculate-indirect-jumps.
14719 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14720 (*call_indirect_aix<mode>): Disable for
14721 -mno-speculate-indirect-jumps.
14722 (*call_indirect_aix<mode>_nospec): New define_insn.
14723 (*call_value_indirect_aix<mode>): Disable for
14724 -mno-speculate-indirect-jumps.
14725 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14726 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14727 -mno-speculate-indirect-jumps.
14728 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14730 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14732 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14733 long double type, set the flags for noting the default long double
14734 type, even if we don't pass or return a long double type.
14736 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14739 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14742 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14744 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14745 support for merge[hl].
14746 (fold_mergehl_helper): New helper function.
14747 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14748 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14749 (altivec_vmrglw_direct): Add xxmrglw insn.
14751 2018-01-17 Andrew Waterman <andrew@sifive.com>
14753 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14754 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14756 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14759 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14760 call the lto_location_cache before reading the
14761 DECL_SOURCE_LOCATION of the types.
14763 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14764 Richard Sandiford <richard.sandiford@linaro.org>
14766 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14767 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14768 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14769 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14770 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14772 * config/aarch64/constraints.md (aarch64_movti_operand):
14774 * config/aarch64/predicates.md (Uti): Add new constraint.
14776 2018-01-17 Carl Love <cel@us.ibm.com>
14778 * config/rs6000/vsx.md (define_expand xl_len_r,
14779 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14780 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14782 (define_expand, define_insn): Move the shift left from the
14783 define_insn to the define_expand for lxvl and stxvl instructions.
14784 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14785 and XL_LEN_R definitions to PURE.
14787 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14789 * config/i386/i386.c (indirect_thunk_name): Declare regno
14790 as unsigned int. Compare regno with INVALID_REGNUM.
14791 (output_indirect_thunk): Ditto.
14792 (output_indirect_thunk_function): Ditto.
14793 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14794 in the call to output_indirect_thunk_function.
14796 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14798 PR middle-end/83884
14799 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14800 rather than the size of inner_type to determine the stack slot size
14801 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14803 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14806 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14809 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14811 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14812 endian Linux systems to optionally enable multilibs for selecting
14813 the long double type if the user configured an explicit type.
14814 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14815 have no long double multilibs if not defined.
14816 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14817 warn if the user used -mabi={ieee,ibm}longdouble and we built
14818 multilibs for long double.
14819 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14820 appropriate multilib option.
14821 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14823 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14824 for building long double multilibs.
14825 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14827 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14829 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14832 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14834 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14837 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14840 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14843 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14845 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14846 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14848 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14850 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14851 different rtl trees depending on TARGET_64BIT.
14852 (rs6000_gen_lvx): Likewise.
14854 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14856 * config/visium/visium.md (nop): Tweak comment.
14857 (hazard_nop): Likewise.
14859 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14861 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14862 -mspeculate-indirect-jumps.
14863 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14864 for -mno-speculate-indirect-jumps.
14865 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14866 (*call_value_indirect_elfv2<mode>): Disable for
14867 -mno-speculate-indirect-jumps.
14868 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14869 (indirect_jump): Emit different RTL for
14870 -mno-speculate-indirect-jumps.
14871 (*indirect_jump<mode>): Disable for
14872 -mno-speculate-indirect-jumps.
14873 (*indirect_jump<mode>_nospec): New define_insn.
14874 (tablejump): Emit different RTL for
14875 -mno-speculate-indirect-jumps.
14876 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14877 (tablejumpsi_nospec): New define_expand.
14878 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14879 (tablejumpdi_nospec): New define_expand.
14880 (*tablejump<mode>_internal1): Disable for
14881 -mno-speculate-indirect-jumps.
14882 (*tablejump<mode>_internal1_nospec): New define_insn.
14883 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14886 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14888 * caller-save.c (insert_save): Drop unnecessary parameter. All
14891 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14892 Richard Biener <rguenth@suse.de>
14895 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14896 return early, inline manually is_gimple_sizepos. Make sure if we
14897 call gimplify_expr we don't end up with a gimple constant.
14898 * tree.c (variably_modified_type_p): Don't return true for
14899 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14900 * gimplify.h (is_gimple_sizepos): Remove.
14902 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14904 PR tree-optimization/83857
14905 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14906 vectorizable_live_operation for pure SLP statements.
14907 (vectorizable_live_operation): Handle PHIs.
14909 2018-01-16 Richard Biener <rguenther@suse.de>
14911 PR tree-optimization/83867
14912 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14913 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14915 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14918 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14919 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14920 If off is not INTEGER_CST, issue a may not be aligned warning
14921 rather than isn't aligned. Use isn%'t rather than isn't.
14922 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14924 <case MULT_EXPR>: Improve the case when bottom and one of the
14925 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14926 operand, in that case check if the other operand is multiple of
14927 bottom divided by the INTEGER_CST operand.
14929 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14932 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14933 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14934 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14935 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14936 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14937 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14938 (pa_function_arg_size): New function.
14940 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14942 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14943 in a separate statement.
14945 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14947 PR tree-optimization/83847
14948 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14949 group gathers and scatters.
14951 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14953 PR rtl-optimization/86620
14954 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14956 PR rtl-optimization/83213
14957 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14958 to last if both are JUMP_INSNs.
14960 PR tree-optimization/83843
14961 * gimple-ssa-store-merging.c
14962 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14963 store_immediate_info for bswap/nop orig_stores.
14965 2018-01-15 Andrew Waterman <andrew@sifive.com>
14967 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14969 <UDIV>: Increase cost if !TARGET_DIV.
14971 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14973 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14974 (define_attr "cr_logical_3op"): New.
14975 (cceq_ior_compare): Adjust.
14976 (cceq_ior_compare_complement): Adjust.
14977 (*cceq_rev_compare): Adjust.
14978 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14979 (is_cracked_insn): Adjust.
14980 (insn_must_be_first_in_group): Adjust.
14981 * config/rs6000/40x.md: Adjust.
14982 * config/rs6000/440.md: Adjust.
14983 * config/rs6000/476.md: Adjust.
14984 * config/rs6000/601.md: Adjust.
14985 * config/rs6000/603.md: Adjust.
14986 * config/rs6000/6xx.md: Adjust.
14987 * config/rs6000/7450.md: Adjust.
14988 * config/rs6000/7xx.md: Adjust.
14989 * config/rs6000/8540.md: Adjust.
14990 * config/rs6000/cell.md: Adjust.
14991 * config/rs6000/e300c2c3.md: Adjust.
14992 * config/rs6000/e500mc.md: Adjust.
14993 * config/rs6000/e500mc64.md: Adjust.
14994 * config/rs6000/e5500.md: Adjust.
14995 * config/rs6000/e6500.md: Adjust.
14996 * config/rs6000/mpc.md: Adjust.
14997 * config/rs6000/power4.md: Adjust.
14998 * config/rs6000/power5.md: Adjust.
14999 * config/rs6000/power6.md: Adjust.
15000 * config/rs6000/power7.md: Adjust.
15001 * config/rs6000/power8.md: Adjust.
15002 * config/rs6000/power9.md: Adjust.
15003 * config/rs6000/rs64.md: Adjust.
15004 * config/rs6000/titan.md: Adjust.
15006 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15008 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15009 ix86_indirect_branch_register logic.
15011 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15013 * config/i386/constraints.md (Bs): Update
15014 ix86_indirect_branch_register check. Don't check
15015 ix86_indirect_branch_register with GOT_memory_operand.
15017 * config/i386/predicates.md (GOT_memory_operand): Don't check
15018 ix86_indirect_branch_register here.
15019 (GOT32_symbol_operand): Likewise.
15021 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15023 * config/i386/predicates.md (constant_call_address_operand):
15024 Rewrite ix86_indirect_branch_register logic.
15025 (sibcall_insn_operand): Likewise.
15027 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15029 * config/i386/constraints.md (Bs): Replace
15030 ix86_indirect_branch_thunk_register with
15031 ix86_indirect_branch_register.
15033 * config/i386/i386.md (indirect_jump): Likewise.
15034 (tablejump): Likewise.
15035 (*sibcall_memory): Likewise.
15036 (*sibcall_value_memory): Likewise.
15037 Peepholes of indirect call and jump via memory: Likewise.
15038 * config/i386/i386.opt: Likewise.
15039 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15040 (GOT_memory_operand): Likewise.
15041 (call_insn_operand): Likewise.
15042 (sibcall_insn_operand): Likewise.
15043 (GOT32_symbol_operand): Likewise.
15045 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15047 PR middle-end/83837
15048 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15049 type rather than type addr's type points to.
15050 (expand_omp_atomic_mutex): Likewise.
15051 (expand_omp_atomic): Likewise.
15053 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15056 * config/i386/i386.c (output_indirect_thunk_function): Use
15057 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15058 for __x86_return_thunk.
15060 2018-01-15 Richard Biener <rguenther@suse.de>
15062 PR middle-end/83850
15063 * expmed.c (extract_bit_field_1): Fix typo.
15065 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15068 * config/arm/iterators.md (VF): New mode iterator.
15069 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15070 Remove integer-related logic from pattern.
15071 (neon_vabd<mode>_3): Likewise.
15073 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15075 PR middle-end/82694
15076 * common.opt (fstrict-overflow): No longer an alias.
15077 (fwrapv-pointer): New option.
15078 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15079 also for pointer types based on flag_wrapv_pointer.
15080 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15081 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15082 opts->x_flag_wrapv got set.
15083 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15084 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15085 POINTER_TYPE_OVERFLOW_UNDEFINED.
15086 * match.pd: Likewise in address comparison pattern.
15087 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15089 2018-01-15 Richard Biener <rguenther@suse.de>
15092 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15093 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15094 Reset type names to their identifier if their TYPE_DECL doesn't
15095 have linkage (and thus is used for ODR and devirt).
15096 (save_debug_info_for_decl): Remove.
15097 (save_debug_info_for_type): Likewise.
15098 (add_tree_to_fld_list): Adjust.
15099 * tree-pretty-print.c (dump_generic_node): Make dumping of
15100 type names more robust.
15102 2018-01-15 Richard Biener <rguenther@suse.de>
15104 * BASE-VER: Bump to 8.0.1.
15106 2018-01-14 Martin Sebor <msebor@redhat.com>
15109 * builtins.c (check_access): Avoid warning when the no-warning bit
15112 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15114 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15115 * ira-color (allocno_hard_regs_compare): Likewise.
15117 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15120 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15121 Use .pushsection/.popsection.
15123 2018-01-14 Martin Sebor <msebor@redhat.com>
15126 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15128 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15130 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15131 entry from extra_headers.
15132 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15133 extra_headers, make the list bitwise identical to the i?86-*-* one.
15135 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15137 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15138 -mcmodel=large with -mindirect-branch=thunk,
15139 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15140 -mfunction-return=thunk-extern.
15141 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15142 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15143 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15145 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15147 * config/i386/i386.c (print_reg): Print the name of the full
15148 integer register without '%'.
15149 (ix86_print_operand): Handle 'V'.
15150 * doc/extend.texi: Document 'V' modifier.
15152 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15154 * config/i386/constraints.md (Bs): Disallow memory operand for
15155 -mindirect-branch-register.
15157 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15158 (GOT_memory_operand): Likewise.
15159 (call_insn_operand): Likewise.
15160 (sibcall_insn_operand): Likewise.
15161 (GOT32_symbol_operand): Likewise.
15162 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15163 for -mindirect-branch-register.
15164 (tablejump): Likewise.
15165 (*sibcall_memory): Likewise.
15166 (*sibcall_value_memory): Likewise.
15167 Disallow peepholes of indirect call and jump via memory for
15168 -mindirect-branch-register.
15169 (*call_pop): Replace m with Bw.
15170 (*call_value_pop): Likewise.
15171 (*sibcall_pop_memory): Replace m with Bs.
15172 * config/i386/i386.opt (mindirect-branch-register): New option.
15173 * doc/invoke.texi: Document -mindirect-branch-register option.
15175 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15177 * config/i386/i386-protos.h (ix86_output_function_return): New.
15178 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15179 set function_return_type.
15180 (indirect_thunk_name): Add ret_p to indicate thunk for function
15182 (output_indirect_thunk_function): Pass false to
15183 indirect_thunk_name.
15184 (ix86_output_indirect_branch_via_reg): Likewise.
15185 (ix86_output_indirect_branch_via_push): Likewise.
15186 (output_indirect_thunk_function): Create alias for function
15187 return thunk if regno < 0.
15188 (ix86_output_function_return): New function.
15189 (ix86_handle_fndecl_attribute): Handle function_return.
15190 (ix86_attribute_table): Add function_return.
15191 * config/i386/i386.h (machine_function): Add
15192 function_return_type.
15193 * config/i386/i386.md (simple_return_internal): Use
15194 ix86_output_function_return.
15195 (simple_return_internal_long): Likewise.
15196 * config/i386/i386.opt (mfunction-return=): New option.
15197 (indirect_branch): Mention -mfunction-return=.
15198 * doc/extend.texi: Document function_return function attribute.
15199 * doc/invoke.texi: Document -mfunction-return= option.
15201 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15203 * config/i386/i386-opts.h (indirect_branch): New.
15204 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15205 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15206 with local indirect jump when converting indirect call and jump.
15207 (ix86_set_indirect_branch_type): New.
15208 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15209 (indirectlabelno): New.
15210 (indirect_thunk_needed): Likewise.
15211 (indirect_thunk_bnd_needed): Likewise.
15212 (indirect_thunks_used): Likewise.
15213 (indirect_thunks_bnd_used): Likewise.
15214 (INDIRECT_LABEL): Likewise.
15215 (indirect_thunk_name): Likewise.
15216 (output_indirect_thunk): Likewise.
15217 (output_indirect_thunk_function): Likewise.
15218 (ix86_output_indirect_branch_via_reg): Likewise.
15219 (ix86_output_indirect_branch_via_push): Likewise.
15220 (ix86_output_indirect_branch): Likewise.
15221 (ix86_output_indirect_jmp): Likewise.
15222 (ix86_code_end): Call output_indirect_thunk_function if needed.
15223 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15225 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15226 (ix86_attribute_table): Add indirect_branch.
15227 * config/i386/i386.h (machine_function): Add indirect_branch_type
15228 and has_local_indirect_jump.
15229 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15231 (tablejump): Likewise.
15232 (*indirect_jump): Use ix86_output_indirect_jmp.
15233 (*tablejump_1): Likewise.
15234 (simple_return_indirect_internal): Likewise.
15235 * config/i386/i386.opt (mindirect-branch=): New option.
15236 (indirect_branch): New.
15239 (thunk-inline): Likewise.
15240 (thunk-extern): Likewise.
15241 * doc/extend.texi: Document indirect_branch function attribute.
15242 * doc/invoke.texi: Document -mindirect-branch= option.
15244 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15247 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15249 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15251 * ipa-inline.c (want_inline_small_function_p): Return false if
15252 inlining has already failed with CIF_FINAL_ERROR.
15253 (update_caller_keys): Call want_inline_small_function_p before
15255 (update_callee_keys): Likewise.
15257 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15259 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15261 (rs6000_quadword_masked_address_p): Likewise.
15262 (quad_aligned_load_p): Likewise.
15263 (quad_aligned_store_p): Likewise.
15264 (const_load_sequence_p): Add comment to describe the outer-most loop.
15265 (mimic_memory_attributes_and_flags): New function.
15266 (rs6000_gen_stvx): Likewise.
15267 (replace_swapped_aligned_store): Likewise.
15268 (rs6000_gen_lvx): Likewise.
15269 (replace_swapped_aligned_load): Likewise.
15270 (replace_swapped_load_constant): Capitalize argument name in
15271 comment describing this function.
15272 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15273 and stores that access quad-word aligned addresses and replace
15274 with stvx or lvx instructions when appropriate.
15275 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15276 New function prototype.
15277 (rs6000_quadword_masked_address_p): Likewise.
15278 (rs6000_gen_lvx): Likewise.
15279 (rs6000_gen_stvx): Likewise.
15280 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15281 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15282 when memory address is aligned.
15283 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15284 this split to select lvx instruction when memory address is aligned.
15285 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15286 instruction when memory address is aligned.
15287 (*vsx_le_perm_load_v16qi): Likewise.
15288 (four unnamed splitters): Modify to select the stvx instruction
15289 when memory is aligned.
15291 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15293 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15294 which appears in the queue multiple times.
15296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15297 Alan Hayward <alan.hayward@arm.com>
15298 David Sherwood <david.sherwood@arm.com>
15300 * tree-vectorizer.h (vec_lower_bound): New structure.
15301 (_loop_vec_info): Add check_nonzero and lower_bounds.
15302 (LOOP_VINFO_CHECK_NONZERO): New macro.
15303 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15304 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15305 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15306 fields. Make seg_len the distance travelled, not including the
15308 (dr_direction_indicator): Declare.
15309 (dr_zero_step_indicator): Likewise.
15310 (dr_known_forward_stride_p): Likewise.
15311 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15313 (runtime_alias_check_p): Allow runtime alias checks with
15315 (operator ==): Compare access_size and align.
15316 (prune_runtime_alias_test_list): Rework for new distinction between
15317 the access_size and seg_len.
15318 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15320 (get_segment_min_max): New function.
15321 (create_intersect_range_checks): Use it.
15322 (dr_step_indicator): New function.
15323 (dr_direction_indicator): Likewise.
15324 (dr_zero_step_indicator): Likewise.
15325 (dr_known_forward_stride_p): Likewise.
15326 * tree-loop-distribution.c (data_ref_segment_size): Return
15327 DR_STEP * (niters - 1).
15328 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15330 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15331 (vect_preserves_scalar_order_p): New function, split out from...
15332 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15333 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15334 (vect_vfa_access_size): New function.
15335 (vect_vfa_align): Likewise.
15336 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15337 (dump_lower_bound): New function.
15338 (vect_check_lower_bound): Likewise.
15339 (vect_small_gap_p): Likewise.
15340 (vectorizable_with_step_bound_p): Likewise.
15341 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15342 depencies if the vectorization factor is 1. Convert the checks
15343 for nonzero steps into checks on the bounds of DR_STEP. Try using
15344 a bunds check for variable steps if the minimum required step is
15345 relatively small. Update calls to the dr_with_seg_len
15346 constructor and to vect_compile_time_alias.
15347 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15349 (vect_loop_versioning): Call it.
15350 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15352 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15355 Alan Hayward <alan.hayward@arm.com>
15356 David Sherwood <david.sherwood@arm.com>
15358 * doc/sourcebuild.texi (vect_scatter_store): Document.
15359 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15361 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15363 * genopinit.c (main): Add supports_vec_scatter_store and
15364 supports_vec_scatter_store_cached to target_optabs.
15365 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15366 IFN_MASK_SCATTER_STORE.
15367 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15369 * internal-fn.h (internal_store_fn_p): Declare.
15370 (internal_fn_stored_value_index): Likewise.
15371 * internal-fn.c (scatter_store_direct): New macro.
15372 (expand_scatter_store_optab_fn): New function.
15373 (direct_scatter_store_optab_supported_p): New macro.
15374 (internal_store_fn_p): New function.
15375 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15376 IFN_MASK_SCATTER_STORE.
15377 (internal_fn_mask_index): Likewise.
15378 (internal_fn_stored_value_index): New function.
15379 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15380 for scatter stores.
15381 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15382 * optabs-query.c (supports_vec_scatter_store_p): New function.
15383 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15384 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15385 true for scatter stores.
15386 (vect_gather_scatter_fn_p): Handle scatter stores too.
15387 (vect_check_gather_scatter): Consider using scatter stores if
15388 supports_vec_scatter_store_p.
15389 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15390 scatter stores too.
15391 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15392 internal_fn_stored_value_index.
15393 (check_load_store_masking): Handle scatter stores too.
15394 (vect_get_store_rhs): Make public.
15395 (vectorizable_call): Use internal_store_fn_p.
15396 (vectorizable_store): Handle scatter store internal functions.
15397 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15398 when deciding whether the end of the group has been reached.
15399 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15400 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15401 (mask_scatter_store<mode>): New insns.
15403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15404 Alan Hayward <alan.hayward@arm.com>
15405 David Sherwood <david.sherwood@arm.com>
15407 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15408 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15409 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15411 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15412 Use vect_truncate_gather_scatter_offset if we can't treat the
15413 operation as a normal gather load or scatter store.
15414 (get_group_load_store_type): Take the gather_scatter_info
15415 as argument. Try using a gather load or scatter store for
15416 single-element groups.
15417 (get_load_store_type): Update calls to get_group_load_store_type
15418 and vect_use_strided_gather_scatters_p.
15420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15421 Alan Hayward <alan.hayward@arm.com>
15422 David Sherwood <david.sherwood@arm.com>
15424 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15425 optional tree argument.
15426 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15428 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15429 but continue to use the current value as a fallback.
15430 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15431 to compare the updates.
15432 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15433 (get_load_store_type): Use it when handling a strided access.
15434 (vect_get_strided_load_store_ops): New function.
15435 (vect_get_data_ptr_increment): Likewise.
15436 (vectorizable_load): Handle strided gather loads. Always pass
15437 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15440 Alan Hayward <alan.hayward@arm.com>
15441 David Sherwood <david.sherwood@arm.com>
15443 * doc/md.texi (gather_load@var{m}): Document.
15444 (mask_gather_load@var{m}): Likewise.
15445 * genopinit.c (main): Add supports_vec_gather_load and
15446 supports_vec_gather_load_cached to target_optabs.
15447 * optabs-tree.c (init_tree_optimization_optabs): Use
15448 ggc_cleared_alloc to allocate target_optabs.
15449 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15450 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15452 * internal-fn.h (internal_load_fn_p): Declare.
15453 (internal_gather_scatter_fn_p): Likewise.
15454 (internal_fn_mask_index): Likewise.
15455 (internal_gather_scatter_fn_supported_p): Likewise.
15456 * internal-fn.c (gather_load_direct): New macro.
15457 (expand_gather_load_optab_fn): New function.
15458 (direct_gather_load_optab_supported_p): New macro.
15459 (direct_internal_fn_optab): New function.
15460 (internal_load_fn_p): Likewise.
15461 (internal_gather_scatter_fn_p): Likewise.
15462 (internal_fn_mask_index): Likewise.
15463 (internal_gather_scatter_fn_supported_p): Likewise.
15464 * optabs-query.c (supports_at_least_one_mode_p): New function.
15465 (supports_vec_gather_load_p): Likewise.
15466 * optabs-query.h (supports_vec_gather_load_p): Declare.
15467 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15468 and memory_type field.
15469 (NUM_PATTERNS): Bump to 15.
15470 * tree-vect-data-refs.c: Include internal-fn.h.
15471 (vect_gather_scatter_fn_p): New function.
15472 (vect_describe_gather_scatter_call): Likewise.
15473 (vect_check_gather_scatter): Try using internal functions for
15474 gather loads. Recognize existing calls to a gather load function.
15475 (vect_analyze_data_refs): Consider using gather loads if
15476 supports_vec_gather_load_p.
15477 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15478 (vect_get_gather_scatter_offset_type): Likewise.
15479 (vect_convert_mask_for_vectype): Likewise.
15480 (vect_add_conversion_to_patterm): Likewise.
15481 (vect_try_gather_scatter_pattern): Likewise.
15482 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15483 (vect_vect_recog_func_ptrs): Add it.
15484 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15485 internal_fn_mask_index and internal_gather_scatter_fn_p.
15486 (check_load_store_masking): Take the gather_scatter_info as an
15487 argument and handle gather loads.
15488 (vect_get_gather_scatter_ops): New function.
15489 (vectorizable_call): Check internal_load_fn_p.
15490 (vectorizable_load): Likewise. Handle gather load internal
15492 (vectorizable_store): Update call to check_load_store_masking.
15493 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15494 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15495 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15496 (aarch64_gather_scale_operand_d): New predicates.
15497 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15498 (mask_gather_load<mode>): New insns.
15500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15501 Alan Hayward <alan.hayward@arm.com>
15502 David Sherwood <david.sherwood@arm.com>
15504 * optabs.def (fold_left_plus_optab): New optab.
15505 * doc/md.texi (fold_left_plus_@var{m}): Document.
15506 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15507 * internal-fn.c (fold_left_direct): Define.
15508 (expand_fold_left_optab_fn): Likewise.
15509 (direct_fold_left_optab_supported_p): Likewise.
15510 * fold-const-call.c (fold_const_fold_left): New function.
15511 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15512 * tree-parloops.c (valid_reduction_p): New function.
15513 (gather_scalar_reductions): Use it.
15514 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15515 (vect_finish_replace_stmt): Declare.
15516 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15517 (needs_fold_left_reduction_p): New function, split out from...
15518 (vect_is_simple_reduction): ...here. Accept reductions that
15519 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15520 (vect_force_simple_reduction): Also store the reduction type in
15521 the assignment's STMT_VINFO_REDUC_TYPE.
15522 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15523 (merge_with_identity): New function.
15524 (vect_expand_fold_left): Likewise.
15525 (vectorize_fold_left_reduction): Likewise.
15526 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15527 scalar phi in place for it. Check for target support and reject
15528 cases that would reassociate the operation. Defer the transform
15529 phase to vectorize_fold_left_reduction.
15530 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15531 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15532 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15536 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15537 call to ifc_temp_var.
15539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15540 Alan Hayward <alan.hayward@arm.com>
15541 David Sherwood <david.sherwood@arm.com>
15543 * target.def (legitimize_address_displacement): Take the original
15544 offset as a poly_int.
15545 * targhooks.h (default_legitimize_address_displacement): Update
15547 * targhooks.c (default_legitimize_address_displacement): Likewise.
15548 * doc/tm.texi: Regenerate.
15549 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15550 as an argument, moving assert of ad->disp == ad->disp_term to...
15551 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15552 Try calling targetm.legitimize_address_displacement before expanding
15553 the address rather than afterwards, and adjust for the new interface.
15554 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15555 Match the new hook interface. Handle SVE addresses.
15556 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15557 new hook interface.
15559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15561 * Makefile.in (OBJS): Add early-remat.o.
15562 * target.def (select_early_remat_modes): New hook.
15563 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15564 * doc/tm.texi: Regenerate.
15565 * targhooks.h (default_select_early_remat_modes): Declare.
15566 * targhooks.c (default_select_early_remat_modes): New function.
15567 * timevar.def (TV_EARLY_REMAT): New timevar.
15568 * passes.def (pass_early_remat): New pass.
15569 * tree-pass.h (make_pass_early_remat): Declare.
15570 * early-remat.c: New file.
15571 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15573 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15576 Alan Hayward <alan.hayward@arm.com>
15577 David Sherwood <david.sherwood@arm.com>
15579 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15580 vfm1 with a bound_epilog parameter.
15581 (vect_do_peeling): Update calls accordingly, and move the prologue
15582 call earlier in the function. Treat the base bound_epilog as 0 for
15583 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15584 this base when peeling for gaps.
15585 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15586 with fully-masked loops.
15587 (vect_estimate_min_profitable_iters): Handle the single peeled
15588 iteration in that case.
15590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15591 Alan Hayward <alan.hayward@arm.com>
15592 David Sherwood <david.sherwood@arm.com>
15594 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15595 single-element interleaving even if the size is not a power of 2.
15596 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15597 accesses for single-element interleaving if the group size is
15600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15601 Alan Hayward <alan.hayward@arm.com>
15602 David Sherwood <david.sherwood@arm.com>
15604 * doc/md.texi (fold_extract_last_@var{m}): Document.
15605 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15606 * optabs.def (fold_extract_last_optab): New optab.
15607 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15608 * internal-fn.c (fold_extract_direct): New macro.
15609 (expand_fold_extract_optab_fn): Likewise.
15610 (direct_fold_extract_optab_supported_p): Likewise.
15611 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15612 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15613 EXTRACT_LAST_REDUCTION.
15614 (get_initial_def_for_reduction): Do not create an initial vector
15615 for EXTRACT_LAST_REDUCTION reductions.
15616 (vectorizable_reduction): Leave the scalar phi in place for
15617 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15618 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15619 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15620 transform phase to vectorizable_condition.
15621 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15623 (vect_finish_stmt_generation): ...here.
15624 (vect_finish_replace_stmt): New function.
15625 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15626 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15628 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15631 Alan Hayward <alan.hayward@arm.com>
15632 David Sherwood <david.sherwood@arm.com>
15634 * doc/md.texi (extract_last_@var{m}): Document.
15635 * optabs.def (extract_last_optab): New optab.
15636 * internal-fn.def (EXTRACT_LAST): New internal function.
15637 * internal-fn.c (cond_unary_direct): New macro.
15638 (expand_cond_unary_optab_fn): Likewise.
15639 (direct_cond_unary_optab_supported_p): Likewise.
15640 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15641 loops using EXTRACT_LAST.
15642 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15643 (extract_last_<mode>): ...this optab.
15644 (vec_extract<mode><Vel>): Update accordingly.
15646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15647 Alan Hayward <alan.hayward@arm.com>
15648 David Sherwood <david.sherwood@arm.com>
15650 * target.def (empty_mask_is_expensive): New hook.
15651 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15652 * doc/tm.texi: Regenerate.
15653 * targhooks.h (default_empty_mask_is_expensive): Declare.
15654 * targhooks.c (default_empty_mask_is_expensive): New function.
15655 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15656 if the target says that empty masks are expensive.
15657 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15659 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15662 Alan Hayward <alan.hayward@arm.com>
15663 David Sherwood <david.sherwood@arm.com>
15665 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15666 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15667 (vect_use_loop_mask_for_alignment_p): New function.
15668 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15669 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15670 niters_skip argument. Make sure that the first niters_skip elements
15671 of the first iteration are inactive.
15672 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15673 Update call to vect_set_loop_masks_directly.
15674 (get_misalign_in_elems): New function, split out from...
15675 (vect_gen_prolog_loop_niters): ...here.
15676 (vect_update_init_of_dr): Take a code argument that specifies whether
15677 the adjustment should be added or subtracted.
15678 (vect_update_init_of_drs): Likewise.
15679 (vect_prepare_for_masked_peels): New function.
15680 (vect_do_peeling): Skip prologue peeling if we're using a mask
15681 instead. Update call to vect_update_inits_of_drs.
15682 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15684 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15685 alignment. Do not include the number of peeled iterations in
15686 the minimum threshold in that case.
15687 (vectorizable_induction): Adjust the start value down by
15688 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15689 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15690 Take the number of skipped iterations into account when calculating
15692 * tree-vect-stmts.c (vect_gen_while_not): New function.
15694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15695 Alan Hayward <alan.hayward@arm.com>
15696 David Sherwood <david.sherwood@arm.com>
15698 * doc/sourcebuild.texi (vect_fully_masked): Document.
15699 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15700 default value to 0.
15701 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15703 (vect_analyze_loop_2): ...here. Don't check the vectorization
15704 factor against the number of loop iterations if the loop is
15707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15708 Alan Hayward <alan.hayward@arm.com>
15709 David Sherwood <david.sherwood@arm.com>
15711 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15712 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15713 (dump_groups): Update accordingly.
15714 (iv_use::mem_type): New member variable.
15715 (address_p): New function.
15716 (record_use): Add a mem_type argument and initialize the new
15718 (record_group_use): Add a mem_type argument. Use address_p.
15719 Remove obsolete null checks of base_object. Update call to record_use.
15720 (find_interesting_uses_op): Update call to record_group_use.
15721 (find_interesting_uses_cond): Likewise.
15722 (find_interesting_uses_address): Likewise.
15723 (get_mem_type_for_internal_fn): New function.
15724 (find_address_like_use): Likewise.
15725 (find_interesting_uses_stmt): Try find_address_like_use before
15726 calling find_interesting_uses_op.
15727 (addr_offset_valid_p): Use the iv mem_type field as the type
15728 of the addressed memory.
15729 (add_autoinc_candidates): Likewise.
15730 (get_address_cost): Likewise.
15731 (split_small_address_groups_p): Use address_p.
15732 (split_address_groups): Likewise.
15733 (add_iv_candidate_for_use): Likewise.
15734 (autoinc_possible_for_pair): Likewise.
15735 (rewrite_groups): Likewise.
15736 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15737 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15738 (get_alias_ptr_type_for_ptr_address): New function.
15739 (rewrite_use_address): Rewrite address uses in calls that were
15740 identified by find_address_like_use.
15742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15743 Alan Hayward <alan.hayward@arm.com>
15744 David Sherwood <david.sherwood@arm.com>
15746 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15748 * gimple-expr.h (is_gimple_addressable: Likewise.
15749 * gimple-expr.c (is_gimple_address): Likewise.
15750 * internal-fn.c (expand_call_mem_ref): New function.
15751 (expand_mask_load_optab_fn): Use it.
15752 (expand_mask_store_optab_fn): Likewise.
15754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15755 Alan Hayward <alan.hayward@arm.com>
15756 David Sherwood <david.sherwood@arm.com>
15758 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15759 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15760 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15761 (cond_umax@var{mode}): Document.
15762 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15763 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15764 (cond_umin_optab, cond_umax_optab): New optabs.
15765 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15766 (COND_IOR, COND_XOR): New internal functions.
15767 * internal-fn.h (get_conditional_internal_fn): Declare.
15768 * internal-fn.c (cond_binary_direct): New macro.
15769 (expand_cond_binary_optab_fn): Likewise.
15770 (direct_cond_binary_optab_supported_p): Likewise.
15771 (get_conditional_internal_fn): New function.
15772 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15773 Cope with reduction statements that are vectorized as calls rather
15775 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15776 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15777 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15778 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15779 (UNSPEC_COND_EOR): New unspecs.
15780 (optab): Add mappings for them.
15781 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15782 (sve_int_op, sve_fp_op): New int attributes.
15784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15785 Alan Hayward <alan.hayward@arm.com>
15786 David Sherwood <david.sherwood@arm.com>
15788 * optabs.def (while_ult_optab): New optab.
15789 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15790 * internal-fn.def (WHILE_ULT): New internal function.
15791 * internal-fn.h (direct_internal_fn_supported_p): New override
15792 that takes two types as argument.
15793 * internal-fn.c (while_direct): New macro.
15794 (expand_while_optab_fn): New function.
15795 (convert_optab_supported_p): Likewise.
15796 (direct_while_optab_supported_p): New macro.
15797 * wide-int.h (wi::udiv_ceil): New function.
15798 * tree-vectorizer.h (rgroup_masks): New structure.
15799 (vec_loop_masks): New typedef.
15800 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15801 and fully_masked_p.
15802 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15803 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15804 (vect_max_vf): New function.
15805 (slpeel_make_loop_iterate_ntimes): Delete.
15806 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15807 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15808 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15809 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15810 internal-fn.h, stor-layout.h and optabs-query.h.
15811 (vect_set_loop_mask): New function.
15812 (add_preheader_seq): Likewise.
15813 (add_header_seq): Likewise.
15814 (interleave_supported_p): Likewise.
15815 (vect_maybe_permute_loop_masks): Likewise.
15816 (vect_set_loop_masks_directly): Likewise.
15817 (vect_set_loop_condition_masked): Likewise.
15818 (vect_set_loop_condition_unmasked): New function, split out from
15819 slpeel_make_loop_iterate_ntimes.
15820 (slpeel_make_loop_iterate_ntimes): Rename to..
15821 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15822 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15823 (vect_do_peeling): Update call accordingly.
15824 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15826 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15827 mask_compare_type, can_fully_mask_p and fully_masked_p.
15828 (release_vec_loop_masks): New function.
15829 (_loop_vec_info): Use it to free the loop masks.
15830 (can_produce_all_loop_masks_p): New function.
15831 (vect_get_max_nscalars_per_iter): Likewise.
15832 (vect_verify_full_masking): Likewise.
15833 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15834 retries, and free the mask rgroups before retrying. Check loop-wide
15835 reasons for disallowing fully-masked loops. Make the final decision
15836 about whether use a fully-masked loop or not.
15837 (vect_estimate_min_profitable_iters): Do not assume that peeling
15838 for the number of iterations will be needed for fully-masked loops.
15839 (vectorizable_reduction): Disable fully-masked loops.
15840 (vectorizable_live_operation): Likewise.
15841 (vect_halve_mask_nunits): New function.
15842 (vect_double_mask_nunits): Likewise.
15843 (vect_record_loop_mask): Likewise.
15844 (vect_get_loop_mask): Likewise.
15845 (vect_transform_loop): Handle the case in which the final loop
15846 iteration might handle a partial vector. Call vect_set_loop_condition
15847 instead of slpeel_make_loop_iterate_ntimes.
15848 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15849 (check_load_store_masking): New function.
15850 (prepare_load_store_mask): Likewise.
15851 (vectorizable_store): Handle fully-masked loops.
15852 (vectorizable_load): Likewise.
15853 (supportable_widening_operation): Use vect_halve_mask_nunits for
15855 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15856 (vect_gen_while): New function.
15857 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15858 (aarch64_uqdec<mode>): New insn.
15860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15861 Alan Hayward <alan.hayward@arm.com>
15862 David Sherwood <david.sherwood@arm.com>
15864 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15865 (reduc_xor_scal_optab): New optabs.
15866 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15867 (reduc_xor_scal_@var{m}): Document.
15868 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15869 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15870 internal functions.
15871 * fold-const-call.c (fold_const_call): Handle them.
15872 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15873 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15874 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15875 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15876 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15877 (UNSPEC_XORV): New unspecs.
15878 (optab): Add entries for them.
15879 (BITWISEV): New int iterator.
15880 (bit_reduc_op): New int attributes.
15882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15883 Alan Hayward <alan.hayward@arm.com>
15884 David Sherwood <david.sherwood@arm.com>
15886 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15887 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15888 * optabs.def (vec_shl_insert_optab): New optab.
15889 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15890 (duplicate_and_interleave): Likewise.
15891 * tree-vect-loop.c: Include internal-fn.h.
15892 (neutral_op_for_slp_reduction): New function, split out from
15893 get_initial_defs_for_reduction.
15894 (get_initial_def_for_reduction): Handle option 2 for variable-length
15895 vectors by loading the neutral value into a vector and then shifting
15896 the initial value into element 0.
15897 (get_initial_defs_for_reduction): Replace the code argument with
15898 the neutral value calculated by neutral_op_for_slp_reduction.
15899 Use gimple_build_vector for constant-length vectors.
15900 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15901 but the first group_size elements have a neutral value.
15902 Use duplicate_and_interleave otherwise.
15903 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15904 Update call to get_initial_defs_for_reduction. Handle SLP
15905 reductions for variable-length vectors by creating one vector
15906 result for each scalar result, with the elements associated
15907 with other scalar results stubbed out with the neutral value.
15908 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15909 Require IFN_VEC_SHL_INSERT for double reductions on
15910 variable-length vectors, or SLP reductions that have
15911 a neutral value. Require can_duplicate_and_interleave_p
15912 support for variable-length unchained SLP reductions if there
15913 is no neutral value, such as for MIN/MAX reductions. Also require
15914 the number of vector elements to be a multiple of the number of
15915 SLP statements when doing variable-length unchained SLP reductions.
15916 Update call to vect_create_epilog_for_reduction.
15917 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15918 and remove initial values.
15919 (duplicate_and_interleave): Make public.
15920 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15921 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15924 Alan Hayward <alan.hayward@arm.com>
15925 David Sherwood <david.sherwood@arm.com>
15927 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15928 (can_duplicate_and_interleave_p): New function.
15929 (vect_get_and_check_slp_defs): Take the vector of statements
15930 rather than just the current one. Remove excess parentheses.
15931 Restriction rejectinon of vect_constant_def and vect_external_def
15932 for variable-length vectors to boolean types, or types for which
15933 can_duplicate_and_interleave_p is false.
15934 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15935 (duplicate_and_interleave): New function.
15936 (vect_get_constant_vectors): Use gimple_build_vector for
15937 constant-length vectors and suitable variable-length constant
15938 vectors. Use duplicate_and_interleave for other variable-length
15939 vectors. Don't defer the update when inserting new statements.
15941 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15942 Alan Hayward <alan.hayward@arm.com>
15943 David Sherwood <david.sherwood@arm.com>
15945 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15946 min_profitable_iters doesn't go negative.
15948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15949 Alan Hayward <alan.hayward@arm.com>
15950 David Sherwood <david.sherwood@arm.com>
15952 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15953 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15954 * optabs.def (vec_mask_load_lanes_optab): New optab.
15955 (vec_mask_store_lanes_optab): Likewise.
15956 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15957 (MASK_STORE_LANES): Likewise.
15958 * internal-fn.c (mask_load_lanes_direct): New macro.
15959 (mask_store_lanes_direct): Likewise.
15960 (expand_mask_load_optab_fn): Handle masked operations.
15961 (expand_mask_load_lanes_optab_fn): New macro.
15962 (expand_mask_store_optab_fn): Handle masked operations.
15963 (expand_mask_store_lanes_optab_fn): New macro.
15964 (direct_mask_load_lanes_optab_supported_p): Likewise.
15965 (direct_mask_store_lanes_optab_supported_p): Likewise.
15966 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15968 (vect_load_lanes_supported): Likewise.
15969 * tree-vect-data-refs.c (strip_conversion): New function.
15970 (can_group_stmts_p): Likewise.
15971 (vect_analyze_data_ref_accesses): Use it instead of checking
15972 for a pair of assignments.
15973 (vect_store_lanes_supported): Take a masked_p parameter.
15974 (vect_load_lanes_supported): Likewise.
15975 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15976 vect_store_lanes_supported and vect_load_lanes_supported.
15977 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15978 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15979 parameter. Don't allow gaps for masked accesses.
15980 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15981 and vect_load_lanes_supported.
15982 (get_load_store_type): Take a masked_p parameter and update
15983 call to get_group_load_store_type.
15984 (vectorizable_store): Update call to get_load_store_type.
15985 Handle IFN_MASK_STORE_LANES.
15986 (vectorizable_load): Update call to get_load_store_type.
15987 Handle IFN_MASK_LOAD_LANES.
15989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15990 Alan Hayward <alan.hayward@arm.com>
15991 David Sherwood <david.sherwood@arm.com>
15993 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15995 * config/aarch64/aarch64-protos.h
15996 (aarch64_sve_struct_memory_operand_p): Declare.
15997 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15998 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15999 (VPRED, vpred): Handle SVE structure modes.
16000 * config/aarch64/constraints.md (Utx): New constraint.
16001 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16002 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16003 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16004 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16005 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16006 structure modes. Split into pieces after RA.
16007 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16008 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16010 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16011 SVE structure modes.
16012 (aarch64_classify_address): Likewise.
16013 (sizetochar): Move earlier in file.
16014 (aarch64_print_operand): Handle SVE register lists.
16015 (aarch64_array_mode): New function.
16016 (aarch64_sve_struct_memory_operand_p): Likewise.
16017 (TARGET_ARRAY_MODE): Redefine.
16019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16020 Alan Hayward <alan.hayward@arm.com>
16021 David Sherwood <david.sherwood@arm.com>
16023 * target.def (array_mode): New target hook.
16024 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16025 * doc/tm.texi: Regenerate.
16026 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16027 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16028 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16029 targetm.array_mode.
16030 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16033 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16034 Alan Hayward <alan.hayward@arm.com>
16035 David Sherwood <david.sherwood@arm.com>
16037 * fold-const.c (fold_binary_loc): Check the argument types
16038 rather than the result type when testing for a vector operation.
16040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16042 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16043 * doc/tm.texi: Regenerate.
16045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16046 Alan Hayward <alan.hayward@arm.com>
16047 David Sherwood <david.sherwood@arm.com>
16049 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16050 (sve): Document new AArch64 extension.
16051 * doc/md.texi (w): Extend the description of the AArch64
16052 constraint to include SVE vectors.
16053 (Upl, Upa): Document new AArch64 predicate constraints.
16054 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16056 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16057 (msve-vector-bits=): New option.
16058 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16059 SVE when these are disabled.
16060 (sve): New extension.
16061 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16062 modes. Adjust their number of units based on aarch64_sve_vg.
16063 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16064 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16065 aarch64_addr_query_type.
16066 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16067 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16068 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16069 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16070 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16071 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16072 (aarch64_simd_imm_zero_p): Delete.
16073 (aarch64_check_zero_based_sve_index_immediate): Declare.
16074 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16075 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16076 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16077 (aarch64_sve_float_mul_immediate_p): Likewise.
16078 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16079 rather than an rtx.
16080 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16081 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16082 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16083 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16084 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16085 (aarch64_regmode_natural_size): Likewise.
16086 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16087 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16089 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16090 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16091 for VG and the SVE predicate registers.
16092 (V_ALIASES): Add a "z"-prefixed alias.
16093 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16094 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16095 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16096 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16097 (REG_CLASS_NAMES): Add entries for them.
16098 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16099 and the predicate registers.
16100 (aarch64_sve_vg): Declare.
16101 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16102 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16103 (REGMODE_NATURAL_SIZE): Define.
16104 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16106 * config/aarch64/aarch64.c: Include cfgrtl.h.
16107 (simd_immediate_info): Add a constructor for series vectors,
16108 and an associated step field.
16109 (aarch64_sve_vg): New variable.
16110 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16111 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16112 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16113 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16114 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16115 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16116 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16117 (aarch64_get_mask_mode): New functions.
16118 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16119 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16120 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16121 predicate modes and predicate registers. Explicitly restrict
16122 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16123 to store a vector mode if it is recognized by
16124 aarch64_classify_vector_mode.
16125 (aarch64_regmode_natural_size): New function.
16126 (aarch64_hard_regno_caller_save_mode): Return the original mode
16128 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16129 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16130 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16131 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16133 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16134 does not overlap dest if the function is frame-related. Handle
16136 (aarch64_split_add_offset): New function.
16137 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16138 them aarch64_add_offset.
16139 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16140 and update call to aarch64_sub_sp.
16141 (aarch64_add_cfa_expression): New function.
16142 (aarch64_expand_prologue): Pass extra temporary registers to the
16143 functions above. Handle the case in which we need to emit new
16144 DW_CFA_expressions for registers that were originally saved
16145 relative to the stack pointer, but now have to be expressed
16146 relative to the frame pointer.
16147 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16149 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16150 IP0 and IP1 values for SVE frames.
16151 (aarch64_expand_vec_series): New function.
16152 (aarch64_expand_sve_widened_duplicate): Likewise.
16153 (aarch64_expand_sve_const_vector): Likewise.
16154 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16155 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16156 into the register, rather than emitting a SET directly.
16157 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16158 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16159 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16160 (offset_9bit_signed_scaled_p): New functions.
16161 (aarch64_replicate_bitmask_imm): New function.
16162 (aarch64_bitmask_imm): Use it.
16163 (aarch64_cannot_force_const_mem): Reject expressions involving
16164 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16165 (aarch64_classify_index): Handle SVE indices, by requiring
16166 a plain register index with a scale that matches the element size.
16167 (aarch64_classify_address): Handle SVE addresses. Assert that
16168 the mode of the address is VOIDmode or an integer mode.
16169 Update call to aarch64_classify_symbol.
16170 (aarch64_classify_symbolic_expression): Update call to
16171 aarch64_classify_symbol.
16172 (aarch64_const_vec_all_in_range_p): New function.
16173 (aarch64_print_vector_float_operand): Likewise.
16174 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16175 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16176 and the FP immediates 1.0 and 0.5.
16177 (aarch64_print_address_internal): Handle SVE addresses.
16178 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16179 (aarch64_regno_regclass): Handle predicate registers.
16180 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16182 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16183 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16184 (aarch64_convert_sve_vector_bits): New function.
16185 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16186 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16187 rather than an rtx.
16188 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16189 Handle SVE vector and predicate modes. Accept VL-based constants
16190 that need only one temporary register, and VL offsets that require
16191 no temporary registers.
16192 (aarch64_conditional_register_usage): Mark the predicate registers
16193 as fixed if SVE isn't available.
16194 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16195 Return true for SVE vector and predicate modes.
16196 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16197 rather than an unsigned int. Handle SVE modes.
16198 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16200 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16202 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16203 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16204 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16205 (aarch64_sve_float_mul_immediate_p): New functions.
16206 (aarch64_sve_valid_immediate): New function.
16207 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16208 Explicitly reject structure modes. Check for INDEX constants.
16209 Handle PTRUE and PFALSE constants.
16210 (aarch64_check_zero_based_sve_index_immediate): New function.
16211 (aarch64_simd_imm_zero_p): Delete.
16212 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16213 vector modes. Accept constants in the range of CNT[BHWD].
16214 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16215 ask for an Advanced SIMD mode.
16216 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16217 (aarch64_simd_vector_alignment): Handle SVE predicates.
16218 (aarch64_vectorize_preferred_vector_alignment): New function.
16219 (aarch64_simd_vector_alignment_reachable): Use it instead of
16221 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16222 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16224 (MAX_VECT_LEN): Delete.
16225 (expand_vec_perm_d): Add a vec_flags field.
16226 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16227 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16228 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16230 (aarch64_evpc_rev): Rename to...
16231 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16232 (aarch64_evpc_rev_global): New function.
16233 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16234 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16236 (aarch64_evpc_sve_tbl): New function.
16237 (aarch64_expand_vec_perm_const_1): Update after rename of
16238 aarch64_evpc_rev. Handle SVE permutes too, trying
16239 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16240 than aarch64_evpc_tbl.
16241 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16242 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16243 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16244 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16245 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16246 (aarch64_expand_sve_vcond): New functions.
16247 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16248 of aarch64_vector_mode_p.
16249 (aarch64_dwarf_poly_indeterminate_value): New function.
16250 (aarch64_compute_pressure_classes): Likewise.
16251 (aarch64_can_change_mode_class): Likewise.
16252 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16253 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16254 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16255 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16256 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16257 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16258 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16259 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16261 (Dn, Dl, Dr): Accept const as well as const_vector.
16262 (Dz): Likewise. Compare against CONST0_RTX.
16263 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16264 of "vector" where appropriate.
16265 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16266 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16267 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16268 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16269 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16270 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16271 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16272 (v_int_equiv): Extend to SVE modes.
16273 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16275 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16276 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16277 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16278 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16279 (SVE_COND_FP_CMP): New int iterators.
16280 (perm_hilo): Handle the new unpack unspecs.
16281 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16283 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16284 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16285 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16286 (aarch64_equality_operator, aarch64_constant_vector_operand)
16287 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16288 (aarch64_sve_nonimmediate_operand): Likewise.
16289 (aarch64_sve_general_operand): Likewise.
16290 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16291 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16292 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16293 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16294 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16295 (aarch64_sve_float_arith_immediate): Likewise.
16296 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16297 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16298 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16299 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16300 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16301 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16302 (aarch64_sve_float_arith_operand): Likewise.
16303 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16304 (aarch64_sve_float_mul_operand): Likewise.
16305 (aarch64_sve_vec_perm_operand): Likewise.
16306 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16307 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16308 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16309 as well as const_vector.
16310 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16311 in file. Use CONST0_RTX and CONSTM1_RTX.
16312 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16313 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16314 Use aarch64_simd_imm_zero.
16315 * config/aarch64/aarch64-sve.md: New file.
16316 * config/aarch64/aarch64.md: Include it.
16317 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16318 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16319 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16320 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16321 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16322 (sve): New attribute.
16323 (enabled): Disable instructions with the sve attribute unless
16325 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16326 aarch64_expand_mov_immediate.
16327 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16328 CNT[BHSD] immediates.
16329 (movti): Split CONST_POLY_INT moves into two halves.
16330 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16331 Split additions that need a temporary here if the destination
16332 is the stack pointer.
16333 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16334 (*add<mode>3_poly_1): New instruction.
16335 (set_clobber_cc): New expander.
16337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16339 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16340 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16341 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16342 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16343 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16344 Change innermode from fixed_mode_size to machine_mode.
16345 (simplify_subreg): Update call accordingly. Handle a constant-sized
16346 subreg of a variable-length CONST_VECTOR.
16348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16349 Alan Hayward <alan.hayward@arm.com>
16350 David Sherwood <david.sherwood@arm.com>
16352 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16353 (add_offset_to_base): New function, split out from...
16354 (create_mem_ref): ...here. When handling a scale other than 1,
16355 check first whether the address is valid without the offset.
16356 Add it into the base if so, leaving the index and scale as-is.
16358 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16361 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16362 fold_for_warn before checking if arg2 is INTEGER_CST.
16364 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16366 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16367 (store_multiple_operation): Delete.
16368 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16369 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16370 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16371 guarded by TARGET_STRING.
16372 (rs6000_output_load_multiple): Delete.
16373 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16374 OPTION_MASK_STRING / TARGET_STRING handling.
16375 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16376 (const rs6000_opt_masks) <"string">: Change mask to 0.
16377 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16378 (MASK_STRING): Delete.
16379 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16381 (load_multiple): Delete.
16388 (store_multiple): Delete.
16395 (movmemsi_8reg): Delete.
16396 (corresponding unnamed define_insn): Delete.
16397 (movmemsi_6reg): Delete.
16398 (corresponding unnamed define_insn): Delete.
16399 (movmemsi_4reg): Delete.
16400 (corresponding unnamed define_insn): Delete.
16401 (movmemsi_2reg): Delete.
16402 (corresponding unnamed define_insn): Delete.
16403 (movmemsi_1reg): Delete.
16404 (corresponding unnamed define_insn): Delete.
16405 * config/rs6000/rs6000.opt (mno-string): New.
16406 (mstring): Replace by deprecation warning stub.
16407 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16409 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16411 * regrename.c (regrename_do_replace): If replacing the same
16412 reg multiple times, try to reuse last created gen_raw_REG.
16415 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16416 main to workaround a bug in GDB.
16418 2018-01-12 Tom de Vries <tom@codesourcery.com>
16421 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16423 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16425 PR rtl-optimization/80481
16426 * ira-color.c (get_cap_member): New function.
16427 (allocnos_conflict_by_live_ranges_p): Use it.
16428 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16429 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16431 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16434 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16435 (*saddl_se_1): Ditto.
16436 (*ssubsi_1): Ditto.
16437 (*ssubl_se_1): Ditto.
16439 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16441 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16442 rather than wi::to_widest for DR_INITs.
16443 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16444 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16445 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16447 (vect_analyze_group_access_1): Note that here.
16449 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16451 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16452 polynomial type sizes.
16454 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16456 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16457 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16458 (gimple_add_tmp_var): Likewise.
16460 2018-01-12 Martin Liska <mliska@suse.cz>
16462 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16463 (gimple_alloc_sizes): Likewise.
16464 (dump_gimple_statistics): Use PRIu64 in printf format.
16465 * gimple.h: Change uint64_t to int.
16467 2018-01-12 Martin Liska <mliska@suse.cz>
16469 * tree-core.h: Use uint64_t instead of int.
16470 * tree.c (tree_node_counts): Likewise.
16471 (tree_node_sizes): Likewise.
16472 (dump_tree_statistics): Use PRIu64 in printf format.
16474 2018-01-12 Martin Liska <mliska@suse.cz>
16476 * Makefile.in: As qsort_chk is implemented in vec.c, add
16477 vec.o to linkage of gencfn-macros.
16478 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16479 passing the info to record_node_allocation_statistics.
16480 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16482 * ggc-common.c (struct ggc_usage): Add operator== and use
16483 it in operator< and compare function.
16484 * mem-stats.h (struct mem_usage): Likewise.
16485 * vec.c (struct vec_usage): Remove operator< and compare
16486 function. Can be simply inherited.
16488 2018-01-12 Martin Jambor <mjambor@suse.cz>
16491 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16492 * tree-ssa-math-opts.c: Include domwalk.h.
16493 (convert_mult_to_fma_1): New function.
16494 (fma_transformation_info): New type.
16495 (fma_deferring_state): Likewise.
16496 (cancel_fma_deferring): New function.
16497 (result_of_phi): Likewise.
16498 (last_fma_candidate_feeds_initial_phi): Likewise.
16499 (convert_mult_to_fma): Added deferring logic, split actual
16500 transformation to convert_mult_to_fma_1.
16501 (math_opts_dom_walker): New type.
16502 (math_opts_dom_walker::after_dom_children): New method, body moved
16503 here from pass_optimize_widening_mul::execute, added deferring logic
16505 (pass_optimize_widening_mul::execute): Moved most of code to
16506 math_opts_dom_walker::after_dom_children.
16507 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16508 * config/i386/i386.c (ix86_option_override_internal): Added
16509 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16511 2018-01-12 Richard Biener <rguenther@suse.de>
16514 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16515 inline instance vars.
16517 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16520 * config/rx/rx.c (rx_is_restricted_memory_address):
16521 Handle SUBREG case.
16523 2018-01-12 Richard Biener <rguenther@suse.de>
16525 PR tree-optimization/80846
16526 * target.def (split_reduction): New target hook.
16527 * targhooks.c (default_split_reduction): New function.
16528 * targhooks.h (default_split_reduction): Declare.
16529 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16530 target requests first reduce vectors by combining low and high
16532 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16533 (get_vectype_for_scalar_type_and_size): Export.
16534 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16535 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16536 * doc/tm.texi: Regenerate.
16537 * config/i386/i386.c (ix86_split_reduction): Implement
16538 TARGET_VECTORIZE_SPLIT_REDUCTION.
16540 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16543 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16544 in PIC mode except for TARGET_VXWORKS_RTP.
16545 * config/sparc/sparc.c: Include cfgrtl.h.
16546 (TARGET_INIT_PIC_REG): Define.
16547 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16548 (sparc_pic_register_p): New predicate.
16549 (sparc_legitimate_address_p): Use it.
16550 (sparc_legitimize_pic_address): Likewise.
16551 (sparc_delegitimize_address): Likewise.
16552 (sparc_mode_dependent_address_p): Likewise.
16553 (gen_load_pcrel_sym): Remove 4th parameter.
16554 (load_got_register): Adjust call to above. Remove obsolete stuff.
16555 (sparc_expand_prologue): Do not call load_got_register here.
16556 (sparc_flat_expand_prologue): Likewise.
16557 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16558 (sparc_use_pseudo_pic_reg): New function.
16559 (sparc_init_pic_reg): Likewise.
16560 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16561 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16563 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16565 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16566 Add item for branch_cost.
16568 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16570 PR rtl-optimization/83565
16571 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16572 not extend the result to a larger mode for rotate operations.
16573 (num_sign_bit_copies1): Likewise.
16575 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16578 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16580 Use values-Xc.o for -pedantic.
16581 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16583 2018-01-12 Martin Liska <mliska@suse.cz>
16586 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16588 (possible_polymorphic_call_targets): Use it.
16589 (ipa_devirt): Likewise.
16591 2018-01-12 Martin Liska <mliska@suse.cz>
16593 * profile-count.h (enum profile_quality): Use 0 as invalid
16594 enum value of profile_quality.
16596 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16598 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16599 -mext-string options.
16601 2018-01-12 Richard Biener <rguenther@suse.de>
16603 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16604 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16605 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16607 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16609 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16611 * configure.ac (--with-long-double-format): Add support for the
16612 configuration option to change the default long double format on
16614 * config.gcc (powerpc*-linux*-*): Likewise.
16615 * configure: Regenerate.
16616 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16617 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16618 used without modification.
16620 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16622 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16623 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16624 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16625 MISC_BUILTIN_SPEC_BARRIER.
16626 (rs6000_init_builtins): Likewise.
16627 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16629 (speculation_barrier): New define_insn.
16630 * doc/extend.texi: Document __builtin_speculation_barrier.
16632 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16635 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16636 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16637 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16639 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16640 integral modes instead of "ss" and "sd".
16641 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16642 vectors with 32-bit and 64-bit elements.
16643 (vecdupssescalarmodesuffix): New mode attribute.
16644 (vec_dup<mode>): Use it.
16646 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16649 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16650 frame if argument is passed on stack.
16652 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16655 * ree.c (combine_reaching_defs): Optimize also
16656 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16657 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16659 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16661 PR middle-end/83189
16662 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16664 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16666 PR middle-end/83718
16667 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16668 after they are computed.
16670 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16672 PR tree-optimization/83695
16673 * gimple-loop-linterchange.cc
16674 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16675 reset cached scev information after interchange.
16676 (pass_linterchange::execute): Remove call to scev_reset_htab.
16678 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16680 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16681 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16682 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16683 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16684 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16685 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16686 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16687 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16688 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16689 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16690 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16691 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16692 (V_lane_reg): Likewise.
16693 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16695 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16696 (vfmal_lane_low<mode>_intrinsic,
16697 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16698 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16699 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16700 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16701 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16702 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16704 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16706 * config/arm/arm-cpus.in (fp16fml): New feature.
16707 (ALL_SIMD): Add fp16fml.
16708 (armv8.2-a): Add fp16fml as an option.
16709 (armv8.3-a): Likewise.
16710 (armv8.4-a): Add fp16fml as part of fp16.
16711 * config/arm/arm.h (TARGET_FP16FML): Define.
16712 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16714 * config/arm/arm-modes.def (V2HF): Define.
16715 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16716 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16717 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16718 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16719 vfmsl_low, vfmsl_high): New set of builtins.
16720 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16721 (vfml_op): New code attribute.
16722 (VFMLHALVES): New int iterator.
16723 (VFML, VFMLSEL): New mode attributes.
16724 (V_reg): Define mapping for V2HF.
16725 (V_hi, V_lo): New mode attributes.
16726 (VF_constraint): Likewise.
16727 (vfml_half, vfml_half_selector): New int attributes.
16728 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16730 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16731 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16733 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16734 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16735 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16736 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16738 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16739 Document new effective target and option set.
16741 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16743 * config/arm/arm-cpus.in (armv8_4): New feature.
16744 (ARMv8_4a): New fgroup.
16745 (armv8.4-a): New arch.
16746 * config/arm/arm-tables.opt: Regenerate.
16747 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16748 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16749 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16750 Add matching rules for -march=armv8.4-a and extensions.
16751 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16753 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16756 * config/rx/rx.md (BW): New mode attribute.
16757 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16759 2018-01-11 Richard Biener <rguenther@suse.de>
16761 PR tree-optimization/83435
16762 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16763 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16764 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16766 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16767 Alan Hayward <alan.hayward@arm.com>
16768 David Sherwood <david.sherwood@arm.com>
16770 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16772 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16773 (aarch64_print_address_internal): Use it to check for a zero offset.
16775 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16776 Alan Hayward <alan.hayward@arm.com>
16777 David Sherwood <david.sherwood@arm.com>
16779 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16780 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16781 Return a poly_int64 rather than a HOST_WIDE_INT.
16782 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16783 rather than a HOST_WIDE_INT.
16784 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16785 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16786 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16787 final_offset from HOST_WIDE_INT to poly_int64.
16788 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16789 to_constant when getting the number of units in an Advanced SIMD
16791 (aarch64_builtin_vectorized_function): Check for a constant number
16793 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16795 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16796 attribute instead of GET_MODE_NUNITS.
16797 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16798 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16799 GET_MODE_SIZE for fixed-size registers.
16800 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16801 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16802 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16803 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16804 (aarch64_print_operand, aarch64_print_address_internal)
16805 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16806 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16807 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16808 Handle polynomial GET_MODE_SIZE.
16809 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16810 wider than SImode without modification.
16811 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16812 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16813 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16814 passing and returning SVE modes.
16815 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16816 rather than GEN_INT.
16817 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16818 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16819 (aarch64_allocate_and_probe_stack_space): Likewise.
16820 (aarch64_layout_frame): Cope with polynomial offsets.
16821 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16822 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16823 polynomial offsets.
16824 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16825 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16826 poly_int64 rather than a HOST_WIDE_INT.
16827 (aarch64_get_separate_components, aarch64_process_components)
16828 (aarch64_expand_prologue, aarch64_expand_epilogue)
16829 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16830 (aarch64_anchor_offset): New function, split out from...
16831 (aarch64_legitimize_address): ...here.
16832 (aarch64_builtin_vectorization_cost): Handle polynomial
16833 TYPE_VECTOR_SUBPARTS.
16834 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16836 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16837 number of elements from the PARALLEL rather than the mode.
16838 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16839 rather than GET_MODE_BITSIZE.
16840 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16841 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16842 (aarch64_expand_vec_perm_const_1): Handle polynomial
16843 d->perm.length () and d->perm elements.
16844 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16845 Apply to_constant to d->perm elements.
16846 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16847 polynomial CONST_VECTOR_NUNITS.
16848 (aarch64_move_pointer): Take amount as a poly_int64 rather
16850 (aarch64_progress_pointer): Avoid temporary variable.
16851 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16852 the mode attribute instead of GET_MODE.
16854 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16855 Alan Hayward <alan.hayward@arm.com>
16856 David Sherwood <david.sherwood@arm.com>
16858 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16859 x exists before using it.
16860 (aarch64_add_constant_internal): Rename to...
16861 (aarch64_add_offset_1): ...this. Replace regnum with separate
16862 src and dest rtxes. Handle the case in which they're different,
16863 including when the offset is zero. Replace scratchreg with an rtx.
16864 Use 2 additions if there is no spare register into which we can
16865 move a 16-bit constant.
16866 (aarch64_add_constant): Delete.
16867 (aarch64_add_offset): Replace reg with separate src and dest
16868 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16869 Use aarch64_add_offset_1.
16870 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16871 an rtx rather than an int. Take the delta as a poly_int64
16872 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16873 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16874 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16875 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16876 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16877 and aarch64_add_sp.
16878 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16879 aarch64_add_constant.
16881 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16883 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16884 Use scalar_float_mode.
16886 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16888 * config/aarch64/aarch64-simd.md
16889 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16890 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16891 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16892 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16893 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16894 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16895 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16896 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16897 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16898 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16900 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16903 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16904 targ_options->x_arm_arch_string is non NULL.
16906 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16908 * config/aarch64/aarch64.h
16909 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16911 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16914 * expmed.c (emit_store_flag_force): Swap if const op0
16915 and change VOIDmode to mode of op0.
16917 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16919 PR rtl-optimization/83761
16920 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16921 than bytes to mode_for_size.
16923 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16925 PR middle-end/83189
16926 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16927 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16930 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16932 PR middle-end/83575
16933 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16934 when in layout mode.
16935 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16936 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16939 2018-01-10 Michael Collison <michael.collison@arm.com>
16941 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16942 * config/aarch64/aarch64-option-extension.def: Add
16943 AARCH64_OPT_EXTENSION of 'fp16fml'.
16944 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16945 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16946 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16947 * config/aarch64/constraints.md (Ui7): New constraint.
16948 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16949 (VFMLA_SEL_W): Ditto.
16952 (VFMLA16_LOW): New int iterator.
16953 (VFMLA16_HIGH): Ditto.
16954 (UNSPEC_FMLAL): New unspec.
16955 (UNSPEC_FMLSL): Ditto.
16956 (UNSPEC_FMLAL2): Ditto.
16957 (UNSPEC_FMLSL2): Ditto.
16958 (f16mac): New code attribute.
16959 * config/aarch64/aarch64-simd-builtins.def
16960 (aarch64_fmlal_lowv2sf): Ditto.
16961 (aarch64_fmlsl_lowv2sf): Ditto.
16962 (aarch64_fmlalq_lowv4sf): Ditto.
16963 (aarch64_fmlslq_lowv4sf): Ditto.
16964 (aarch64_fmlal_highv2sf): Ditto.
16965 (aarch64_fmlsl_highv2sf): Ditto.
16966 (aarch64_fmlalq_highv4sf): Ditto.
16967 (aarch64_fmlslq_highv4sf): Ditto.
16968 (aarch64_fmlal_lane_lowv2sf): Ditto.
16969 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16970 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16971 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16972 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16973 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16974 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16975 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16976 (aarch64_fmlal_lane_highv2sf): Ditto.
16977 (aarch64_fmlsl_lane_highv2sf): Ditto.
16978 (aarch64_fmlal_laneq_highv2sf): Ditto.
16979 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16980 (aarch64_fmlalq_lane_highv4sf): Ditto.
16981 (aarch64_fmlsl_lane_highv4sf): Ditto.
16982 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16983 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16984 * config/aarch64/aarch64-simd.md:
16985 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16986 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16987 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16988 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16989 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16990 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16991 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16992 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16993 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16994 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16995 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16996 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16997 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16998 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16999 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17000 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17001 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17002 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17003 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17004 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17005 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17006 (vfmlsl_low_u32): Ditto.
17007 (vfmlalq_low_u32): Ditto.
17008 (vfmlslq_low_u32): Ditto.
17009 (vfmlal_high_u32): Ditto.
17010 (vfmlsl_high_u32): Ditto.
17011 (vfmlalq_high_u32): Ditto.
17012 (vfmlslq_high_u32): Ditto.
17013 (vfmlal_lane_low_u32): Ditto.
17014 (vfmlsl_lane_low_u32): Ditto.
17015 (vfmlal_laneq_low_u32): Ditto.
17016 (vfmlsl_laneq_low_u32): Ditto.
17017 (vfmlalq_lane_low_u32): Ditto.
17018 (vfmlslq_lane_low_u32): Ditto.
17019 (vfmlalq_laneq_low_u32): Ditto.
17020 (vfmlslq_laneq_low_u32): Ditto.
17021 (vfmlal_lane_high_u32): Ditto.
17022 (vfmlsl_lane_high_u32): Ditto.
17023 (vfmlal_laneq_high_u32): Ditto.
17024 (vfmlsl_laneq_high_u32): Ditto.
17025 (vfmlalq_lane_high_u32): Ditto.
17026 (vfmlslq_lane_high_u32): Ditto.
17027 (vfmlalq_laneq_high_u32): Ditto.
17028 (vfmlslq_laneq_high_u32): Ditto.
17029 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17030 (AARCH64_FL_FOR_ARCH8_4): New.
17031 (AARCH64_ISA_F16FML): New ISA flag.
17032 (TARGET_F16FML): New feature flag for fp16fml.
17033 (doc/invoke.texi): Document new fp16fml option.
17035 2018-01-10 Michael Collison <michael.collison@arm.com>
17037 * config/aarch64/aarch64-builtins.c:
17038 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17039 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17040 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17041 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17042 (AARCH64_ISA_SHA3): New ISA flag.
17043 (TARGET_SHA3): New feature flag for sha3.
17044 * config/aarch64/iterators.md (sha512_op): New int attribute.
17045 (CRYPTO_SHA512): New int iterator.
17046 (UNSPEC_SHA512H): New unspec.
17047 (UNSPEC_SHA512H2): Ditto.
17048 (UNSPEC_SHA512SU0): Ditto.
17049 (UNSPEC_SHA512SU1): Ditto.
17050 * config/aarch64/aarch64-simd-builtins.def
17051 (aarch64_crypto_sha512hqv2di): New builtin.
17052 (aarch64_crypto_sha512h2qv2di): Ditto.
17053 (aarch64_crypto_sha512su0qv2di): Ditto.
17054 (aarch64_crypto_sha512su1qv2di): Ditto.
17055 (aarch64_eor3qv8hi): Ditto.
17056 (aarch64_rax1qv2di): Ditto.
17057 (aarch64_xarqv2di): Ditto.
17058 (aarch64_bcaxqv8hi): Ditto.
17059 * config/aarch64/aarch64-simd.md:
17060 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17061 (aarch64_crypto_sha512su0qv2di): Ditto.
17062 (aarch64_crypto_sha512su1qv2di): Ditto.
17063 (aarch64_eor3qv8hi): Ditto.
17064 (aarch64_rax1qv2di): Ditto.
17065 (aarch64_xarqv2di): Ditto.
17066 (aarch64_bcaxqv8hi): Ditto.
17067 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17068 (vsha512h2q_u64): Ditto.
17069 (vsha512su0q_u64): Ditto.
17070 (vsha512su1q_u64): Ditto.
17071 (veor3q_u16): Ditto.
17072 (vrax1q_u64): Ditto.
17073 (vxarq_u64): Ditto.
17074 (vbcaxq_u16): Ditto.
17075 * config/arm/types.md (crypto_sha512): New type attribute.
17076 (crypto_sha3): Ditto.
17077 (doc/invoke.texi): Document new sha3 option.
17079 2018-01-10 Michael Collison <michael.collison@arm.com>
17081 * config/aarch64/aarch64-builtins.c:
17082 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17083 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17084 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17085 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17086 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17087 (AARCH64_ISA_SM4): New ISA flag.
17088 (TARGET_SM4): New feature flag for sm4.
17089 * config/aarch64/aarch64-simd-builtins.def
17090 (aarch64_sm3ss1qv4si): Ditto.
17091 (aarch64_sm3tt1aq4si): Ditto.
17092 (aarch64_sm3tt1bq4si): Ditto.
17093 (aarch64_sm3tt2aq4si): Ditto.
17094 (aarch64_sm3tt2bq4si): Ditto.
17095 (aarch64_sm3partw1qv4si): Ditto.
17096 (aarch64_sm3partw2qv4si): Ditto.
17097 (aarch64_sm4eqv4si): Ditto.
17098 (aarch64_sm4ekeyqv4si): Ditto.
17099 * config/aarch64/aarch64-simd.md:
17100 (aarch64_sm3ss1qv4si): Ditto.
17101 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17102 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17103 (aarch64_sm4eqv4si): Ditto.
17104 (aarch64_sm4ekeyqv4si): Ditto.
17105 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17106 (sm3part_op): Ditto.
17107 (CRYPTO_SM3TT): Ditto.
17108 (CRYPTO_SM3PART): Ditto.
17109 (UNSPEC_SM3SS1): New unspec.
17110 (UNSPEC_SM3TT1A): Ditto.
17111 (UNSPEC_SM3TT1B): Ditto.
17112 (UNSPEC_SM3TT2A): Ditto.
17113 (UNSPEC_SM3TT2B): Ditto.
17114 (UNSPEC_SM3PARTW1): Ditto.
17115 (UNSPEC_SM3PARTW2): Ditto.
17116 (UNSPEC_SM4E): Ditto.
17117 (UNSPEC_SM4EKEY): Ditto.
17118 * config/aarch64/constraints.md (Ui2): New constraint.
17119 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17120 * config/arm/types.md (crypto_sm3): New type attribute.
17121 (crypto_sm4): Ditto.
17122 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17123 (vsm3tt1aq_u32): Ditto.
17124 (vsm3tt1bq_u32): Ditto.
17125 (vsm3tt2aq_u32): Ditto.
17126 (vsm3tt2bq_u32): Ditto.
17127 (vsm3partw1q_u32): Ditto.
17128 (vsm3partw2q_u32): Ditto.
17129 (vsm4eq_u32): Ditto.
17130 (vsm4ekeyq_u32): Ditto.
17131 (doc/invoke.texi): Document new sm4 option.
17133 2018-01-10 Michael Collison <michael.collison@arm.com>
17135 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17136 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17137 (AARCH64_FL_FOR_ARCH8_4): New.
17138 (AARCH64_FL_V8_4): New flag.
17139 (doc/invoke.texi): Document new armv8.4-a option.
17141 2018-01-10 Michael Collison <michael.collison@arm.com>
17143 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17144 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17145 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17146 * config/aarch64/aarch64-option-extension.def: Add
17147 AARCH64_OPT_EXTENSION of 'sha2'.
17148 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17149 (crypto): Disable sha2 and aes if crypto disabled.
17150 (crypto): Enable aes and sha2 if enabled.
17151 (simd): Disable sha2 and aes if simd disabled.
17152 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17154 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17155 (TARGET_SHA2): New feature flag for sha2.
17156 (TARGET_AES): New feature flag for aes.
17157 * config/aarch64/aarch64-simd.md:
17158 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17159 conditional on TARGET_AES.
17160 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17161 (aarch64_crypto_sha1hsi): Make pattern conditional
17163 (aarch64_crypto_sha1hv4si): Ditto.
17164 (aarch64_be_crypto_sha1hv4si): Ditto.
17165 (aarch64_crypto_sha1su1v4si): Ditto.
17166 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17167 (aarch64_crypto_sha1su0v4si): Ditto.
17168 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17169 (aarch64_crypto_sha256su0v4si): Ditto.
17170 (aarch64_crypto_sha256su1v4si): Ditto.
17171 (doc/invoke.texi): Document new aes and sha2 options.
17173 2018-01-10 Martin Sebor <msebor@redhat.com>
17175 PR tree-optimization/83781
17176 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17179 2018-01-11 Martin Sebor <msebor@gmail.com>
17180 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17182 PR tree-optimization/83501
17183 PR tree-optimization/81703
17185 * tree-ssa-strlen.c (get_string_cst): Rename...
17186 (get_string_len): ...to this. Handle global constants.
17187 (handle_char_store): Adjust.
17189 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17190 Jim Wilson <jimw@sifive.com>
17192 * config/riscv/riscv-protos.h (riscv_output_return): New.
17193 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17194 (riscv_attribute_table, riscv_output_return),
17195 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17196 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17197 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17199 (riscv_expand_prologue): Add early return for naked function.
17200 (riscv_expand_epilogue): Likewise.
17201 (riscv_function_ok_for_sibcall): Return false for naked function.
17202 (riscv_set_current_function): New.
17203 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17204 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17205 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17206 * doc/extend.texi (RISC-V Function Attributes): New.
17208 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17210 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17211 check for 128-bit long double before checking TCmode.
17212 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17213 128-bit long doubles before checking TFmode or TCmode.
17214 (FLOAT128_IBM_P): Likewise.
17216 2018-01-10 Martin Sebor <msebor@redhat.com>
17218 PR tree-optimization/83671
17219 * builtins.c (c_strlen): Unconditionally return zero for the empty
17221 Use -Warray-bounds for warnings.
17222 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17223 for non-constant array indices with COMPONENT_REF, arrays of
17224 arrays, and pointers to arrays.
17225 (gimple_fold_builtin_strlen): Determine and set length range for
17226 non-constant character arrays.
17228 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17230 PR middle-end/81897
17231 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17234 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17236 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17238 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17241 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17242 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17243 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17244 indexed_or_indirect_operand predicate.
17245 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17246 (*vsx_le_perm_load_v8hi): Likewise.
17247 (*vsx_le_perm_load_v16qi): Likewise.
17248 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17249 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17250 (*vsx_le_perm_store_v8hi): Likewise.
17251 (*vsx_le_perm_store_v16qi): Likewise.
17252 (eight unnamed splitters): Likewise.
17254 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17256 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17257 * config/rs6000/emmintrin.h: Likewise.
17258 * config/rs6000/mmintrin.h: Likewise.
17259 * config/rs6000/xmmintrin.h: Likewise.
17261 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17264 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17266 * tree.c (tree_nop_conversion): Return true for location wrapper
17268 (maybe_wrap_with_location): New function.
17269 (selftest::check_strip_nops): New function.
17270 (selftest::test_location_wrappers): New function.
17271 (selftest::tree_c_tests): Call it.
17272 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17273 (maybe_wrap_with_location): New decl.
17274 (EXPR_LOCATION_WRAPPER_P): New macro.
17275 (location_wrapper_p): New inline function.
17276 (tree_strip_any_location_wrapper): New inline function.
17278 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17281 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17282 stack_realign_offset for the largest alignment of stack slot
17284 (ix86_find_max_used_stack_alignment): New function.
17285 (ix86_finalize_stack_frame_flags): Use it. Set
17286 max_used_stack_alignment if we don't realign stack.
17287 * config/i386/i386.h (machine_function): Add
17288 max_used_stack_alignment.
17290 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17292 * config/arm/arm.opt (-mbranch-cost): New option.
17293 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17296 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17299 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17300 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17302 2018-01-10 Richard Biener <rguenther@suse.de>
17305 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17306 early out so it also covers the case where we have a non-NULL
17309 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17311 PR tree-optimization/83753
17312 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17313 for non-strided grouped accesses if the number of elements is 1.
17315 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17318 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17319 * i386.h (TARGET_USE_GATHER): Define.
17320 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17322 2018-01-10 Martin Liska <mliska@suse.cz>
17325 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17326 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17328 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17329 CLEANUP_NO_PARTITIONING is not set.
17331 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17333 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17334 for vectors, as a partial revert of r254296.
17335 * rtl.h (const_vec_p): Delete.
17336 (const_vec_duplicate_p): Don't test for vector CONSTs.
17337 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17338 * expmed.c (make_tree): Likewise.
17341 * common.md (E, F): Use CONSTANT_P instead of checking for
17343 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17344 checking for CONST_VECTOR.
17346 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17348 PR middle-end/83575
17349 * predict.c (force_edge_cold): Handle in more sane way edges
17350 with no prediction.
17352 2018-01-09 Carl Love <cel@us.ibm.com>
17354 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17356 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17357 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17358 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17359 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17360 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17361 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17362 * config/rs6000/rs6000-protos.h: Add extern defition for
17363 rs6000_generate_float2_double_code.
17364 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17366 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17367 (float2_v2df): Add define_expand.
17369 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17372 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17373 op_mode in the force_to_mode call.
17375 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17377 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17378 instead of checking each element individually.
17379 (aarch64_evpc_uzp): Likewise.
17380 (aarch64_evpc_zip): Likewise.
17381 (aarch64_evpc_ext): Likewise.
17382 (aarch64_evpc_rev): Likewise.
17383 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17384 instead of checking each element individually. Return true without
17386 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17387 whether all selected elements come from the same input, instead of
17388 checking each element individually. Remove calls to gen_rtx_REG,
17389 start_sequence and end_sequence and instead assert that no rtl is
17392 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17394 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17395 order of HIGH and CONST checks.
17397 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17399 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17400 if the destination isn't an SSA_NAME.
17402 2018-01-09 Richard Biener <rguenther@suse.de>
17404 PR tree-optimization/83668
17405 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17407 (canonicalize_loop_form): ... here, renamed from ...
17408 (canonicalize_loop_closed_ssa_form): ... this and amended to
17409 swap successor edges for loop exit blocks to make us use
17410 the RPO order we need for initial schedule generation.
17412 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17414 PR tree-optimization/64811
17415 * match.pd: When optimizing comparisons with Inf, avoid
17416 introducing or losing exceptions from comparisons with NaN.
17418 2018-01-09 Martin Liska <mliska@suse.cz>
17421 * asan.c (shadow_mem_size): Add gcc_assert.
17423 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17425 Don't save registers in main().
17428 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17429 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17430 * config/avr/avr.c (avr_set_current_function): Don't error if
17431 naked, OS_task or OS_main are specified at the same time.
17432 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17434 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17436 * common/config/avr/avr-common.c (avr_option_optimization_table):
17437 Switch on -mmain-is-OS_task for optimizing compilations.
17439 2018-01-09 Richard Biener <rguenther@suse.de>
17441 PR tree-optimization/83572
17442 * graphite.c: Include cfganal.h.
17443 (graphite_transform_loops): Connect infinite loops to exit
17444 and remove fake edges at the end.
17446 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17448 * ipa-inline.c (edge_badness): Revert accidental checkin.
17450 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17453 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17454 symbols; not inline clones.
17456 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17459 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17460 hard registers. Formatting fixes.
17462 PR preprocessor/83722
17463 * gcc.c (try_generate_repro): Pass
17464 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17465 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17468 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17469 Kito Cheng <kito.cheng@gmail.com>
17471 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17472 (riscv_leaf_function_p): Delete.
17473 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17475 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17477 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17479 (do_ifelse): New function.
17480 (do_isel): New function.
17481 (do_sub3): New function.
17482 (do_add3): New function.
17483 (do_load_mask_compare): New function.
17484 (do_overlap_load_compare): New function.
17485 (expand_compare_loop): New function.
17486 (expand_block_compare): Call expand_compare_loop() when appropriate.
17487 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17488 option description.
17489 (-mblock-compare-inline-loop-limit): New option.
17491 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17494 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17495 Reverse order of second and third operands in first alternative.
17496 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17497 of first and second elements in UNSPEC_VPERMR vector.
17498 (altivec_expand_vec_perm_le): Likewise.
17500 2018-01-08 Jeff Law <law@redhat.com>
17502 PR rtl-optimizatin/81308
17503 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17504 (process_switch): If group_case_labels makes a change, then set
17506 (pass_convert_switch::execute): If a switch is converted, then
17507 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17509 PR rtl-optimization/81308
17510 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17513 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17515 PR target/83663 - Revert r255946
17516 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17517 generation for cases where splatting a value is not useful.
17518 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17519 across a vec_duplicate and a paradoxical subreg forming a vector
17520 mode to a vec_concat.
17522 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17524 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17525 -march=armv8.3-a variants.
17526 * config/arm/t-multilib: Likewise.
17527 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17529 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17531 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17533 (cceq_ior_compare_complement): Give it a name so I can use it, and
17534 change boolean_or_operator predicate to boolean_operator so it can
17535 be used to generate a crand.
17536 (eqne): New code iterator.
17537 (bd/bd_neg): New code_attrs.
17538 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17539 a single define_insn.
17540 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17541 decrement (bdnzt/bdnzf/bdzt/bdzf).
17542 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17543 with the new names of the branch decrement patterns, and added the
17544 names of the branch decrement conditional patterns.
17546 2018-01-08 Richard Biener <rguenther@suse.de>
17548 PR tree-optimization/83563
17549 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17552 2018-01-08 Richard Biener <rguenther@suse.de>
17554 PR middle-end/83713
17555 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17557 2018-01-08 Richard Biener <rguenther@suse.de>
17559 PR tree-optimization/83685
17560 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17561 references to abnormals.
17563 2018-01-08 Richard Biener <rguenther@suse.de>
17566 * dwarf2out.c (output_indirect_strings): Handle empty
17567 skeleton_debug_str_hash.
17568 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17570 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17572 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17573 (emit_store_direct): Likewise.
17574 (arc_trampoline_adjust_address): Likewise.
17575 (arc_asm_trampoline_template): New function.
17576 (arc_initialize_trampoline): Use asm_trampoline_template.
17577 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17578 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17579 * config/arc/arc.md (flush_icache): Delete pattern.
17581 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17583 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17584 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17587 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17590 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17591 by not USED_FOR_TARGET.
17592 (make_pass_resolve_sw_modes): Likewise.
17594 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17596 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17599 2018-01-08 Richard Biener <rguenther@suse.de>
17601 PR middle-end/83580
17602 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17604 2018-01-08 Richard Biener <rguenther@suse.de>
17606 PR middle-end/83517
17607 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17609 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17611 PR middle-end/81897
17612 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17613 basic blocks with a small number of successors.
17614 (convert_control_dep_chain_into_preds): Improve handling of
17616 (dump_predicates): Split apart into...
17617 (dump_pred_chain): ...here...
17618 (dump_pred_info): ...and here.
17619 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17620 (can_chain_union_be_invalidated_p): Improve check for invalidation
17622 (uninit_uses_cannot_happen): Avoid unnecessary if
17623 convert_control_dep_chain_into_preds yielded nothing.
17625 2018-01-06 Martin Sebor <msebor@redhat.com>
17627 PR tree-optimization/83640
17628 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17629 subtracting negative offset from size.
17630 (builtin_access::overlap): Adjust offset bounds of the access to fall
17631 within the size of the object if possible.
17633 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17635 PR rtl-optimization/83699
17636 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17637 extract_bit_field_as_subreg to cases in which the extracted
17638 value is also a vector.
17640 * lra-constraints.c (process_alt_operands): Test for the equivalence
17641 substitutions when detecting a possible reload cycle.
17643 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17646 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17647 by default if flag_selective_schedling{,2}. Formatting fixes.
17649 PR rtl-optimization/83682
17650 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17651 if it has non-VECTOR_MODE element mode.
17652 (vec_duplicate_p): Likewise.
17654 PR middle-end/83694
17655 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17656 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17658 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17661 * config/i386/i386-builtin.def
17662 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17663 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17664 Require also OPTION_MASK_ISA_AVX512F in addition to
17665 OPTION_MASK_ISA_GFNI.
17666 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17667 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17668 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17669 to OPTION_MASK_ISA_GFNI.
17670 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17671 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17672 OPTION_MASK_ISA_AVX512BW.
17673 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17674 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17675 addition to OPTION_MASK_ISA_GFNI.
17676 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17677 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17678 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17679 to OPTION_MASK_ISA_GFNI.
17680 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17681 a requirement for all ISAs rather than any of them with a few
17683 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17685 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17686 bitmasks to be enabled with 3 exceptions, instead of requiring any
17687 enabled ISA with lots of exceptions.
17688 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17689 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17690 Change avx512bw in isa attribute to avx512f.
17691 * config/i386/sgxintrin.h: Add license boilerplate.
17692 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17693 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17694 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17695 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17697 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17698 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17699 temporarily sse2 rather than sse if not enabled already.
17702 * config/i386/sse.md (VI248_VLBW): Rename to ...
17703 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17704 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17705 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17706 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17707 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17708 mode iterator instead of VI248_VLBW.
17710 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17712 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17713 (record_modified): Skip clobbers; add debug output.
17714 (param_change_prob): Use sreal frequencies.
17716 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17718 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17719 punt for user-aligned variables.
17721 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17723 * tree-chrec.c (chrec_contains_symbols): Return true for
17726 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17729 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17730 of (x|y) == x for BICS pattern.
17732 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17734 PR tree-optimization/83605
17735 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17736 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17739 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17741 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17742 * config/epiphany/rtems.h: New file.
17744 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17745 Uros Bizjak <ubizjak@gmail.com>
17748 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17749 QIreg_operand instead of register_operand predicate.
17750 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17751 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17752 comments instead of -fmitigate[-_]rop.
17754 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17757 * cgraphunit.c (symbol_table::compile): Switch to text_section
17758 before calling assembly_start debug hook.
17759 * run-rtl-passes.c (run_rtl_passes): Likewise.
17762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17764 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17765 range_int_cst_p rather than !symbolic_range_p before calling
17766 extract_range_from_multiplicative_op_1.
17768 2018-01-04 Jeff Law <law@redhat.com>
17770 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17771 redundant test in assertion.
17773 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17775 * doc/rtl.texi: Document machine_mode wrapper classes.
17777 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17779 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17780 using tree_to_uhwi.
17782 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17784 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17785 the VEC_PERM_EXPR fold to fail.
17787 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17790 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17791 to switched_sections.
17793 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17796 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17797 test for d.testing.
17799 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17802 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17803 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17805 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17808 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17809 is BLKmode and bitpos not zero or mode change is needed.
17811 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17814 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17817 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17820 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17821 instead of MULT rtx. Update all corresponding splitters.
17822 (*saddl_se): Ditto.
17823 (*ssub<modesuffix>): Ditto.
17824 (*ssubl_se): Ditto.
17825 (*cmp_sadd_di): Update split patterns.
17826 (*cmp_sadd_si): Ditto.
17827 (*cmp_sadd_sidi): Ditto.
17828 (*cmp_ssub_di): Ditto.
17829 (*cmp_ssub_si): Ditto.
17830 (*cmp_ssub_sidi): Ditto.
17831 * config/alpha/predicates.md (const23_operand): New predicate.
17832 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17833 Look for ASHIFT, not MULT inner operand.
17834 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17836 2018-01-04 Martin Liska <mliska@suse.cz>
17838 PR gcov-profile/83669
17839 * gcov.c (output_intermediate_file): Add version to intermediate
17841 * doc/gcov.texi: Document new field 'version' in intermediate
17842 file format. Fix location of '-k' option of gcov command.
17844 2018-01-04 Martin Liska <mliska@suse.cz>
17847 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17849 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17851 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17853 2018-01-03 Martin Sebor <msebor@redhat.com>
17855 PR tree-optimization/83655
17856 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17857 checking calls with invalid arguments.
17859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17861 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17862 (vectorizable_mask_load_store): Delete.
17863 (vectorizable_call): Return false for masked loads and stores.
17864 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17865 instead of gimple_assign_rhs1.
17866 (vectorizable_load): Handle IFN_MASK_LOAD.
17867 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17871 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17873 (vectorizable_mask_load_store): ...here.
17874 (vectorizable_load): ...and here.
17876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17878 * tree-vect-stmts.c (vect_build_all_ones_mask)
17879 (vect_build_zero_merge_argument): New functions, split out from...
17880 (vectorizable_load): ...here.
17882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17884 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17886 (vectorizable_mask_load_store): ...here.
17887 (vectorizable_store): ...and here.
17889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17891 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17893 (vectorizable_mask_load_store): ...here.
17895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17897 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17898 (vect_model_store_cost): Take a vec_load_store_type instead of a
17900 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17901 (vect_model_store_cost): Take a vec_load_store_type instead of a
17903 (vectorizable_mask_load_store): Update accordingly.
17904 (vectorizable_store): Likewise.
17905 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17909 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17910 IFN_MASK_LOAD calls here rather than...
17911 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17914 Alan Hayward <alan.hayward@arm.com>
17915 David Sherwood <david.sherwood@arm.com>
17917 * expmed.c (extract_bit_field_1): For vector extracts,
17918 fall back to extract_bit_field_as_subreg if vec_extract
17921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17922 Alan Hayward <alan.hayward@arm.com>
17923 David Sherwood <david.sherwood@arm.com>
17925 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17926 they are variable or constant sized.
17927 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17928 slots for constant-sized data.
17930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17931 Alan Hayward <alan.hayward@arm.com>
17932 David Sherwood <david.sherwood@arm.com>
17934 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17935 handling COND_EXPRs with boolean comparisons, try to find a better
17936 basis for the mask type than the boolean itself.
17938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17940 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17941 is calculated and how it can be overridden.
17942 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17943 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17945 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17949 Alan Hayward <alan.hayward@arm.com>
17950 David Sherwood <david.sherwood@arm.com>
17952 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17953 Remove the mode argument.
17954 (aarch64_simd_valid_immediate): Remove the mode and inverse
17956 * config/aarch64/iterators.md (bitsize): New iterator.
17957 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17958 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17959 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17960 aarch64_simd_valid_immediate.
17961 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17962 (aarch64_reg_or_bic_imm): Likewise.
17963 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17964 with an insn_type enum and msl with a modifier_type enum.
17965 Replace element_width with a scalar_mode. Change the shift
17966 to unsigned int. Add constructors for scalar_float_mode and
17967 scalar_int_mode elements.
17968 (aarch64_vect_float_const_representable_p): Delete.
17969 (aarch64_can_const_movi_rtx_p)
17970 (aarch64_simd_scalar_immediate_valid_for_move)
17971 (aarch64_simd_make_constant): Update call to
17972 aarch64_simd_valid_immediate.
17973 (aarch64_advsimd_valid_immediate_hs): New function.
17974 (aarch64_advsimd_valid_immediate): Likewise.
17975 (aarch64_simd_valid_immediate): Remove mode and inverse
17976 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17977 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17978 and aarch64_float_const_representable_p on the result.
17979 (aarch64_output_simd_mov_immediate): Remove mode argument.
17980 Update call to aarch64_simd_valid_immediate and use of
17981 simd_immediate_info.
17982 (aarch64_output_scalar_simd_mov_immediate): Update call
17985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17986 Alan Hayward <alan.hayward@arm.com>
17987 David Sherwood <david.sherwood@arm.com>
17989 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17990 (mode_nunits): Likewise CONST_MODE_NUNITS.
17991 * machmode.def (ADJUST_NUNITS): Document.
17992 * genmodes.c (mode_data::need_nunits_adj): New field.
17993 (blank_mode): Update accordingly.
17994 (adj_nunits): New variable.
17995 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17997 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17998 listed in adj_nunits.
17999 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18000 listed in adj_nunits. Don't emit case statements for such modes.
18001 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18002 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18003 nothing if adj_nunits is nonnull.
18004 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18005 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18006 (emit_mode_fbit): Update use of print_maybe_const_decl.
18007 (emit_move_size): Likewise. Treat the array as non-const
18009 (emit_mode_adjustments): Handle adj_nunits.
18011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18013 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18014 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18015 (VECTOR_MODES): Use it.
18016 (make_vector_modes): Take the prefix as an argument.
18018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18019 Alan Hayward <alan.hayward@arm.com>
18020 David Sherwood <david.sherwood@arm.com>
18022 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18023 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18024 for MODE_VECTOR_BOOL.
18025 * machmode.def (VECTOR_BOOL_MODE): Document.
18026 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18027 (make_vector_bool_mode): New function.
18028 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18030 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18031 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18033 * stor-layout.c (int_mode_for_mode): Likewise.
18034 * tree.c (build_vector_type_for_mode): Likewise.
18035 * varasm.c (output_constant_pool_2): Likewise.
18036 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18037 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18038 for MODE_VECTOR_BOOL.
18039 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18040 of mode class checks.
18041 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18042 instead of a list of mode class checks.
18043 (expand_vector_scalar_condition): Likewise.
18044 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18047 Alan Hayward <alan.hayward@arm.com>
18048 David Sherwood <david.sherwood@arm.com>
18050 * machmode.h (mode_size): Change from unsigned short to
18052 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18053 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18054 or if measurement_type is not polynomial.
18055 (fixed_size_mode::includes_p): Check for constant-sized modes.
18056 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18057 return a poly_uint16 rather than an unsigned short.
18058 (emit_mode_size): Change the type of mode_size from unsigned short
18059 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18060 (emit_mode_adjustments): Cope with polynomial vector sizes.
18061 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18063 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18065 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18066 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18067 * caller-save.c (setup_save_areas): Likewise.
18068 (replace_reg_with_saved_mem): Likewise.
18069 * calls.c (emit_library_call_value_1): Likewise.
18070 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18071 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18072 (gen_lowpart_for_combine): Likewise.
18073 * convert.c (convert_to_integer_1): Likewise.
18074 * cse.c (equiv_constant, cse_insn): Likewise.
18075 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18076 (cselib_subst_to_values): Likewise.
18077 * dce.c (word_dce_process_block): Likewise.
18078 * df-problems.c (df_word_lr_mark_ref): Likewise.
18079 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18080 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18081 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18082 (rtl_for_decl_location): Likewise.
18083 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18084 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18085 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18086 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18087 (expand_expr_real_1): Likewise.
18088 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18089 (pad_below): Likewise.
18090 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18091 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18092 * ira.c (get_subreg_tracking_sizes): Likewise.
18093 * ira-build.c (ira_create_allocno_objects): Likewise.
18094 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18095 (ira_sort_regnos_for_alter_reg): Likewise.
18096 * ira-costs.c (record_operand_costs): Likewise.
18097 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18098 (resolve_simple_move): Likewise.
18099 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18100 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18101 (lra_constraints): Likewise.
18102 (CONST_POOL_OK_P): Reject variable-sized modes.
18103 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18104 (add_pseudo_to_slot, lra_spill): Likewise.
18105 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18106 * optabs-query.c (get_best_extraction_insn): Likewise.
18107 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18108 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18109 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18110 * recog.c (offsettable_address_addr_space_p): Likewise.
18111 * regcprop.c (maybe_mode_change): Likewise.
18112 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18113 * regrename.c (build_def_use): Likewise.
18114 * regstat.c (dump_reg_info): Likewise.
18115 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18116 (find_reloads, find_reloads_subreg_address): Likewise.
18117 * reload1.c (eliminate_regs_1): Likewise.
18118 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18119 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18120 (simplify_binary_operation_1, simplify_subreg): Likewise.
18121 * targhooks.c (default_function_arg_padding): Likewise.
18122 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18123 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18124 (verify_gimple_assign_ternary): Likewise.
18125 * tree-inline.c (estimate_move_cost): Likewise.
18126 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18127 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18128 (get_address_cost_ainc): Likewise.
18129 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18130 (vect_supportable_dr_alignment): Likewise.
18131 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18132 (vectorizable_reduction): Likewise.
18133 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18134 (vectorizable_operation, vectorizable_load): Likewise.
18135 * tree.c (build_same_sized_truth_vector_type): Likewise.
18136 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18137 * var-tracking.c (emit_note_insn_var_location): Likewise.
18138 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18139 (ADDR_VEC_ALIGN): Likewise.
18141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18142 Alan Hayward <alan.hayward@arm.com>
18143 David Sherwood <david.sherwood@arm.com>
18145 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18147 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18148 or if measurement_type is polynomial.
18149 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18150 * combine.c (make_extraction): Likewise.
18151 * dse.c (find_shift_sequence): Likewise.
18152 * dwarf2out.c (mem_loc_descriptor): Likewise.
18153 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18154 (extract_bit_field, extract_low_bits): Likewise.
18155 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18156 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18157 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18158 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18159 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18160 * reload.c (find_reloads): Likewise.
18161 * reload1.c (alter_reg): Likewise.
18162 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18163 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18164 * tree-if-conv.c (predicate_mem_writes): Likewise.
18165 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18166 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18167 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18168 * valtrack.c (dead_debug_insert_temp): Likewise.
18169 * varasm.c (mergeable_constant_section): Likewise.
18170 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18173 Alan Hayward <alan.hayward@arm.com>
18174 David Sherwood <david.sherwood@arm.com>
18176 * expr.c (expand_assignment): Cope with polynomial mode sizes
18177 when assigning to a CONCAT.
18179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18180 Alan Hayward <alan.hayward@arm.com>
18181 David Sherwood <david.sherwood@arm.com>
18183 * machmode.h (mode_precision): Change from unsigned short to
18185 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18187 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18188 or if measurement_type is not polynomial.
18189 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18190 in which the mode is already known to be a scalar_int_mode.
18191 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18192 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18194 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18195 for GET_MODE_PRECISION.
18196 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18197 for GET_MODE_PRECISION.
18198 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18200 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18201 (expand_field_assignment, make_extraction): Likewise.
18202 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18203 (get_last_value): Likewise.
18204 * convert.c (convert_to_integer_1): Likewise.
18205 * cse.c (cse_insn): Likewise.
18206 * expr.c (expand_expr_real_1): Likewise.
18207 * lra-constraints.c (simplify_operand_subreg): Likewise.
18208 * optabs-query.c (can_atomic_load_p): Likewise.
18209 * optabs.c (expand_atomic_load): Likewise.
18210 (expand_atomic_store): Likewise.
18211 * ree.c (combine_reaching_defs): Likewise.
18212 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18213 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18214 * tree.h (type_has_mode_precision_p): Likewise.
18215 * ubsan.c (instrument_si_overflow): Likewise.
18217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18218 Alan Hayward <alan.hayward@arm.com>
18219 David Sherwood <david.sherwood@arm.com>
18221 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18222 polynomial numbers of units.
18223 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18224 (valid_vector_subparts_p): New function.
18225 (build_vector_type): Remove temporary shim and take the number
18226 of units as a poly_uint64 rather than an int.
18227 (build_opaque_vector_type): Take the number of units as a
18228 poly_uint64 rather than an int.
18229 * tree.c (build_vector_from_ctor): Handle polynomial
18230 TYPE_VECTOR_SUBPARTS.
18231 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18232 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18233 (build_vector_from_val): If the number of units is variable,
18234 use build_vec_duplicate_cst for constant operands and
18235 VEC_DUPLICATE_EXPR otherwise.
18236 (make_vector_type): Remove temporary is_constant ().
18237 (build_vector_type, build_opaque_vector_type): Take the number of
18238 units as a poly_uint64 rather than an int.
18239 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18241 * cfgexpand.c (expand_debug_expr): Likewise.
18242 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18243 (store_constructor, expand_expr_real_1): Likewise.
18244 (const_scalar_mask_from_tree): Likewise.
18245 * fold-const-call.c (fold_const_reduction): Likewise.
18246 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18247 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18248 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18249 (fold_relational_const): Likewise.
18250 (native_interpret_vector): Likewise. Change the size from an
18251 int to an unsigned int.
18252 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18253 TYPE_VECTOR_SUBPARTS.
18254 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18255 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18256 duplicating a non-constant operand into a variable-length vector.
18257 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18258 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18259 * ipa-icf.c (sem_variable::equals): Likewise.
18260 * match.pd: Likewise.
18261 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18262 * print-tree.c (print_node): Likewise.
18263 * stor-layout.c (layout_type): Likewise.
18264 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18265 * tree-cfg.c (verify_gimple_comparison): Likewise.
18266 (verify_gimple_assign_binary): Likewise.
18267 (verify_gimple_assign_ternary): Likewise.
18268 (verify_gimple_assign_single): Likewise.
18269 * tree-pretty-print.c (dump_generic_node): Likewise.
18270 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18271 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18272 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18273 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18274 (vect_shift_permute_load_chain): Likewise.
18275 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18276 (expand_vector_condition, optimize_vector_constructor): Likewise.
18277 (lower_vec_perm, get_compute_type): Likewise.
18278 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18279 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18280 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18281 (vect_recog_mask_conversion_pattern): Likewise.
18282 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18283 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18284 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18285 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18286 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18287 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18288 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18289 (supportable_widening_operation): Likewise.
18290 (supportable_narrowing_operation): Likewise.
18291 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18293 * varasm.c (output_constant): Likewise.
18295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18296 Alan Hayward <alan.hayward@arm.com>
18297 David Sherwood <david.sherwood@arm.com>
18299 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18300 so that both the length == 3 and length != 3 cases set up their
18301 own permute vectors. Add comments explaining why we know the
18302 number of elements is constant.
18303 (vect_permute_load_chain): Likewise.
18305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18306 Alan Hayward <alan.hayward@arm.com>
18307 David Sherwood <david.sherwood@arm.com>
18309 * machmode.h (mode_nunits): Change from unsigned char to
18311 (ONLY_FIXED_SIZE_MODES): New macro.
18312 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18313 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18314 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18316 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18317 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18318 or if measurement_type is not polynomial.
18319 * genmodes.c (ZERO_COEFFS): New macro.
18320 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18322 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18323 Use ZERO_COEFFS when emitting initializers.
18324 * data-streamer.h (bp_pack_poly_value): New function.
18325 (bp_unpack_poly_value): Likewise.
18326 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18327 for GET_MODE_NUNITS.
18328 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18329 for GET_MODE_NUNITS.
18330 * tree.c (make_vector_type): Remove temporary shim and make
18331 the real function take the number of units as a poly_uint64
18332 rather than an int.
18333 (build_vector_type_for_mode): Handle polynomial nunits.
18334 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18335 * emit-rtl.c (const_vec_series_p_1): Likewise.
18336 (gen_rtx_CONST_VECTOR): Likewise.
18337 * fold-const.c (test_vec_duplicate_folding): Likewise.
18338 * genrecog.c (validate_pattern): Likewise.
18339 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18340 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18341 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18342 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18343 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18344 * rtlanal.c (subreg_get_info): Likewise.
18345 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18346 (vect_grouped_load_supported): Likewise.
18347 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18348 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18349 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18350 (simplify_const_unary_operation, simplify_binary_operation_1)
18351 (simplify_const_binary_operation, simplify_ternary_operation)
18352 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18353 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18354 instead of CONST_VECTOR_NUNITS.
18355 * varasm.c (output_constant_pool_2): Likewise.
18356 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18357 explicit-encoded elements in the XVEC for variable-length vectors.
18359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18361 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18364 Alan Hayward <alan.hayward@arm.com>
18365 David Sherwood <david.sherwood@arm.com>
18367 * coretypes.h (fixed_size_mode): Declare.
18368 (fixed_size_mode_pod): New typedef.
18369 * builtins.h (target_builtins::x_apply_args_mode)
18370 (target_builtins::x_apply_result_mode): Change type to
18371 fixed_size_mode_pod.
18372 * builtins.c (apply_args_size, apply_result_size, result_vector)
18373 (expand_builtin_apply_args_1, expand_builtin_apply)
18374 (expand_builtin_return): Update accordingly.
18376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18378 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18379 * cselib.c (cselib_hash_rtx): Likewise.
18380 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18381 CONST_VECTOR encoding.
18383 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18384 Jeff Law <law@redhat.com>
18387 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18388 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18389 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18390 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18393 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18394 explicitly probe *sp in a noreturn function if there were any callee
18395 register saves or frame pointer is needed.
18397 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18400 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18401 BLKmode for ternary, binary or unary expressions.
18404 * var-tracking.c (delete_vta_debug_insn): New inline function.
18405 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18406 insns from get_insns () to NULL instead of each bb separately.
18407 Use delete_vta_debug_insn. No longer static.
18408 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18409 delete_vta_debug_insns callers.
18410 * rtl.h (delete_vta_debug_insns): Declare.
18411 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18412 instead of variable_tracking_main.
18414 2018-01-03 Martin Sebor <msebor@redhat.com>
18416 PR tree-optimization/83603
18417 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18418 arguments past the endof the argument list in functions declared
18419 without a prototype.
18420 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18421 Avoid checking when arguments are null.
18423 2018-01-03 Martin Sebor <msebor@redhat.com>
18426 * doc/extend.texi (attribute const): Fix a typo.
18427 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18428 issuing -Wsuggest-attribute for void functions.
18430 2018-01-03 Martin Sebor <msebor@redhat.com>
18432 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18433 offset_int::from instead of wide_int::to_shwi.
18434 (maybe_diag_overlap): Remove assertion.
18435 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18436 * gimple-ssa-sprintf.c (format_directive): Same.
18437 (parse_directive): Same.
18438 (sprintf_dom_walker::compute_format_length): Same.
18439 (try_substitute_return_value): Same.
18441 2018-01-03 Jeff Law <law@redhat.com>
18443 PR middle-end/83654
18444 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18445 non-constant residual for zero at runtime and avoid probing in
18446 that case. Reorganize code for trailing problem to mirror handling
18449 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18451 PR tree-optimization/83501
18452 * tree-ssa-strlen.c (get_string_cst): New.
18453 (handle_char_store): Call get_string_cst.
18455 2018-01-03 Martin Liska <mliska@suse.cz>
18457 PR tree-optimization/83593
18458 * tree-ssa-strlen.c: Include tree-cfg.h.
18459 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18460 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18461 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18463 (strlen_dom_walker::before_dom_children): Call
18464 gimple_purge_dead_eh_edges. Dump tranformation with details
18466 (strlen_dom_walker::before_dom_children): Update call by adding
18467 new argument cleanup_eh.
18468 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18470 2018-01-03 Martin Liska <mliska@suse.cz>
18473 * cif-code.def (VARIADIC_THUNK): New enum value.
18474 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18477 2018-01-03 Jan Beulich <jbeulich@suse.com>
18479 * sse.md (mov<mode>_internal): Tighten condition for when to use
18480 vmovdqu<ssescalarsize> for TI and OI modes.
18482 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18484 Update copyright years.
18486 2018-01-03 Martin Liska <mliska@suse.cz>
18489 * ipa-visibility.c (function_and_variable_visibility): Skip
18490 functions with noipa attribure.
18492 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18494 * gcc.c (process_command): Update copyright notice dates.
18495 * gcov-dump.c (print_version): Ditto.
18496 * gcov.c (print_version): Ditto.
18497 * gcov-tool.c (print_version): Ditto.
18498 * gengtype.c (create_file): Ditto.
18499 * doc/cpp.texi: Bump @copying's copyright year.
18500 * doc/cppinternals.texi: Ditto.
18501 * doc/gcc.texi: Ditto.
18502 * doc/gccint.texi: Ditto.
18503 * doc/gcov.texi: Ditto.
18504 * doc/install.texi: Ditto.
18505 * doc/invoke.texi: Ditto.
18507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18509 * vector-builder.h (vector_builder::m_full_nelts): Change from
18510 unsigned int to poly_uint64.
18511 (vector_builder::full_nelts): Update prototype accordingly.
18512 (vector_builder::new_vector): Likewise.
18513 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18514 (vector_builder::operator ==): Likewise.
18515 (vector_builder::finalize): Likewise.
18516 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18517 Take the number of elements as a poly_uint64 rather than an
18519 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18520 from unsigned int to poly_uint64.
18521 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18522 (vec_perm_indices::new_vector): Likewise.
18523 (vec_perm_indices::length): Likewise.
18524 (vec_perm_indices::nelts_per_input): Likewise.
18525 (vec_perm_indices::input_nelts): Likewise.
18526 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18527 number of elements per input as a poly_uint64 rather than an
18528 unsigned int. Use the original encoding for variable-length
18529 vectors, rather than clamping each individual element.
18530 For the second and subsequent elements in each pattern,
18531 clamp the step and base before clamping their sum.
18532 (vec_perm_indices::series_p): Handle polynomial element counts.
18533 (vec_perm_indices::all_in_range_p): Likewise.
18534 (vec_perm_indices_to_tree): Likewise.
18535 (vec_perm_indices_to_rtx): Likewise.
18536 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18537 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18538 (tree_vector_builder::new_binary_operation): Handle polynomial
18539 element counts. Return false if we need to know the number
18540 of elements at compile time.
18541 * fold-const.c (fold_vec_perm): Punt if the number of elements
18542 isn't known at compile time.
18544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18546 * vec-perm-indices.h (vec_perm_builder): Change element type
18547 from HOST_WIDE_INT to poly_int64.
18548 (vec_perm_indices::element_type): Update accordingly.
18549 (vec_perm_indices::clamp): Handle polynomial element_types.
18550 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18551 (vec_perm_indices::all_in_range_p): Likewise.
18552 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18554 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18555 polynomial vec_perm_indices element types.
18556 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18557 * fold-const.c (fold_vec_perm): Likewise.
18558 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18559 * tree-vect-generic.c (lower_vec_perm): Likewise.
18560 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18561 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18562 element type to HOST_WIDE_INT.
18564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18565 Alan Hayward <alan.hayward@arm.com>
18566 David Sherwood <david.sherwood@arm.com>
18568 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18569 rather than an int. Use plus_constant.
18570 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18571 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18574 Alan Hayward <alan.hayward@arm.com>
18575 David Sherwood <david.sherwood@arm.com>
18577 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18578 a HOST_WIDE_INT to a poly_int64.
18580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18581 Alan Hayward <alan.hayward@arm.com>
18582 David Sherwood <david.sherwood@arm.com>
18584 * calls.c (load_register_parameters): Cope with polynomial
18585 mode sizes. Require a constant size for BLKmode parameters
18586 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18587 forces a parameter to be padded at the lsb end in order to
18588 fill a complete number of words, require the parameter size
18589 to be ordered wrt UNITS_PER_WORD.
18591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18592 Alan Hayward <alan.hayward@arm.com>
18593 David Sherwood <david.sherwood@arm.com>
18595 * reload1.c (spill_stack_slot_width): Change element type
18596 from unsigned int to poly_uint64_pod.
18597 (alter_reg): Treat mode sizes as polynomial.
18599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18600 Alan Hayward <alan.hayward@arm.com>
18601 David Sherwood <david.sherwood@arm.com>
18603 * reload.c (complex_word_subreg_p): New function.
18604 (reload_inner_reg_of_subreg, push_reload): Use it.
18606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18607 Alan Hayward <alan.hayward@arm.com>
18608 David Sherwood <david.sherwood@arm.com>
18610 * lra-constraints.c (process_alt_operands): Reject matched
18611 operands whose sizes aren't ordered.
18612 (match_reload): Refer to this check here.
18614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18615 Alan Hayward <alan.hayward@arm.com>
18616 David Sherwood <david.sherwood@arm.com>
18618 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18619 that the mode size is in the set {1, 2, 4, 8, 16}.
18621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18622 Alan Hayward <alan.hayward@arm.com>
18623 David Sherwood <david.sherwood@arm.com>
18625 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18626 Use plus_constant instead of gen_rtx_PLUS.
18628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18629 Alan Hayward <alan.hayward@arm.com>
18630 David Sherwood <david.sherwood@arm.com>
18632 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18633 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18634 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18635 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18636 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18637 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18638 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18639 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18640 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18641 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18643 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18644 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18645 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18646 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18647 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18648 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18649 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18650 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18651 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18652 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18654 * expr.c (emit_move_resolve_push): Treat the input and result
18655 of PUSH_ROUNDING as a poly_int64.
18656 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18657 (emit_push_insn): Likewise.
18658 * lra-eliminations.c (mark_not_eliminable): Likewise.
18659 * recog.c (push_operand): Likewise.
18660 * reload1.c (elimination_effects): Likewise.
18661 * rtlanal.c (nonzero_bits1): Likewise.
18662 * calls.c (store_one_arg): Likewise. Require the padding to be
18663 known at compile time.
18665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18666 Alan Hayward <alan.hayward@arm.com>
18667 David Sherwood <david.sherwood@arm.com>
18669 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18670 Use plus_constant instead of gen_rtx_PLUS.
18672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18673 Alan Hayward <alan.hayward@arm.com>
18674 David Sherwood <david.sherwood@arm.com>
18676 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18677 rather than an int.
18679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18680 Alan Hayward <alan.hayward@arm.com>
18681 David Sherwood <david.sherwood@arm.com>
18683 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18684 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18685 via stack temporaries. Treat the mode size as polynomial too.
18687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18688 Alan Hayward <alan.hayward@arm.com>
18689 David Sherwood <david.sherwood@arm.com>
18691 * expr.c (expand_expr_real_2): When handling conversions involving
18692 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18693 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18694 as a poly_uint64 too.
18696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18697 Alan Hayward <alan.hayward@arm.com>
18698 David Sherwood <david.sherwood@arm.com>
18700 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18703 Alan Hayward <alan.hayward@arm.com>
18704 David Sherwood <david.sherwood@arm.com>
18706 * combine.c (can_change_dest_mode): Handle polynomial
18707 REGMODE_NATURAL_SIZE.
18708 * expmed.c (store_bit_field_1): Likewise.
18709 * expr.c (store_constructor): Likewise.
18710 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18711 and polynomial REGMODE_NATURAL_SIZE.
18712 (gen_lowpart_common): Likewise.
18713 * reginfo.c (record_subregs_of_mode): Likewise.
18714 * rtlanal.c (read_modify_subreg_p): Likewise.
18716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18717 Alan Hayward <alan.hayward@arm.com>
18718 David Sherwood <david.sherwood@arm.com>
18720 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18721 numbers of elements.
18723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18724 Alan Hayward <alan.hayward@arm.com>
18725 David Sherwood <david.sherwood@arm.com>
18727 * match.pd: Cope with polynomial numbers of vector elements.
18729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18730 Alan Hayward <alan.hayward@arm.com>
18731 David Sherwood <david.sherwood@arm.com>
18733 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18734 in a POINTER_PLUS_EXPR.
18736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18737 Alan Hayward <alan.hayward@arm.com>
18738 David Sherwood <david.sherwood@arm.com>
18740 * omp-simd-clone.c (simd_clone_subparts): New function.
18741 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18742 (ipa_simd_modify_function_body): Likewise.
18744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18745 Alan Hayward <alan.hayward@arm.com>
18746 David Sherwood <david.sherwood@arm.com>
18748 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18749 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18750 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18751 (expand_vector_condition, vector_element): Likewise.
18752 (subparts_gt): New function.
18753 (get_compute_type): Use subparts_gt.
18754 (count_type_subparts): Delete.
18755 (expand_vector_operations_1): Use subparts_gt instead of
18756 count_type_subparts.
18758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18759 Alan Hayward <alan.hayward@arm.com>
18760 David Sherwood <david.sherwood@arm.com>
18762 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18763 (vect_compile_time_alias): ...this new function. Do the calculation
18764 on poly_ints rather than trees.
18765 (vect_prune_runtime_alias_test_list): Update call accordingly.
18767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18768 Alan Hayward <alan.hayward@arm.com>
18769 David Sherwood <david.sherwood@arm.com>
18771 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18773 (vect_schedule_slp_instance): Likewise.
18775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18776 Alan Hayward <alan.hayward@arm.com>
18777 David Sherwood <david.sherwood@arm.com>
18779 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18780 constant and extern definitions for variable-length vectors.
18781 (vect_get_constant_vectors): Note that the number of units
18782 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
18789 of units as polynomial. Choose between WIDE and NARROW based
18792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18793 Alan Hayward <alan.hayward@arm.com>
18794 David Sherwood <david.sherwood@arm.com>
18796 * tree-vect-stmts.c (simd_clone_subparts): New function.
18797 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18800 Alan Hayward <alan.hayward@arm.com>
18801 David Sherwood <david.sherwood@arm.com>
18803 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18804 vectors as polynomial. Use build_index_vector for
18805 IFN_GOMP_SIMD_LANE.
18807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18808 Alan Hayward <alan.hayward@arm.com>
18809 David Sherwood <david.sherwood@arm.com>
18811 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18812 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18813 for variable-length vectors.
18814 (vectorizable_mask_load_store): Treat the number of units as
18815 polynomial, asserting that it is constant if the condition has
18816 already been enforced.
18817 (vectorizable_store, vectorizable_load): Likewise.
18819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18820 Alan Hayward <alan.hayward@arm.com>
18821 David Sherwood <david.sherwood@arm.com>
18823 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18824 of units as polynomial. Punt if we can't tell at compile time
18825 which vector contains the final result.
18827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18828 Alan Hayward <alan.hayward@arm.com>
18829 David Sherwood <david.sherwood@arm.com>
18831 * tree-vect-loop.c (vectorizable_induction): Treat the number
18832 of units as polynomial. Punt on SLP inductions. Use an integer
18833 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18834 cast of such a series for variable-length floating-point
18837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18838 Alan Hayward <alan.hayward@arm.com>
18839 David Sherwood <david.sherwood@arm.com>
18841 * tree.h (build_index_vector): Declare.
18842 * tree.c (build_index_vector): New function.
18843 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18844 of units as polynomial, forcibly converting it to a constant if
18845 vectorizable_reduction has already enforced the condition.
18846 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18847 to create a {1,2,3,...} vector.
18848 (vectorizable_reduction): Treat the number of units as polynomial.
18849 Choose vectype_in based on the largest scalar element size rather
18850 than the smallest number of units. Enforce the restrictions
18853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18854 Alan Hayward <alan.hayward@arm.com>
18855 David Sherwood <david.sherwood@arm.com>
18857 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18858 number of units as polynomial.
18860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18861 Alan Hayward <alan.hayward@arm.com>
18862 David Sherwood <david.sherwood@arm.com>
18864 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18865 * target.def (autovectorize_vector_sizes): Return the vector sizes
18866 by pointer, using vector_sizes rather than a bitmask.
18867 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18868 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18869 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18871 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18872 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18873 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18874 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18875 * omp-general.c (omp_max_vf): Likewise.
18876 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18877 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18878 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18879 * tree-vect-slp.c (vect_slp_bb): Likewise.
18880 * doc/tm.texi: Regenerate.
18881 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18883 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18884 the vector size as a poly_uint64 rather than an unsigned int.
18885 (current_vector_size): Change from an unsigned int to a poly_uint64.
18886 (get_vectype_for_scalar_type): Update accordingly.
18887 * tree.h (build_truth_vector_type): Take the size and number of
18888 units as a poly_uint64 rather than an unsigned int.
18889 (build_vector_type): Add a temporary overload that takes
18890 the number of units as a poly_uint64 rather than an unsigned int.
18891 * tree.c (make_vector_type): Likewise.
18892 (build_truth_vector_type): Take the number of units as a poly_uint64
18893 rather than an unsigned int.
18895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18896 Alan Hayward <alan.hayward@arm.com>
18897 David Sherwood <david.sherwood@arm.com>
18899 * target.def (get_mask_mode): Take the number of units and length
18900 as poly_uint64s rather than unsigned ints.
18901 * targhooks.h (default_get_mask_mode): Update accordingly.
18902 * targhooks.c (default_get_mask_mode): Likewise.
18903 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18904 * doc/tm.texi: Regenerate.
18906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18907 Alan Hayward <alan.hayward@arm.com>
18908 David Sherwood <david.sherwood@arm.com>
18910 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18911 * omp-general.c (omp_max_vf): Likewise.
18912 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18913 (expand_omp_simd): Handle polynomial safelen.
18914 * omp-low.c (omplow_simd_context): Add a default constructor.
18915 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18916 (lower_rec_simd_input_clauses): Update accordingly.
18917 (lower_rec_input_clauses): Likewise.
18919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18920 Alan Hayward <alan.hayward@arm.com>
18921 David Sherwood <david.sherwood@arm.com>
18923 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18924 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18925 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18926 (vect_analyze_slp_cost): Likewise.
18927 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18928 (vect_model_load_cost): Likewise.
18930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18931 Alan Hayward <alan.hayward@arm.com>
18932 David Sherwood <david.sherwood@arm.com>
18934 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18935 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18936 from an unsigned int * to a poly_uint64_pod *.
18937 (calculate_unrolling_factor): New function.
18938 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18941 Alan Hayward <alan.hayward@arm.com>
18942 David Sherwood <david.sherwood@arm.com>
18944 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18945 from an unsigned int to a poly_uint64.
18946 (_loop_vec_info::slp_unrolling_factor): Likewise.
18947 (_loop_vec_info::vectorization_factor): Change from an int
18949 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18950 (vect_get_num_vectors): New function.
18951 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18952 (vect_get_num_copies): Use vect_get_num_vectors.
18953 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18954 to an unsigned int *.
18955 (vect_analyze_data_refs): Change min_vf from an int * to a
18957 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18958 than an unsigned HOST_WIDE_INT.
18959 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18960 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18961 to an unsigned int *.
18962 (vect_analyze_data_ref_dependences): Likewise.
18963 (vect_compute_data_ref_alignment): Handle polynomial vf.
18964 (vect_enhance_data_refs_alignment): Likewise.
18965 (vect_prune_runtime_alias_test_list): Likewise.
18966 (vect_shift_permute_load_chain): Likewise.
18967 (vect_supportable_dr_alignment): Likewise.
18968 (dependence_distance_ge_vf): Take the vectorization factor as a
18969 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18970 (vect_analyze_data_refs): Change min_vf from an int * to a
18972 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18973 vfm1 as a poly_uint64 rather than an int. Make the same change
18974 for the returned bound_scalar.
18975 (vect_gen_vector_loop_niters): Handle polynomial vf.
18976 (vect_do_peeling): Likewise. Update call to
18977 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18978 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18980 * tree-vect-loop.c (vect_determine_vectorization_factor)
18981 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18982 (vect_get_known_peeling_cost): Likewise.
18983 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18984 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18985 (vect_transform_loop): Likewise. Use the lowest possible VF when
18986 updating the upper bounds of the loop.
18987 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18988 rather than an int.
18989 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18990 polynomial unroll factors.
18991 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18992 (vect_make_slp_decision): Likewise.
18993 (vect_supported_load_permutation_p): Likewise, and polynomial
18995 (vect_analyze_slp_cost): Handle polynomial vf.
18996 (vect_slp_analyze_node_operations): Likewise.
18997 (vect_slp_analyze_bb_1): Likewise.
18998 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18999 than an unsigned HOST_WIDE_INT.
19000 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19001 (vectorizable_load): Handle polynomial vf.
19002 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19004 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19007 Alan Hayward <alan.hayward@arm.com>
19008 David Sherwood <david.sherwood@arm.com>
19010 * match.pd: Handle bit operations involving three constants
19011 and try to fold one pair.
19013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19015 * tree-vect-loop-manip.c: Include gimple-fold.h.
19016 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19017 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19018 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19019 Add a path that uses a step of VF instead of 1, but disable it
19021 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19022 and niters_no_overflow parameters. Update calls to
19023 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19024 Create a new SSA name if the latter choses to use a ste other
19025 than zero, and return it via niters_vector_mult_vf_var.
19026 * tree-vect-loop.c (vect_transform_loop): Update calls to
19027 vect_do_peeling, vect_gen_vector_loop_niters and
19028 slpeel_make_loop_iterate_ntimes.
19029 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19030 (vect_gen_vector_loop_niters): Update declarations after above changes.
19032 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19034 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19035 128-bit round to integer instructions.
19036 (ceil<mode>2): Likewise.
19037 (btrunc<mode>2): Likewise.
19038 (round<mode>2): Likewise.
19040 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19042 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19043 unaligned VSX load/store on P8/P9.
19044 (expand_block_clear): Allow the use of unaligned VSX
19045 load/store on P8/P9.
19047 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19049 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19051 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19052 swap associated with both a load and a store.
19054 2018-01-02 Andrew Waterman <andrew@sifive.com>
19056 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19057 * config/riscv/riscv.md (clear_cache): Use it.
19059 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19061 * web.c: Remove out-of-date comment.
19063 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19065 * expr.c (fixup_args_size_notes): Check that any existing
19066 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19067 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19068 (emit_single_push_insn): ...here.
19070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19072 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19073 (const_vector_encoded_nelts): New function.
19074 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19075 (const_vector_int_elt, const_vector_elt): Declare.
19076 * emit-rtl.c (const_vector_int_elt_1): New function.
19077 (const_vector_elt): Likewise.
19078 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19079 of CONST_VECTOR_ELT.
19081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19083 * expr.c: Include rtx-vector-builder.h.
19084 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19085 directly on the tree encoding.
19086 (const_vector_from_tree): Likewise.
19087 * optabs.c: Include rtx-vector-builder.h.
19088 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19089 sequence of "u" values.
19090 * vec-perm-indices.c: Include rtx-vector-builder.h.
19091 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19092 directly on the vec_perm_indices encoding.
19094 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19096 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19097 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19098 * rtx-vector-builder.h: New file.
19099 * rtx-vector-builder.c: Likewise.
19100 * rtl.h (rtx_def::u2): Add a const_vector field.
19101 (CONST_VECTOR_NPATTERNS): New macro.
19102 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19103 (CONST_VECTOR_DUPLICATE_P): Likewise.
19104 (CONST_VECTOR_STEPPED_P): Likewise.
19105 (CONST_VECTOR_ENCODED_ELT): Likewise.
19106 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19107 (unwrap_const_vec_duplicate): Likewise.
19108 (const_vec_series_p): Check for a non-duplicated vector encoding.
19109 Say that the function only returns true for integer vectors.
19110 * emit-rtl.c: Include rtx-vector-builder.h.
19111 (gen_const_vec_duplicate_1): Delete.
19112 (gen_const_vector): Call gen_const_vec_duplicate instead of
19113 gen_const_vec_duplicate_1.
19114 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19115 (gen_const_vec_duplicate): Use rtx_vector_builder.
19116 (gen_const_vec_series): Likewise.
19117 (gen_rtx_CONST_VECTOR): Likewise.
19118 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19119 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19120 Build a new vector rather than modifying a CONST_VECTOR in-place.
19121 (handle_special_swappables): Update call accordingly.
19122 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19123 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19124 Build a new vector rather than modifying a CONST_VECTOR in-place.
19125 (handle_special_swappables): Update call accordingly.
19127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19129 * simplify-rtx.c (simplify_const_binary_operation): Use
19130 CONST_VECTOR_ELT instead of XVECEXP.
19132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19134 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19135 the selector elements to be different from the data elements
19136 if the selector is a VECTOR_CST.
19137 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19138 ssizetype for the selector.
19140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19142 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19143 before testing each element individually.
19144 * tree-vect-generic.c (lower_vec_perm): Likewise.
19146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19148 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19149 * selftest-run-tests.c (selftest::run_tests): Call it.
19150 * vector-builder.h (vector_builder::operator ==): New function.
19151 (vector_builder::operator !=): Likewise.
19152 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19153 (vec_perm_indices::all_from_input_p): New function.
19154 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19155 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19156 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19157 instead of reading the VECTOR_CST directly. Detect whether both
19158 vector inputs are the same before constructing the vec_perm_indices,
19159 and update the number of inputs argument accordingly. Use the
19160 utility functions added above. Only construct sel2 if we need to.
19162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19164 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19165 the broadcast of the low byte.
19166 (expand_mult_highpart): Use an explicit encoding for the permutes.
19167 * optabs-query.c (can_mult_highpart_p): Likewise.
19168 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19169 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19170 (vectorizable_bswap): Likewise.
19171 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19172 explicit encoding for the power-of-2 permutes.
19173 (vect_permute_store_chain): Likewise.
19174 (vect_grouped_load_supported): Likewise.
19175 (vect_permute_load_chain): Likewise.
19177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19179 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19180 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19181 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19182 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19183 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19184 (vect_gen_perm_mask_any): Likewise.
19186 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19188 * int-vector-builder.h: New file.
19189 * vec-perm-indices.h: Include int-vector-builder.h.
19190 (vec_perm_indices): Redefine as an int_vector_builder.
19191 (auto_vec_perm_indices): Delete.
19192 (vec_perm_builder): Redefine as a stand-alone class.
19193 (vec_perm_indices::vec_perm_indices): New function.
19194 (vec_perm_indices::clamp): Likewise.
19195 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19196 (vec_perm_indices::new_vector): New function.
19197 (vec_perm_indices::new_expanded_vector): Update for new
19198 vec_perm_indices class.
19199 (vec_perm_indices::rotate_inputs): New function.
19200 (vec_perm_indices::all_in_range_p): Operate directly on the
19201 encoded form, without computing elided elements.
19202 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19203 encoding. Update for new vec_perm_indices class.
19204 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19205 the given vec_perm_builder.
19206 (expand_vec_perm_var): Update vec_perm_builder constructor.
19207 (expand_mult_highpart): Use vec_perm_builder instead of
19208 auto_vec_perm_indices.
19209 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19210 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19211 or double series encoding as appropriate.
19212 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19213 vec_perm_indices instead of auto_vec_perm_indices.
19214 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19215 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19216 (vect_permute_store_chain): Likewise.
19217 (vect_grouped_load_supported): Likewise.
19218 (vect_permute_load_chain): Likewise.
19219 (vect_shift_permute_load_chain): Likewise.
19220 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19221 (vect_transform_slp_perm_load): Likewise.
19222 (vect_schedule_slp_instance): Likewise.
19223 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19224 (vectorizable_mask_load_store): Likewise.
19225 (vectorizable_bswap): Likewise.
19226 (vectorizable_store): Likewise.
19227 (vectorizable_load): Likewise.
19228 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19229 vec_perm_indices instead of auto_vec_perm_indices. Use
19230 tree_to_vec_perm_builder to read the vector from a tree.
19231 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19232 vec_perm_builder instead of a vec_perm_indices.
19233 (have_whole_vector_shift): Use vec_perm_builder and
19234 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19235 truncation to calc_vec_perm_mask_for_shift.
19236 (vect_create_epilog_for_reduction): Likewise.
19237 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19238 from auto_vec_perm_indices to vec_perm_indices.
19239 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19240 instead of changing individual elements.
19241 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19242 the vector in d.perm.
19243 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19244 from auto_vec_perm_indices to vec_perm_indices.
19245 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19246 instead of changing individual elements.
19247 (arm_vectorize_vec_perm_const): Use new_vector to install
19248 the vector in d.perm.
19249 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19250 Update vec_perm_builder constructor.
19251 (rs6000_expand_interleave): Likewise.
19252 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19253 (rs6000_expand_interleave): Likewise.
19255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19257 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19258 to qimode could truncate the indices.
19259 * optabs.c (expand_vec_perm_var): Likewise.
19261 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19263 * Makefile.in (OBJS): Add vec-perm-indices.o.
19264 * vec-perm-indices.h: New file.
19265 * vec-perm-indices.c: Likewise.
19266 * target.h (vec_perm_indices): Replace with a forward class
19268 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19269 * optabs.h: Include vec-perm-indices.h.
19270 (expand_vec_perm): Delete.
19271 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19272 (expand_vec_perm_const): Declare.
19273 * target.def (vec_perm_const_ok): Replace with...
19274 (vec_perm_const): ...this new hook.
19275 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19276 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19277 * doc/tm.texi: Regenerate.
19278 * optabs.def (vec_perm_const): Delete.
19279 * doc/md.texi (vec_perm_const): Likewise.
19280 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19281 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19282 expand_vec_perm for constant permutation vectors. Assert that
19283 the mode of variable permutation vectors is the integer equivalent
19284 of the mode that is being permuted.
19285 * optabs-query.h (selector_fits_mode_p): Declare.
19286 * optabs-query.c: Include vec-perm-indices.h.
19287 (selector_fits_mode_p): New function.
19288 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19289 is defined, instead of checking whether the vec_perm_const_optab
19290 exists. Use targetm.vectorize.vec_perm_const instead of
19291 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19292 fit in the vector mode before using a variable permute.
19293 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19294 vec_perm_indices instead of an rtx.
19295 (expand_vec_perm): Replace with...
19296 (expand_vec_perm_const): ...this new function. Take the selector
19297 as a vec_perm_indices rather than an rtx. Also take the mode of
19298 the selector. Update call to shift_amt_for_vec_perm_mask.
19299 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19300 Use vec_perm_indices::new_expanded_vector to expand the original
19301 selector into bytes. Check whether the indices fit in the vector
19302 mode before using a variable permute.
19303 (expand_vec_perm_var): Make global.
19304 (expand_mult_highpart): Use expand_vec_perm_const.
19305 * fold-const.c: Includes vec-perm-indices.h.
19306 * tree-ssa-forwprop.c: Likewise.
19307 * tree-vect-data-refs.c: Likewise.
19308 * tree-vect-generic.c: Likewise.
19309 * tree-vect-loop.c: Likewise.
19310 * tree-vect-slp.c: Likewise.
19311 * tree-vect-stmts.c: Likewise.
19312 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19314 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19315 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19316 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19317 (aarch64_vectorize_vec_perm_const): ...this new function.
19318 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19319 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19320 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19321 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19322 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19323 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19324 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19326 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19327 check for NEON modes.
19328 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19329 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19330 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19331 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19333 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19334 the old VEC_PERM_CONST conditions.
19335 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19336 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19337 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19338 (ia64_vectorize_vec_perm_const_ok): Merge into...
19339 (ia64_vectorize_vec_perm_const): ...this new function.
19340 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19341 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19342 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19343 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19344 * config/mips/mips.c (mips_expand_vec_perm_const)
19345 (mips_vectorize_vec_perm_const_ok): Merge into...
19346 (mips_vectorize_vec_perm_const): ...this new function.
19347 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19348 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19349 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19350 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19351 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19352 (rs6000_expand_vec_perm_const): Delete.
19353 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19355 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19356 (altivec_expand_vec_perm_const_le): Take each operand individually.
19357 Operate on constant selectors rather than rtxes.
19358 (altivec_expand_vec_perm_const): Likewise. Update call to
19359 altivec_expand_vec_perm_const_le.
19360 (rs6000_expand_vec_perm_const): Delete.
19361 (rs6000_vectorize_vec_perm_const_ok): Delete.
19362 (rs6000_vectorize_vec_perm_const): New function.
19363 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19364 an element count and rtx array.
19365 (rs6000_expand_extract_even): Update call accordingly.
19366 (rs6000_expand_interleave): Likewise.
19367 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19368 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19369 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19370 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19371 (rs6000_expand_vec_perm_const): Delete.
19372 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19373 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19374 (altivec_expand_vec_perm_const_le): Take each operand individually.
19375 Operate on constant selectors rather than rtxes.
19376 (altivec_expand_vec_perm_const): Likewise. Update call to
19377 altivec_expand_vec_perm_const_le.
19378 (rs6000_expand_vec_perm_const): Delete.
19379 (rs6000_vectorize_vec_perm_const_ok): Delete.
19380 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19381 reference to the SPE evmerge intructions.
19382 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19383 an element count and rtx array.
19384 (rs6000_expand_extract_even): Update call accordingly.
19385 (rs6000_expand_interleave): Likewise.
19386 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19387 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19389 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19391 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19393 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19394 vector mode and that that mode matches the mode of the data
19396 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19397 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19398 directly using expand_vec_perm_1 when forcing selectors into
19400 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19402 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19404 * optabs-query.h (can_vec_perm_p): Delete.
19405 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19406 * optabs-query.c (can_vec_perm_p): Split into...
19407 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19408 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19409 particular selector is valid.
19410 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19411 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19412 (vect_grouped_load_supported): Likewise.
19413 (vect_shift_permute_load_chain): Likewise.
19414 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19415 (vect_transform_slp_perm_load): Likewise.
19416 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19417 (vectorizable_bswap): Likewise.
19418 (vect_gen_perm_mask_checked): Likewise.
19419 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19420 implementations of variable permutation vectors into account
19421 when deciding which selector to use.
19422 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19423 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19424 with a false third argument.
19425 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19426 to test whether the constant selector is valid and can_vec_perm_var_p
19427 to test whether a variable selector is valid.
19429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19431 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19432 * optabs-query.c (can_vec_perm_p): Likewise.
19433 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19434 instead of vec_perm_indices.
19435 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19436 (vect_gen_perm_mask_checked): Likewise,
19437 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19438 (vect_gen_perm_mask_checked): Likewise,
19440 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19442 * optabs-query.h (qimode_for_vec_perm): Declare.
19443 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19444 (qimode_for_vec_perm): ...this new function.
19445 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19447 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19449 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19450 does not have a conditional at the top.
19452 2018-01-02 Richard Biener <rguenther@suse.de>
19454 * ipa-inline.c (big_speedup_p): Fix expression.
19456 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19459 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19462 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19466 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19467 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19468 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19469 cond_taken_branch_cost 3->4.
19471 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19473 PR tree-optimization/83581
19474 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19475 TODO_cleanup_cfg if any changes have been made.
19477 PR middle-end/83608
19478 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19479 convert_modes if target mode has the right side, but different mode
19482 PR middle-end/83609
19483 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19484 last argument when extracting from CONCAT. If either from_real or
19485 from_imag is NULL, use expansion through memory. If result is not
19486 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19487 the parts directly to inner mode, if even that fails, use expansion
19490 PR middle-end/83623
19491 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19492 check for bswap in mode rather than HImode and use that in expand_unop
19495 Copyright (C) 2018 Free Software Foundation, Inc.
19497 Copying and distribution of this file, with or without modification,
19498 are permitted in any medium without royalty provided the copyright
19499 notice and this notice are preserved.