1 2018-07-12 Richard Biener <rguenther@suse.de>
3 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4 (vn_lookup_simplify_result): Remove recursion limit applied
6 (vn_nary_build_or_lookup_1): Adjust.
7 (try_to_simplify): Likewise.
8 * gimple-match-head.c (gimple_resimplify1): Instead apply one
10 (gimple_resimplify2): Likewise.
11 (gimple_resimplify3): Likewise.
12 (gimple_resimplify4): Likewise.
14 2018-07-11 Jakub Jelinek <jakub@redhat.com>
16 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
17 Use __mmask64 type instead of __mmask8 for __M argument.
18 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
19 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
21 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
22 __mmask16 for __M argument.
23 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
24 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
25 to __mmask16 instead of __mmask8.
26 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
27 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
28 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
29 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
30 instead of __mmask16 for __U argument.
31 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
32 __mmask16 instead of __mmask8 for __U argument.
33 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
35 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
37 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
39 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
41 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
43 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
45 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
47 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
49 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
51 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
53 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
54 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
55 return type as well as __M argument type and all casts from __mmask8
57 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
58 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
59 return type as well as __M argument type and all casts from __mmask8
61 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
62 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
63 return type as well as __M argument type and all casts from __mmask8
65 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
66 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
67 return type as well as __M argument type and all casts from __mmask8
69 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
70 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
73 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
75 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
76 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
79 2018-07-11 Paul Koning <ni1d@arrl.net>
81 * doc/md.texi (define_subst): Document how multiple occurrences of
82 the same argument in the replacement pattern are handled.
84 2018-07-11 Paul Koning <ni1d@arrl.net>
86 * doc/extend.texi (Common Variable Attributes): Move "mode" into
88 (Common Type Attributes): Add "mode" attribute.
90 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
92 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
93 stream DECL_ORIGINAL_TYPE.
94 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
95 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
96 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
97 Do not walk original type.
98 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
100 (write_ts_decl_non_common_tree_pointers): Do not stream
102 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
103 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
105 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
107 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
108 thread twice from the same starting edge.
110 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
112 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
113 * gimple.c (gimple_call_nonnull_result_p): ...here...
114 (gimple_call_nonnull_arg): ...and here.
115 * gimple.h (gimple_call_nonnull_result_p): New.
116 (gimple_call_nonnull_arg): New.
118 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
120 * config/arm/arm-cpus.in: Move information from fpu field of each
121 cpu definition to the isa field.
122 * config/arm/parsecpu.awk (fpu): Delete match rule.
123 (gen_comm_data): Don't add bits from the CPU's FPU entry.
125 2018-07-11 Richard Biener <rguenther@suse.de>
128 * dwarf2out.c (init_sections_and_labels): Use
129 output_asm_line_debug_info consistently.
130 (dwarf2out_early_finish): Likewise.
131 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
134 2018-07-11 Richard Biener <rguenther@suse.de>
136 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
137 Rework father_bb setting in a way to avoid propagating constants
138 multiple times on a loop body.
140 2018-07-10 Mark Wielaard <mark@klomp.org>
143 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
145 2018-07-10 Richard Biener <rguenther@suse.de>
147 * hash-map.h (hash_map::iterator::operator*): Return
148 references to key and value.
150 2018-07-10 Jakub Jelinek <jakub@redhat.com>
153 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
154 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
155 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
156 (gimplify_omp_for): For composite loops, move outer
157 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
158 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
159 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
160 TREE_LIST for both the original class iterator and the "last" helper
161 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
162 loop, remember has_decl_expr from outer composite loops for the
163 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
165 2018-07-09 Martin Sebor <msebor@redhat.com>
169 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
170 accessing implicitly initialized array elements.
171 * expr.c (string_constant): Handle string initializers of
172 character arrays within aggregates.
173 * gimple-fold.c (fold_array_ctor_reference): Add argument.
174 Store element offset. As a special case, handle zero size.
175 (fold_nonarray_ctor_reference): Same.
176 (fold_ctor_reference): Add argument. Store subobject offset.
177 * gimple-fold.h (fold_ctor_reference): Add argument.
179 2018-07-09 Paul Koning <ni1d@arrl.net>
181 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
182 (pdp11_insn_cost): New function.
183 (pdp11_md_asm_adjust): New function.
184 (TARGET_INVALID_WITHIN_DOLOOP): Define.
185 (pdp11_rtx_costs): Update to match machine better.
186 (output_addr_const_pdp11): Correct format mismatch warnings.
187 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
188 * config/pdp11/pdp11.md: General change to add base_cost and/or
189 length attributes for use by new pdp11_insn_cost function.
190 (MIN_BRANCH): Correct definition.
192 (doloop_end): Use standard pattern name for looping pattern.
193 (doloop_end_nocc): New.
194 (movsf): Add another constraint alternative.
195 (zero_extendqihi2): Add constraint alternatives for not in place
197 (zero_extendhisi2): Remove.
198 (shift patterns): Add CC handling variants.
202 (define_peephole2): New peephole to recognize mov that sets CC for
205 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
207 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
208 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
211 2018-07-09 Jakub Jelinek <jakub@redhat.com>
214 * real.c (real_nextafter): Return true if result is denormal.
216 2018-07-09 Martin Liska <mliska@suse.cz>
218 * common.opt: Add back wrongly removed attribute.
220 2018-07-09 Richard Biener <rguenther@suse.de>
223 * dwarf2out.c (gen_block_die): For an early generated DIE
224 always output high/low PC attributes.
226 2018-07-09 Tom de Vries <tdevries@suse.de>
228 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
229 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
230 onto VAR_DECL with abstract origin.
232 2018-07-07 Jim Wilson <jimw@sifive.com>
234 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
236 2018-07-07 Tom de Vries <tdevries@suse.de>
238 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
240 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
242 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
244 (combine_bound): Use wide-int overflow calculation instead of
246 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
248 * fold-const.c (int_const_binop_2): Same.
249 (extract_muldiv_1): Same.
250 (fold_div_compare): Same.
251 (fold_abs_const): Same.
253 * poly-int.h (add): Same.
257 * predict.c (predict_iv_comparison): Same.
258 * profile-count.c (slow_safe_scale_64bit): Same.
259 * simplify-rtx.c (simplify_const_binary_operation): Same.
260 * tree-chrec.c (tree_fold_binomial): Same.
261 * tree-data-ref.c (split_constant_offset_1): Same.
262 * tree-if-conv.c (idx_within_array_bound): Same.
263 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
264 * tree-ssa-phiopt.c (minmax_replacement): Same.
265 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
266 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
267 * vr-values.c (vr_values::adjust_range_with_scev): Same.
268 * wide-int.cc (wi::add_large): Same.
269 (wi::mul_internal): Same.
270 (wi::sub_large): Same.
271 (wi::divmod_internal): Same.
272 * wide-int.h: Change overflow type to overflow_type for neg, add,
273 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
274 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
275 mul_internal, divmod_internal.
276 (overflow_type): New enum.
277 (accumulate_overflow): New.
279 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
281 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
282 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
284 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
286 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
287 argument is checked for zero before entering loop, avoid checking again.
289 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
291 * gimplify.h (generic_expr_could_trap_p): Set as global function.
292 * gimplify.h (generic_expr_could_trap_p): Likwise.
293 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
295 2018-07-06 Jakub Jelinek <jakub@redhat.com>
297 PR tree-optimization/86401
298 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
299 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
300 (fold_bit_and_mask): ... here. New helper function for match.pd.
301 * fold-const.h (fold_bit_and_mask): Declare.
302 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
304 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
307 * target.def (translate_mode_attribute): New hook.
308 * targhooks.h (default_translate_mode_attribute): Declare.
309 * targhooks.c (default_translate_mode_attribute): New function.
310 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
311 * doc/tm.texi: Regenerate.
312 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
313 (rs6000_translate_mode_attribute): New function.
315 2018-07-06 Paul Koning <ni1d@arrl.net>
317 * doc/md.texi (define_split): Document DONE and FAIL.
318 (define_peephole2): Ditto.
320 2018-07-05 Jeff Law <law@redhat.com>
322 PR tree-optimization/86010
323 * tree-ssa-dse.c (compute_trims): More aggressively trim at
324 both the head and tail of mem* and str* calls.
326 2018-07-05 Jim Wilson <jimw@sifive.com>
328 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
330 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
332 * config/aarch64/aarch64-simd.md: correct flags text for
333 MIN_EXPR replacement.
335 2018-07-05 James Clarke <jrtc27@jrtc27.com>
337 * configure: Regenerated.
339 2018-07-05 Carl Love <cel@us.ibm.com>
341 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
342 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
343 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
344 VSX_BUILTIN_DOUBLEL_V4SF.
346 2018-07-05 Martin Sebor <msebor@redhat.com>
349 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
350 than its domain to compute its the upper bound of a char array.
352 2018-07-05 Nathan Sidwell <nathan@acm.org>
354 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
355 * doc/cpp.texi: Update comment.
356 * doc/tm.texi: Rebuilt.
357 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
358 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
359 * doc/extend.texi (Backwards Compatibility): Clarify it is system
360 headers affected by extern "C".
361 * system.h: Poison NO_IMPLICIT_EXTERN_C.
362 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
363 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
364 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
365 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
366 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
367 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
368 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
369 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
370 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
371 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
372 config/sparc/sp64-elf.h, config/spu/spu.h,
373 config/stormy16/stormy16.h, config/v850/v850.h,
374 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
375 define NO_IMPLICIT_EXTERN_C.
376 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
378 2018-07-05 Tamar Christina <tamar.christina@arm.com>
381 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
382 instead of GET_MODE_SIZE when comparing Units.
384 2018-07-05 Tamar Christina <tamar.christina@arm.com>
387 * rtlanal.c (set_noop_p): Constrain on mode change,
388 include hard-reg-set.h
390 2018-07-05 Tamar Christina <tamar.christina@arm.com>
392 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
394 2018-07-05 Jakub Jelinek <jakub@redhat.com>
397 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
400 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
402 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
404 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
407 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
409 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
411 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
414 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
416 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
418 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
421 2018-07-04 Martin Liska <mliska@suse.cz>
423 * tree-switch-conversion.c: Define
424 max_ratio_for_speed and max_ratio_for_size constants.
426 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
427 Martin Liska <mliska@suse.cz>
432 * common.opt: Rename align options with 'str_' prefix.
433 * common/config/i386/i386-common.c (set_malign_value): New
435 (ix86_handle_option): Use it to set -falign-* options/
436 * config/aarch64/aarch64-protos.h (struct tune_params): Change
437 type from int to string.
438 * config/aarch64/aarch64.c: Update default values from int
440 * config/alpha/alpha.c (alpha_override_options_after_change):
442 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
443 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
444 max skip conditionally.
445 * config/i386/freebsd.h (SUBALIGN_LOG): New.
446 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
447 max skip conditionally.
448 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
449 max skip conditionally.
450 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
451 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
452 max skip conditionally.
453 * config/i386/i386.c (struct ptt): Change type from int to
455 (ix86_default_align): Set default values.
456 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
457 max skip conditionally.
458 * config/i386/iamcu.h (SUBALIGN_LOG): New.
459 (ASM_OUTPUT_MAX_SKIP_ALIGN):
460 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
461 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
462 max skip conditionally.
463 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
464 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
465 * config/i386/x86-64.h (SUBALIGN_LOG): New.
466 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
467 max skip conditionally.
468 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
469 * config/ia64/ia64.c (ia64_option_override): Set default values
470 for alignment options.
471 * config/m68k/m68k.c: Handle new str_align_* options.
472 * config/mips/mips.c (mips_set_compression_mode): Change
474 (mips_option_override): Set default values for options.
475 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
477 * config/rs6000/rs6000.c (rs6000_option_override_internal):
479 * config/rx/rx.c (rx_option_override): Likewise.
480 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
481 (LABEL_ALIGN): Use align_labels_log.
482 (LOOP_ALIGN): Use align_loops_align.
483 * config/s390/s390.c (s390_asm_output_function_label): Use new
485 * config/sh/sh.c (sh_override_options_after_change):
486 Change type of constants.
487 * config/spu/spu.c (spu_sched_init): Likewise.
488 * config/sparc/sparc.c (sparc_option_override): Set default
490 * config/visium/visium.c (visium_option_override): Likewise.
491 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
492 emit p2align format with last argument if it's not needed.
493 * doc/invoke.texi: Document extended format of -falign-*.
494 * final.c: Use align_labels alignment.
495 * flags.h (struct target_flag_state): Change type to use
497 (struct align_flags_tuple): New.
498 (struct align_flags): Likewise.
499 (align_loops_log): Redefine macro to use new types.
500 (align_loops_max_skip): Redefine macro to use new types.
501 (align_jumps_log): Redefine macro to use new types.
502 (align_jumps_max_skip): Redefine macro to use new types.
503 (align_labels_log): Redefine macro to use new types.
504 (align_labels_max_skip): Redefine macro to use new types.
505 (align_functions_log): Redefine macro to use new types.
506 (align_loops): Redefine macro to use new types.
507 (align_jumps): Redefine macro to use new types.
508 (align_labels): Redefine macro to use new types.
509 (align_functions): Redefine macro to use new types.
510 (align_functions_max_skip): Redefine macro to use new types.
511 (align_loops_value): New macro.
512 (align_jumps_value): New macro.
513 (align_labels_value): New macro.
514 (align_functions_value): New macro.
515 * function.c (invoke_set_current_function_hook): Propagate
516 alignment values from flags to global variables default in
518 * ipa-icf.c (sem_function::equals_wpa): Use
519 cl_optimization_option_eq instead of memcmp.
520 * lto-streamer.h (cl_optimization_stream_out): Support streaming
522 (cl_optimization_stream_in): Likewise.
523 * optc-save-gen.awk: Support strings in cl_optimization.
524 * opth-gen.awk: Likewise.
525 * opts.c (finish_options): Remove error checking of invalid
527 (MAX_CODE_ALIGN): Remove.
528 (MAX_CODE_ALIGN_VALUE): Likewise.
529 (parse_and_check_align_values): New function.
530 (check_alignment_argument): Likewise.
531 (common_handle_option): Use check_alignment_argument.
532 * opts.h (parse_and_check_align_values): Declare.
533 * toplev.c (init_alignments): Remove.
534 (read_log_maxskip): New.
535 (parse_N_M): Likewise.
536 (parse_alignment_opts): Likewise.
537 (backend_init_target): Remove usage of init_alignments.
538 * toplev.h (parse_alignment_opts): Declare.
539 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
541 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
542 * tree.c (cl_option_hasher::equal): New.
543 * varasm.c: Use new global macros.
545 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
547 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
548 Use a simpler align directive also if MAXSKIP = ALIGN-1.
549 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
550 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
551 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
552 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
553 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
554 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
555 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
556 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
557 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
558 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
559 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
561 2018-07-04 Martin Liska <mliska@suse.cz>
562 Jonathan Wakely <jwakely@redhat.com>
564 * coverage.c: Use correct type.
565 * doc/invoke.texi: Language correction.
567 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
570 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
571 ENDBRANCH for non-tail call which may return via indirect branch.
572 * doc/extend.texi: Document indirect_return attribute.
574 2018-07-03 Martin Sebor <msebor@redhat.com>
576 PR tree-optimization/86274
577 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
579 (format_floating): Correct handling of infinities and NaNs.
581 2018-07-03 Martin Sebor <msebor@redhat.com>
583 * print-tree.c (print_real_cst): New function.
584 (print_node_brief): Call it.
587 2018-07-03 Jeff Law <law@redhat.com>
589 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
590 into a single pattern.
592 * config/h8300/h8300.md (ors code_iterator): New.
593 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
594 a single pattern and single splitter.
595 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
596 (iorqi3_1, xorqi3_1): Likewise.
597 (iorqi3, xorqi3 expanders): Similarly.
599 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
600 (movmd_internal) into a single pattern using the P mode iterator.
601 (movmd splitters): Similarly.
602 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
603 (movsd splitters): Similarly.
605 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
606 ADDB, ADDW and ADDL into a single ADD attribute which selects the
607 right table based on the size of the operand.
608 * config/h8300/h8300.md (length_table): Corresponding changes. All
609 references to "addb", "addw" and "addl" changed to "add".
610 (btst patterns): Merge two variants into a single pattern.
611 (tstqi, tsthi): Likewise.
612 (addhi3_incdec, addsi3_incdec): Likewise.
613 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
614 (mulhi3, mulsi3): Likewise.
615 (udivhi3, udivsi3): Likewise.
616 (divhi3, divsi3): Likewise.
617 (andorqi3, andorhi3, andorsi3): Likewise.
619 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
622 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
623 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
625 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
627 PR tree-optimization/85694
628 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
629 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
631 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
632 (<u>avg<mode>3_ceil): New patterns.
634 2018-07-03 David Malcolm <dmalcolm@redhat.com>
636 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
637 scan-tree-dump directive.
638 * gcc.dg/vect/slp-perm-2.c: Likewise.
639 * gcc.dg/vect/slp-perm-3.c: Likewise.
640 * gcc.dg/vect/slp-perm-5.c: Likewise.
641 * gcc.dg/vect/slp-perm-6.c: Likewise.
642 * gcc.dg/vect/slp-perm-7.c: Likewise.
643 * gcc.dg/vect/slp-perm-8.c: Likewise.
645 2018-07-03 Marek Polacek <polacek@redhat.com>
648 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
650 2018-07-03 Richard Biener <rguenther@suse.de>
653 * tree-ssa-structalias.c (find_func_clobbers): Properly
654 handle indirect calls.
656 2018-07-03 Jeff Law <law@redhat.com>
658 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
659 (shifts): New code iterator.
660 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
661 expander. Fix HImode handling on H8/SX.
662 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
663 (subqi3, subhi3, subsi3 expanders): Likewise.
664 (andqi3, andhi3, andsi3 expanders): Likewise.
665 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
666 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
667 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
668 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
669 (zero_extendqihi2, zero_extendqisi2): Likewise.
670 (extendqihi2, extendqisi2): Likewise.
671 (rotlqi3, rotlhi3, rotlsi3): Likewise.
672 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
673 (rotlqi3_1, rotlhi3_1): Likewise.
674 (logicalhi3_sn, logicalsi3_sn): Likewise.
675 (logicalhi3, logicalsi3): Likewise.
677 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
679 * tree-vect-patterns.c (vect_recog_rotate_pattern)
680 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
681 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
682 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
683 type to append_pattern_def_seq instead of creating a stmt_vec_info
685 (build_mask_conversion): Likewise. Remove vinfo argument.
686 (vect_add_conversion_to_patterm): Likewise, renaming to...
687 (vect_add_conversion_to_pattern): ...this.
688 (vect_recog_mask_conversion_pattern): Update call to
689 build_mask_conversion. Pass the vector type to
690 append_pattern_def_seq here too.
691 (vect_recog_gather_scatter_pattern): Update call to
692 vect_add_conversion_to_pattern.
694 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
696 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
697 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
698 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
699 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
700 STMT_VINFO_PATTERN_DEF_SEQ to null here.
701 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
702 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
703 append_pattern_def_seq instead of new_pattern_def_seq.
704 (vect_recog_divmod_pattern): Do both of the above.
705 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
708 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
710 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
711 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
712 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
713 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
714 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
715 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
716 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
717 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
718 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
719 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
720 parameter with a single stmt_vec_info.
721 (vect_recog_func_ptr): Likewise.
722 (vect_recog_gather_scatter_pattern): Likewise, folding in...
723 (vect_try_gather_scatter_pattern): ...this.
724 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
725 the stmt_vec_info of the statement to be matched. Don't clear
726 STMT_VINFO_RELATED_STMT.
727 (vect_pattern_recog): Update call accordingly.
729 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
731 PR tree-optimization/85694
732 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
733 (uavgM3_ceil): Document new optabs.
734 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
735 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
737 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
738 (savg_ceil_optab): New optabs.
739 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
740 (vect_vect_recog_func_ptrs): Add it.
741 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
742 constant directly from the associated lhs.
744 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
746 * tree-vect-patterns.c (vect_split_statement): New function.
747 (vect_convert_input): Use it to try to split an existing cast.
749 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
751 * poly-int.h (print_hex): New function.
752 * dumpfile.h (dump_dec, dump_hex): Declare.
753 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
754 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
755 min_input_precision, operation_precision and operation_sign.
756 * tree-vect-patterns.c (vect_get_range_info): New function.
757 (vect_same_loop_or_bb_p, vect_single_imm_use)
758 (vect_operation_fits_smaller_type): Delete.
759 (vect_look_through_possible_promotion): Add an optional
760 single_use_p parameter.
761 (vect_recog_over_widening_pattern): Rewrite to use new
762 stmt_vec_info infomration. Handle one operation at a time.
763 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
764 (vect_truncatable_operation_p, vect_set_operation_type)
765 (vect_set_min_input_precision): New functions.
766 (vect_determine_min_output_precision_1): Likewise.
767 (vect_determine_min_output_precision): Likewise.
768 (vect_determine_precisions_from_range): Likewise.
769 (vect_determine_precisions_from_users): Likewise.
770 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
771 (vect_vect_recog_func_ptrs): Put over_widening first.
773 (vect_pattern_recog): Call vect_determine_precisions.
775 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
777 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
778 statements that have been replaced by further pattern statements.
779 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
781 2018-07-03 Richard Biener <rguenther@suse.de>
783 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
784 always set *dt. Dump vectype in vectype overload.
785 * dumpfile.h (dump_gimple_expr): New function.
786 (dump_gimple_expr_loc): Likewise.
787 * dumpfile.c (dump_gimple_expr): New function.
788 (dump_gimple_expr_loc): Likewise.
790 2018-07-02 Jeff Law <law@redhat.com>
792 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
793 the H8/300, H8/300H and H8/S variants into a single pattern.
794 (movhi_h8300, movqi_h8300hs): Similarly.
795 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
796 (QHI mode iterator): New.
798 * config/h8300/h8300.md: Remove trailing whitespace.
800 2018-07-02 Jim Wilson <jimw@sifive.com>
802 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
803 instead of emit_insn for interrupt returns.
804 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
805 (riscv_sret, riscv_uret): Likewise.
807 2018-07-02 David Malcolm <dmalcolm@redhat.com>
809 * pretty-print.c (selftest::test_pp_format): Move save and restore
810 of quotes to class auto_fix_quotes, and add an instance.
811 * selftest.c: Include "intl.h".
812 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
813 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
814 * selftest.h (selftest::auto_fix_quotes): New class.
816 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
818 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
819 (aarch64_sve_prepare_conditional_op): Remove.
820 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
821 Allow aarch64_simd_reg_or_zero as select operand; remove
822 the aarch64_sve_prepare_conditional_op call.
823 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
824 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
825 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
826 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
827 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
828 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
829 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
830 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
831 and a splitters to match all of the *_any patterns.
832 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
834 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
835 (SVE_COND_FP_BINARY_REV): Remove.
836 (sve_int_op_rev, sve_fp_op_rev): New.
837 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
838 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
839 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
840 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
841 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
842 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
843 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
844 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
845 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
847 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
848 Remove match_dup 1 from the inner unspec.
849 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
851 * config/aarch64/aarch64.md (movprfx): New attr.
852 (length): Default movprfx to 8.
853 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
854 (*madd<SVE_I>, *msub<SVE_I): Likewise.
855 (*<su>mul<SVE_I>3_highpart): Likewise.
856 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
857 (*v<ASHIFT><SVE_I>3): Likewise.
858 (*<su><MAXMIN><SVE_I>3): Likewise.
859 (*<su><MAXMIN><SVE_F>3): Likewise.
860 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
861 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
862 (*div<SVE_F>4): Likewise.
864 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
866 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
869 2018-07-02 Richard Biener <rguenther@suse.de>
871 PR tree-optimization/86363
872 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
873 memset argument refers to a non-variable address.
875 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
877 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
878 {PLUS,MINUS}_EXPR code to...
879 (adjust_symbolic_bound): ...here,
880 (combine_bound): ...here,
881 (set_value_range_with_overflow): ...and here.
883 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
885 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
887 (extract_range_from_abs_expr): ...here.
889 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
891 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
892 -fno-omit-frame-pointer when not optimizing.
894 2018-07-02 Martin Liska <mliska@suse.cz>
897 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
898 (propagate_nothrow): Likewise.
900 2018-07-02 Martin Liska <mliska@suse.cz>
903 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
905 2018-07-02 David Malcolm <dmalcolm@redhat.com>
907 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
908 function in r262149, changing "loc" param from source_location to
909 const dump_location_t &.
910 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
911 declaration, as above.
913 2018-07-01 Paul Koning <ni1d@arrl.net>
915 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
916 -munit-asm, -mgnu-asm, -mdec-asm.
917 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
918 (pdp11_output_labelref): New.
919 (pdp11_output_def): New.
920 (pdp11_output_addr_vec_elt): New.
921 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
922 %# and %@ format codes.
923 (pdp11_option_override): New.
924 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
925 (pdp11_output_ident): New.
926 (pdp11_asm_named_section): New.
927 (pdp11_asm_init_sections): New.
928 (pdp11_file_start): New.
929 (pdp11_file_end): New.
930 (output_ascii): Use .ascii/.asciz for -mdec-asm.
931 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
932 %o, like %c but octal.
933 (pdp11_option_override): New.
934 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
936 (DATA_SECTION_ASM_OP): Ditto.
937 (READONLY_DATA_SECTION_ASM_OP): New.
938 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
939 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
940 (ASM_OUTPUT_LABELREF): Ditto.
941 (ASM_OUTPUT_DEF): Ditto.
942 (ASM_OUTPUT_EXTERNAL): New.
943 (ASM_OUTPUT_SOURCE_FILENAME): New.
944 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
945 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
946 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
947 %# and %@ format codes.
948 * config/pdp11/pdp11.opt (mgnu-asm): New.
949 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
950 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
951 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
953 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
955 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
956 dereferencing path[] beyond its length.
958 (debug_all_paths): New.
959 (rewire_first_differing_edge): New.
960 (adjust_paths_after_duplication): New.
961 (duplicate_thread_path): Call adjust_paths_after_duplication.
963 (thread_through_all_blocks): Add new argument to
964 duplicate_thread_path.
966 2018-06-30 Jim Wilson <jimw@sifive.com>
968 * config/riscv/predicates.md (p2m1_shift_operand): New.
969 (high_mask_shift_operand): New.
970 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
971 pattern using p2m1_shift_operand.
972 (lshsi3_zero_extend_3+2): New combiner pattern using
973 high_mask_shift_operand.
975 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
977 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
979 (vect_recog_rotate_pattern): ...here.
980 (vect_convert_input): Try to insert casts of invariants in the
982 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
983 preheader to be empty.
985 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
987 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
988 vector type. If given, install it in the new statement's
990 (vect_element_precision): New function.
991 (vect_unpromoted_value): New struct.
992 (vect_unpromoted_value::vect_unpromoted_value): New function.
993 (vect_unpromoted_value::set_op): Likewise.
994 (vect_look_through_possible_promotion): Likewise.
995 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
996 (vect_widened_op_tree, vect_convert_input): Likewise.
997 (vect_convert_inputs, vect_convert_output): Likewise.
998 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
999 to handle the optional cast of the multiplication result and
1000 vect_widened_op_tree to detect the widened multiplication itself.
1001 Do not require the input and output of promotion casts to have
1002 the same sign, but base the signedness of the operation on the
1003 input rather than the result. If the pattern includes two
1004 promotions, check that those promotions have the same sign.
1005 Do not restrict the MULT_EXPR handling to a double-width result;
1006 handle quadruple-width results and wider. Use vect_convert_inputs
1007 to convert the inputs to the common type.
1008 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1009 to handle the optional cast of the ABS result. Also allow a sign
1010 change or a sign extension between the ABS and MINUS.
1011 Use vect_widened_op_tree to detect the widened subtraction and use
1012 vect_convert_inputs to convert the inputs to the common type.
1013 (vect_handle_widen_op_by_const): Delete.
1014 (vect_recog_widen_op_pattern): New function.
1015 (vect_recog_widen_mult_pattern): Use it.
1016 (vect_recog_widen_shift_pattern): Likewise.
1017 (vect_recog_widen_sum_pattern): Use
1018 vect_look_through_possible_promotion to handle the promoted
1021 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1023 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1024 the containing gimple_seq *.
1025 * gimple-iterator.h (gsi_for_stmt): Declare it.
1026 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1027 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1028 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1029 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1030 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1032 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1034 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1035 statement being replaced is part of an existing pattern
1036 definition sequence, inserting the new pattern statements before
1038 (vect_pattern_recog_1): Don't return a bool. If the statement
1039 is already part of a pattern, instead apply pattern matching
1040 to the pattern definition statements. Don't clear the
1041 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1042 (vect_pattern_recog): Don't break after the first match;
1043 continue processing the pattern definition statements instead.
1044 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1046 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1048 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1049 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1050 (vect_recog_widen_sum_pattern): Use it.
1052 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1054 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1055 phi is not a pattern statement and has not been replaced by
1056 a pattern statement.
1057 * tree-vect-patterns.c (type_conversion_p): Don't check
1058 STMT_VINFO_IN_PATTERN_P.
1059 (vect_recog_vector_vector_shift_pattern): Likewise.
1060 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1061 the pattern statement rather than the original statement; check
1062 directly for a WIDEN_MULT_EXPR here.
1063 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1064 vect_is_simple_use to return the pattern statement rather
1065 than the original statement; use is_pattern_stmt_p to check
1066 for such a pattern statement.
1067 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1068 to return the pattern statement rather than the original statement;
1069 don't do the same transformation here.
1070 (vect_is_simple_use): If the defining statement has been replaced
1071 by a pattern statement, return the pattern statement instead.
1072 Remove the corresponding (local) transformation from the vectype
1075 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1077 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1078 end and default to null.
1079 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1080 (vectorizable_reduction): Update calls accordingly, dropping the
1081 gimple ** argument if the passed-back statement isn't needed.
1082 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1083 (vect_recog_rotate_pattern): Likewise.
1084 (vect_recog_mask_conversion_pattern): Likewise.
1085 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1086 (vect_mask_constant_operand_p): Likewise.
1087 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1088 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1089 (get_group_load_store_type, get_load_store_type): Likewise.
1090 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1091 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1092 (vectorizable_conversion, vectorizable_assignment): Likewise.
1093 (vectorizable_shift, vectorizable_operation): Likewise.
1094 (vectorizable_store, vect_is_simple_cond): Likewise.
1095 (vectorizable_condition, vectorizable_comparison): Likewise.
1096 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1097 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1098 and move it to the end. Cope with null def_stmt_outs.
1100 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1102 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1104 2018-06-29 Jeff Law <law@redhat.com>
1106 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1107 displacements for TARGET_V850E2V3 and newer.
1108 (TARGET_LRA_P): Remove. Defaults to LRA now.
1109 * config/v850/v850.md (sign23byte_load): Remove.
1110 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1111 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1113 2018-06-29 Martin Liska <mliska@suse.cz>
1116 * coverage.c (coverage_init): Mangle full path name.
1117 * doc/invoke.texi: Document the change.
1118 * gcov-io.c (mangle_path): New.
1119 * gcov-io.h (mangle_path): Likewise.
1120 * gcov.c (mangle_name): Use mangle_path for path mangling.
1122 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1124 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1125 if starting source register is not even.
1127 2018-06-29 Martin Liska <mliska@suse.cz>
1129 PR tree-optimization/86263
1130 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1131 Make edge redirection.
1133 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1135 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1136 (dump_scope_depth): New variable.
1137 (get_dump_scope_depth): New function.
1138 (dump_begin_scope): New function.
1139 (dump_end_scope): New function.
1140 * dumpfile.h (get_dump_scope_depth): New declaration.
1141 (dump_begin_scope): New declaration.
1142 (dump_end_scope): New declaration.
1143 (class auto_dump_scope): New class.
1144 (AUTO_DUMP_SCOPE): New macro.
1145 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1148 2018-06-29 Richard Biener <rguenther@suse.de>
1150 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1151 compute_all_dependences succeeds.
1152 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1153 exceed --param loop-max-datarefs-for-datadeps.
1155 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1157 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1159 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1162 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1163 alternative 0 in preferred_for_speed attribute.
1165 2018-06-28 Paul Koning <ni1d@arrl.net>
1167 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1168 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1169 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1170 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1171 for shift insn patterns.
1173 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1175 * cgraph.c (cgraph_node::get_body): Replace assignments to
1176 "dump_file" with calls to set_dump_file.
1177 * dumpfile.c (alt_dump_file): Make static, and group with...
1178 (alt_flags): ...this definition.
1179 (dumps_are_enabled): New variable.
1180 (refresh_dumps_are_enabled): New function.
1181 (set_dump_file): New function.
1182 (set_alt_dump_file): New function.
1183 (gcc::dump_manager::dump_start): Replace assignments to
1184 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1186 (gcc::dump_manager::dump_finish): Likewise.
1187 * dumpfile.h (alt_dump_file): Delete decl.
1188 (dumps_are_enabled): New variable decl.
1189 (set_dump_file): New function decl.
1190 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1192 * tree-nested.c (lower_nested_functions): Replace assignments to
1193 "dump_file" with calls to set_dump_file.
1195 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1197 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1198 goto_locus of each outgoing edge of each basic block.
1200 2018-06-28 Richard Biener <rguenther@suse.de>
1202 * dwarf2out.c (decl_scope_table): Remove.
1203 (push_decl_scope): Likewise.
1204 (pop_decl_scope): Likewise.
1205 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1206 (gen_struct_or_union_type_die): Likewise.
1207 (gen_tagged_type_die): Likewise.
1208 (dwarf2out_init): Do not initialize decl_scope_table.
1209 (dwarf2out_c_finalize): Do not free it.
1211 2018-06-28 Richard Biener <rguenther@suse.de>
1213 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1214 deciding whether to not re-use a DIE.
1216 2018-06-28 Richard Biener <rguenther@suse.de>
1218 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1219 DW_AT_abstract_origin attribute.
1221 2018-06-28 Martin Liska <mliska@suse.cz>
1223 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1224 Use newly introduced constants.
1225 * tree-switch-conversion.h (struct jump_table_cluster):
1226 Define max_ratio_for_size and max_ratio_for_speed.
1228 2018-06-28 Martin Liska <mliska@suse.cz>
1230 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1231 Add new checking assert to catch invalid state.
1232 (jump_table_cluster::can_be_handled): Handle single case
1234 (jump_table_cluster::is_beneficial): Bail out for such case.
1235 (bit_test_cluster::find_bit_tests):
1236 Add new checking assert to catch invalid state.
1237 (bit_test_cluster::can_be_handled): Handle single case
1239 (bit_test_cluster::is_beneficial): Bail out for such case.
1240 (switch_decision_tree::analyze_switch_statement):
1243 2018-06-28 Martin Liska <mliska@suse.cz>
1245 * common.opt: Introduce -completion option.
1246 * gcc.c (driver_handle_option): Handle it.
1247 (driver::main): Print completions if completion
1249 * opt-suggestions.c (option_proposer::get_completions):
1251 (option_proposer::suggest_completion): Likewise.
1252 (option_proposer::find_param_completions): Likewise.
1253 (verify_autocompletions): Likewise.
1254 (test_completion_valid_options): Likewise.
1255 (test_completion_valid_params): Likewise.
1256 (in_completion_p): Likewise.
1257 (empty_completion_p): Likewise.
1258 (test_completion_partial_match): Likewise.
1259 (test_completion_garbage): Likewise.
1260 (opt_proposer_c_tests): Likewise.
1261 * opt-suggestions.h: Declare new functions.
1262 * opts.c (common_handle_option): Handle OPT__completion_.
1263 * selftest-run-tests.c (selftest::run_tests): Add
1264 opt_proposer_c_tests.
1265 * selftest.c (assert_str_startswith): New.
1266 * selftest.h (assert_str_startswith): Likewise.
1267 (opt_proposer_c_tests): New.
1268 (ASSERT_STR_STARTSWITH): Likewise.
1270 2018-06-28 Martin Liska <mliska@suse.cz>
1272 * Makefile.in: Add opt-suggestions.o.
1273 * gcc-main.c: Include opt-suggestions.h.
1274 * gcc.c (driver::driver): Likewise.
1275 (driver::~driver): Remove m_option_suggestions.
1276 (driver::build_option_suggestions): Moved to option_proposer.
1277 (driver::suggest_option): Likewise.
1278 (driver::handle_unrecognized_options): Use option_proposer.
1279 * gcc.h (class driver): Add new memver m_option_proposer.
1280 * opt-suggestions.c: New file.
1281 * opt-suggestions.h: New file.
1283 2018-06-28 Martin Liska <mliska@suse.cz>
1285 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1286 (auto_string_vec::~auto_string_vec): Likewise.
1288 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1290 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1291 prevent_decl_creation_for_types fields up and add reset_location field.
1292 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1293 statement if id->reset_location is true.
1294 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1295 id->reset_location is true.
1296 (copy_phis_for_bb): Force input_location on the arguments if
1297 id->reset_location is true.
1298 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1299 is set on the function to be inlined.
1301 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1303 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1305 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1307 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1308 registers for Pmode.
1309 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1310 hard registers for the clobbered pseudo.
1312 2018-06-27 Paul Koning <ni1d@arrl.net>
1314 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1315 mutually exclusive options.
1316 * config/pdp11/constraints.md (h): New constraint.
1317 (O): Update definition to match shift code generation.
1318 (D): New constraint.
1319 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1321 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1323 (output_jump): Change arguments.
1324 (pdp11_fixed_cc_regs): New function.
1325 (pdp11_cc_mode): Ditto.
1326 (pdp11_expand_shift): Ditto.
1327 (pdp11_assemble_shift): Ditto.
1328 (pdp11_small_shift): Ditto.
1329 (pdp11_branch_cost): Remove.
1330 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1332 (pdp11_register_move_cost): Update for CC registers.
1333 (pdp11_rtx_costs): Add case for LSHIFTRT.
1334 (pdp11_output_jump): Add CCNZ mode conditional branches.
1335 (notice_update_cc_on_set): Remove.
1336 (pdp11_cc_mode): New function.
1337 (simple_memory_operand): Correct pre/post decrement case.
1338 (no_side_effect_operand): New function.
1339 (pdp11_regno_reg_class): Add CC_REGS class.
1340 (pdp11_fixed_cc_regs): New function.
1341 (pdp11_small_shift): New function.
1342 (pdp11_expand_shift): New function to expand shift insns.
1343 (pdp11_assemble_shift): New function to output shifts.
1344 (pdp11_branch_cost): Remove.
1345 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1346 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1347 (WCHAR_TYPE): Ditto.
1348 (PTRDIFF_TYPE): Ditto.
1349 (ADJUST_INSN_LENGTH): New macro.
1350 (FIXED_REGISTERS): Add CC registers.
1351 (CALL_USED_REGISTERS): Ditto.
1353 (REG_CLASS_NAMES): Ditto.
1354 (REG_CLASS_CONTENTS): Ditto.
1355 (SELECT_CC_MODE): Use new function.
1356 (TARGET_FLAGS_REGNUM): New macro.
1357 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1358 (cc0_reg_rtx): Remove.
1359 (CC_STATUS_MDEP): Remove.
1360 (CC_STATUS_MDEFP_INIT): Remove.
1361 (CC_IN_FPU): Remove.
1362 (NOTICE_UPDATE_CC): Remove.
1363 (REGISTER_NAMES): Add CC registers.
1364 (BRANCH_COST): Change to constant 1.
1365 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1367 * config/pdp11/pdp11.opt (mbcopy): Remove.
1368 (mbcopy-builtin): Remove.
1369 (mbranch-cheap): Remove.
1370 (mbranch-expensive): Remove.
1371 * config/pdp11/predicates.md (expand_shift_operand): Update to
1372 match shift code generation.
1373 (ccnz_operator): New predicate.
1374 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1375 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1376 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1378 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1379 description of O constraint.
1381 2018-06-27 Jeff Law <law@redhat.com>
1382 Austin Law <austinklaw@gmail.com>
1384 * config/v850/v850.md (addsi3_set_flags): New pattern.
1385 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1386 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1387 (zero_extendhisi2_v850_set_flags): Likewise.
1388 (zero_extendqisi2_v850_set_flags): Likewise.
1389 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1390 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1391 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1393 * config/v850/v850-protos.h (notice_update_cc): Remove.
1394 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1395 (v850_print_operand): Handle 'D' and "d".
1396 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1397 Add handling of arithmetic/logical operations compared against zero.
1398 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1399 Do not look at v850_compare_op, instead get mode from last argument.
1400 (v850_gen_compare): Remove
1401 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1402 after reload for prologue insns.
1403 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1405 (construct_save_jarl): Likewise.
1406 (TARGET_FLAGS_REGNUM): Define.
1407 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1408 (NOTICE_UPDATE_CC): Remove.
1409 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1410 than cc0. Conditionalize on reload_completed.
1411 (cmpsi_insn, setfcc_insn): Likewise.
1412 (tst1 splitter): Turn into define_and_split which sets the flags
1414 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1415 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1416 (cstoresf4, cstoredf4): Clobber the flags.
1417 (cmpsi, cmpsf, cmpdf): Remove expanders.
1418 (setf_insn): Remove pattern.
1419 (addsi3): Turn into define_and_split which clobbers the flags after
1420 reload and a suitable pattern (addsi3_clobber_flags) for use after
1422 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1423 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1424 (ashrsi3, ashrsi3_v850e2): Likewise.
1425 (bins): Clobber the flags.
1426 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1427 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1428 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1429 (call_value_internal_short, call_value_internal_long): Likewise.
1430 (callt_save_interrupt, callt_return_interrupt): Likewise.
1431 (save_interrupt, return_interrupt): Likewise.
1432 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1433 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1434 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1435 (All FP comparisons): Only allow after reload has completed.
1437 (divh, divhu): Tweak output template.
1438 (branch_z_normal, branch_z_invert): Remove
1439 (branch_nz_normal, branch_nz_invert): Likewise.
1440 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1442 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1443 * config/v850/v850.c (notice_update_cc): Remove.
1444 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1445 (CC_NO_CARRY): Likewise.
1446 (NOTICE_UPDATE_CC): Define to nothing.
1447 * config/v850/v850.md: Remove block comment on cc0 handling
1448 Remove "cc" attribute from all patterns. Remove cc_status handling
1449 from all patterns. Minor formatting fixes.
1451 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1453 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1454 (cortex-a76.cortex-a55): Likewise.
1455 * config/aarch64/aarch64-tune.md: Regenerate.
1456 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1457 cortex-a76.cortex-a55.
1459 2018-06-27 Jeff Law <law@redhat.com>
1461 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1462 (MULTILIB_DIRNAMES): Similarly.
1464 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1466 * gimple.h (gimple_return_retbnd): Delete.
1467 (gimple_return_set_retbnd): Likewise.
1468 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1469 gimple_return_set_retbnd.
1470 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1471 gimple_return_retbnd and adjust.
1472 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1473 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1474 Explicitly return NULL in a couple more cases. Move assertion
1475 on debug statements and remove unreachable code.
1476 (reset_debug_binding): Do not test id->retbnd.
1477 (expand_call_inline): Do not set it.
1479 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1481 * configure.ac: Add --disable-gcov option.
1482 * configure: Regenerate.
1483 * Makefile.in: Honour @enable_gcov@.
1484 * doc/install.texi: Document --disable-gcov.
1486 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1488 * config/arm/arm-cpus.in (cortex-a76): New entry.
1489 (cortex-a76.cortex-a55): Likewise.
1490 * config/arm/arm-tables.opt: Regenerate.
1491 * config/arm/arm-tune.md: Likewise.
1492 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1493 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1494 cortex-a76.cortex-a55.
1496 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1499 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1501 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1503 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1505 (EPILOGUE_USES): Likewise.
1507 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1509 * tree-inline.c (remap_location): New function extracted from...
1510 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1511 (copy_phis_for_bb): ...here. Call remap_location.
1512 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1514 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1516 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1517 unaligned vsx for 16B memset.
1519 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1522 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1523 ieee128_float_type_node to long_double_type_node unless
1524 TARGET_LONG_DOUBLE_128 is set.
1526 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1528 * cfgloop.c (get_loop_location): Convert return type from
1529 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1530 by implicit construction from rtx_insn *, and using
1531 dump_user_location_t::from_function_decl for the fallback case.
1532 * cfgloop.h (get_loop_location): Convert return type from
1533 location_t to dump_user_location_t.
1534 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1535 dump_printf_loc to pass in a dump_location_t rather than a
1536 location_t, via the gimple stmt.
1537 * coverage.c (get_coverage_counts): Update calls to
1538 dump_printf_loc to pass in dump_location_t rather than a
1540 * doc/optinfo.texi (Dump types): Convert example of
1541 dump_printf_loc from taking "locus" to taking "insn". Update
1542 description of the "_loc" calls to cover dump_location_t.
1543 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1545 (dump_user_location_t::dump_user_location_t): New constructors,
1546 from gimple *stmt and rtx_insn *.
1547 (dump_user_location_t::from_function_decl): New function.
1548 (dump_loc): Make static.
1549 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1550 const dump_location_t &.
1551 (dump_generic_expr_loc): Delete.
1552 (dump_printf_loc): Convert param "loc" from location_t to
1553 const dump_location_t &.
1554 (selftest::test_impl_location): New function.
1555 (selftest::dumpfile_c_tests): New function.
1556 * dumpfile.h: Include "profile-count.h".
1557 (class dump_user_location_t): New class.
1558 (struct dump_impl_location_t): New struct.
1559 (class dump_location_t): New class.
1560 (dump_printf_loc): Convert 2nd param from source_location to
1561 const dump_location_t &.
1562 (dump_generic_expr_loc): Delete.
1563 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1564 const dump_location_t &.
1565 * gimple-fold.c (fold_gimple_assign): Update call to
1566 dump_printf_loc to pass in a dump_location_t rather than a
1567 location_t, via the gimple stmt.
1568 (gimple_fold_call): Likewise.
1569 * gimple-loop-interchange.cc
1570 (loop_cand::analyze_iloop_reduction_var): Update for change to
1571 check_reduction_path.
1572 (tree_loop_interchange::interchange): Update for change to
1574 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1575 change in return-type of find_loop_location.
1576 (graphite_regenerate_ast_isl): Likewise.
1577 * graphite-optimize-isl.c (optimize_isl): Likewise.
1578 * graphite.c (graphite_transform_loops): Likewise.
1579 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1580 pass in a dump_location_t rather than a location_t, via the
1582 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1583 * ipa.c (walk_polymorphic_call_targets): Likewise.
1584 * loop-unroll.c (report_unroll): Convert "locus" param from
1585 location_t to dump_location_t.
1586 (decide_unrolling): Update for change to get_loop_location's
1588 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1589 location_t to dump_user_location_t.
1590 (grid_find_single_omp_among_assignments_1): Updates calls to
1591 dump_printf_loc to pass in a dump_location_t rather than a
1592 location_t, via the gimple stmt.
1593 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1594 location_t to dump_location_t. Updates calls to dump_printf_loc
1595 to pass in a dump_location_t rather than a location_t, via the
1597 (grid_inner_loop_gridifiable_p): Likewise.
1598 (grid_dist_follows_simple_pattern): Likewise.
1599 (grid_gfor_follows_tiling_pattern): Likewise.
1600 (grid_target_follows_gridifiable_pattern): Likewise.
1601 (grid_attempt_target_gridification): Convert initialization
1602 of local "grid" from memset to zero-initialization; FIXME: does
1603 this require C++11? Update call to dump_printf_loc to pass in a
1604 optinfo_location rather than a location_t, via the gimple stmt.
1605 * profile.c (read_profile_edge_counts): Updates call to
1606 dump_printf_loc to pass in a dump_location_t rather than a
1608 (compute_branch_probabilities): Likewise.
1609 * selftest-run-tests.c (selftest::run_tests): Call
1611 * selftest.h (dumpfile_c_tests): New decl.
1612 * tree-loop-distribution.c (pass_loop_distribution::execute):
1613 Update for change in return type of find_loop_location.
1614 * tree-parloops.c (parallelize_loops): Likewise.
1615 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1616 "locus" from location_t to dump_user_location_t.
1617 (canonicalize_loop_induction_variables): Likewise.
1618 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1619 for change in return type of find_loop_location.
1620 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1621 to dump_printf_loc to pass in a dump_location_t rather than a
1622 location_t, via the stmt.
1623 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1625 * tree-vect-loop-manip.c (find_loop_location): Convert return
1626 type from source_location to dump_user_location_t.
1627 (vect_do_peeling): Update for above change.
1628 (vect_loop_versioning): Update for change in type of
1630 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1631 from location_t to dump_user_location_t.
1632 (vect_estimate_min_profitable_iters): Update for change in type
1634 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1635 location_t to dump_location_t.
1636 (vect_slp_bb): Update for change in type of vect_location.
1637 * tree-vectorizer.c (vect_location): Convert from source_location
1638 to dump_user_location_t.
1639 (try_vectorize_loop_1): Update for change in vect_location's type.
1640 (vectorize_loops): Likewise.
1641 (increase_alignment): Likewise.
1642 * tree-vectorizer.h (vect_location): Convert from source_location
1643 to dump_user_location_t.
1644 (find_loop_location): Convert return type from source_location to
1645 dump_user_location_t.
1646 (check_reduction_path): Convert 1st param from location_t to
1647 dump_user_location_t.
1648 * value-prof.c (check_counter): Update call to dump_printf_loc to
1649 pass in a dump_user_location_t rather than a location_t; update
1650 call to error_at for change in type of "locus".
1651 (check_ic_target): Update call to dump_printf_loc to
1652 pass in a dump_user_location_t rather than a location_t, via the
1655 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1657 * config/s390/s390.h (enum processor_flags): Do not use
1658 default tune parameter when -march was specified.
1660 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1663 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1664 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1667 2018-06-26 Richard Biener <rguenther@suse.de>
1669 PR tree-optimization/86287
1671 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1672 (vect_analyze_loop): Initialize n_stmts.
1674 2018-06-26 Richard Biener <rguenther@suse.de>
1677 * fold-const.c (fold_convertible_p): Pointer extension
1680 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1683 * dwarf2out.c (loc_list_has_views): Adjust comments.
1684 (dw_loc_list): Split single cross-partition range with
1687 2018-06-25 Jeff Law <law@redhat.com>
1689 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1690 on -mbig-switch by default.
1692 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1694 (const_float_0_operand): Remove unused predicate.
1695 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1696 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1697 (recipsf2): New expander. Original pattern now called
1699 (recipdf2, recipdf2_insn): Similarly.
1700 (rsqrtsf2, rsqrtsf2_insn): Similarly
1701 (rsqrtdf2, rsqrtdf2_insn): Similarly
1703 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1705 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1706 Simplify logic for FreeBSD (twice).
1708 2018-06-25 Martin Sebor <msebor@redhat.com>
1710 PR tree-optimization/86204
1711 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1712 a strnlen result if it's less than the length of the string.
1714 2018-06-25 Martin Sebor <msebor@redhat.com>
1716 PR tree-optimization/85700
1717 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1718 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1719 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1721 2018-06-25 Martin Sebor <msebor@redhat.com>
1723 * doc/extend.texi (Zero-length arrays): Update and clarify.
1725 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1727 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1728 added IEEE/IBM long double multilib support on PowerPC little
1729 endian Linux systems.
1730 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1731 (MULTILIB_DEFAULTS): Likewise.
1732 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1734 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1735 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1736 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1738 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1741 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1742 (REORDER_45): Likewise.
1744 2018-06-25 Jeff Law <law@redhat.com>
1746 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1747 dividend to 32 bits. Adjust length.
1748 (udivmodhi4): Cleanup output template. Fix length.
1750 2018-06-25 Carl Love <cel@us.ibm.com>
1752 * config/rs6000/vsx.md: Change word selector to prefered location.
1754 2018-06-25 Richard Biener <rguenther@suse.de>
1756 PR tree-optimization/86304
1757 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1758 epilogue-if-converted loops as well.
1760 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1762 * lto-section-out.c (lto_begin_section): Do not print section
1763 name for noaddr and unnumbered dumps.
1765 2018-06-25 Richard Biener <rguenther@suse.de>
1767 * tree-vectorizer.h (struct vec_info_shared): New structure
1768 with parts split out from struct vec_info and loop_nest from
1769 struct _loop_vec_info.
1770 (struct vec_info): Adjust accordingly.
1771 (struct _loop_vec_info): Likewise.
1772 (LOOP_VINFO_LOOP_NEST): Adjust.
1773 (LOOP_VINFO_DATAREFS): Likewise.
1774 (LOOP_VINFO_DDRS): Likewise.
1775 (struct _bb_vec_info): Likewise.
1776 (BB_VINFO_DATAREFS): Likewise.
1777 (BB_VINFO_DDRS): Likewise.
1778 (struct _stmt_vec_info): Add dr_aux member.
1779 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1780 (DR_MISALIGNMENT_UNINITIALIZED): New.
1781 (set_dr_misalignment): Adjust.
1782 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1783 (vect_analyze_loop): Adjust prototype.
1784 (vect_analyze_loop_form): Likewise.
1785 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1786 Compute dependences lazily.
1787 (vect_record_base_alignments): Use shared datarefs/ddrs.
1788 (vect_verify_datarefs_alignment): Likewise.
1789 (vect_analyze_data_refs_alignment): Likewise.
1790 (vect_analyze_data_ref_accesses): Likewise.
1791 (vect_analyze_data_refs): Likewise.
1792 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1793 constructor parameter for shared part.
1794 (vect_analyze_loop_form): Pass in shared part and adjust.
1795 (vect_analyze_loop_2): Pass in storage for the number of
1796 stmts. Move loop nest finding to the caller. Compute
1798 (vect_analyze_loop): Pass in shared part.
1799 (vect_transform_loop): Verify shared datarefs are unchanged.
1800 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1801 constructor parameter for shared part.
1802 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1803 (vect_slp_bb): Verify shared datarefs are unchanged before
1805 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1807 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1808 DR_MISALIGNMENT_UNINITIALIZED.
1809 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1810 parameter for shared part.
1811 (vec_info::~vec_info): Adjust.
1812 (vec_info_shared::vec_info_shared): New.
1813 (vec_info_shared::~vec_info_shared): Likewise.
1814 (vec_info_shared::save_datarefs): Likewise.
1815 (vec_info_shared::check_datarefs): Likewise.
1816 (try_vectorize_loop_1): Construct shared part live for analyses
1817 of a single loop for multiple vector sizes.
1818 * tree-parloops.c (gather_scalar_reductions): Adjust.
1820 2018-06-25 Richard Biener <rguenther@suse.de>
1822 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1823 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1824 (vect_analyze_data_refs): Remove similar code from here and
1825 simplify accordingly.
1827 2018-06-25 Richard Biener <rguenther@suse.de>
1829 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1830 for reverse storage order accesses rather than asserting
1831 they cannot happen here.
1833 2018-06-25 Tom de Vries <tdevries@suse.de>
1836 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1837 Use data16 instead of .byte for insn prefix.
1839 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1842 * parser.c (make_char_string_pack): Pass this literal chars
1843 through cpp_interpret_string.
1844 (cp_parser_userdef_numeric_literal): Check the result of
1845 make_char_string_pack.
1847 2018-06-24 Maya Rashish <coypu@sdf.org>
1849 * ginclude/stddef.h: Simplify conditions around avoiding
1850 re-definition of __size_t.
1852 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1854 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1855 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1857 2018-06-22 Maya Rashish <coypu@sdf.org>
1859 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1860 made to OpenBSD and NetBSD through the years.
1862 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1864 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1865 behavior of vec_pack (vector double, vector double) to match
1866 behavior of vec_float2 (vector double, vector double).
1868 2018-06-22 Olivier Hainque <hainque@adacore.com>
1870 * gimplify.c (gimplify_function_tree): Prevent creation
1871 of a trampoline for the address of the current function
1872 passed to entry/exit instrumentation hooks.
1874 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1877 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1880 2018-06-22 Martin Liska <mliska@suse.cz>
1882 PR tree-optimization/86263
1883 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1884 Bail out if is_enabled is false.
1885 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1887 (jump_table_cluster::is_enabled): New function.
1889 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1891 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1892 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1893 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1894 (lto_input_ts_binfo_tree_pointers): Likewise.
1895 * tree-streamer-out.c (streamer_write_tree_bitfields,
1896 write_ts_binfo_tree_pointers): Likewise.
1897 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1899 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1901 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1903 2018-06-22 Martin Liska <mliska@suse.cz>
1905 * symbol-summary.h (get): Make it pure and inline move
1906 functionality from ::get function.
1907 (get): Remove and inline into ::get and ::get_create.
1908 (get_create): Move code from ::get function.
1910 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1913 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1914 -x assembler-with-cpp.
1916 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1918 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1919 _FILE_OFFSET_BITS=64 for C++.
1921 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1923 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1924 conversion insn that shows up when pr85657-3.c is compiled using
1925 IEEE 128-bit long double.
1926 (neg<mode>2_internal): Use the correct mode to check whether the
1927 mode is IBM extended.
1928 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1929 multiply and divide external functions from being created more
1932 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1934 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1936 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1937 the edge can be forwarded.
1938 (cfg_layout_merge_blocks): Likewise.
1940 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1942 * except.c (finish_eh_generation): Commit edge insertions only after
1943 the EH edges have been redirected from post-landing to landing pads.
1945 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1947 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1948 create_tmp_var_for to create the FRAME decl.
1949 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1951 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1953 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1954 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1955 debug statement when resetting its value.
1956 (expand_call_inline): Copy the locus of the call onto the assignment
1957 of the return value, if any. Use local variable in more cases.
1959 2018-06-21 Martin Liska <mliska@suse.cz>
1961 * ipa-pure-const.c (propagate_nothrow): Use
1962 funct_state_summaries->get.
1963 (dump_malloc_lattice): Likewise.
1964 (propagate_malloc): Likewise.
1966 2018-06-21 Richard Biener <rguenther@suse.de>
1968 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1969 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1970 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1972 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1973 BLOCK_ABSTRACT_ORIGIN unconditionally.
1975 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1977 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1979 * ipa-reference.c (ipa_reference_c_finalize): Delete
1980 ipa_ref_opt_sum_summaries and set it to NULL.
1982 2018-06-21 Tom de Vries <tdevries@suse.de>
1984 PR tree-optimization/85859
1985 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1986 test with comment from bb_no_side_effects_p.
1988 2018-06-21 Richard Biener <rguenther@suse.de>
1990 PR tree-optimization/86232
1991 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1992 max for constant niter.
1994 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1996 * config/aarch64/aarch64-simd.md
1997 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1999 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2001 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2002 Make opernads of the unspec commutative.
2004 2018-06-21 Richard Biener <rguenther@suse.de>
2006 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2007 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2008 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2009 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2011 (vect_record_base_alignments): Do not record base alignment
2012 for gathers or scatters.
2013 (vect_compute_data_ref_alignment): Drop return value that is always
2014 true. Bail out early for gathers or scatters.
2015 (vect_enhance_data_refs_alignment): Bail out early for gathers
2017 (vect_find_same_alignment_drs): Likewise.
2018 (vect_analyze_data_refs_alignment): Remove dead code.
2019 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2020 (vect_analyze_data_refs): For possible gathers or scatters do
2021 not create an alternate DR, just check their possible validity
2022 and mark them. Adjust DECL_NONALIASED handling to not rely
2024 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2025 update inits of gathers or scatters.
2026 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2027 Also copy gather/scatter flag to pattern vinfo.
2029 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2031 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2032 behavior of vec_packsu (vector unsigned long long, vector unsigned
2033 long long) to match behavior of vec_packs with same signature.
2035 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2036 Thomas Schwinge <thomas@codesourcery.com>
2037 Cesar Philippidis <cesar@codesourcery.com>
2039 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2040 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2041 (gimplify_adjust_omp_clauses): Likewise.
2042 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2043 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2044 (gimplify_omp_target_update): Update handling of acc update and
2046 * omp-low.c (install_var_field): Remove unused parameter
2047 base_pointers_restrict.
2048 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2049 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2051 (omp_target_base_pointers_restrict_p): Delete.
2052 (scan_omp_target): Update call to scan_sharing_clauses.
2053 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2055 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2056 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2057 (convert_local_omp_clauses): Likewise.
2058 * tree-pretty-print.c (dump_omp_clause): Likewise.
2059 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2061 (omp_clause_code_name): Likewise.
2063 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2066 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2069 PR tree-optimization/86231
2070 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2071 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2073 2018-06-20 Tom de Vries <tdevries@suse.de>
2075 PR tree-optimization/86097
2076 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2077 iv type if signedness of iv type is not the same as that of *nit.
2079 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2081 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2082 EDGE_EH edges, verify they are all EDGE_EH.
2084 2018-06-20 Maya Rashish <coypu@sdf.org>
2086 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2088 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2090 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2091 * config/aarch64/aarch64.c (xgene1_tunings): Add
2092 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2093 (aarch64_mode_valid_for_sched_fusion_p):
2094 Allow 16-byte modes.
2095 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2096 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2098 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2100 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2101 * config/aarch64/iterators.md (VQ2): New mode iterator.
2103 2018-06-20 Martin Liska <mliska@suse.cz>
2105 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2106 Change default ratio from 10 to 8.
2108 2018-06-20 Martin Liska <mliska@suse.cz>
2110 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2112 (bit_test_cluster::find_bit_tests): Likewise.
2113 (switch_decision_tree::analyze_switch_statement): Find clusters.
2114 * tree-switch-conversion.h (struct jump_table_cluster): Document
2117 2018-06-20 Martin Liska <mliska@suse.cz>
2119 * tree-switch-conversion.c (switch_conversion::collect):
2120 Record m_uniq property.
2121 (switch_conversion::expand): Bail out for special conditions.
2122 (group_cluster::~group_cluster): New.
2123 (group_cluster::group_cluster): Likewise.
2124 (group_cluster::dump): Likewise.
2125 (jump_table_cluster::emit): New.
2126 (switch_decision_tree::fix_phi_operands_for_edges): New.
2127 (struct case_node): Remove struct.
2128 (jump_table_cluster::can_be_handled): New.
2129 (case_values_threshold): Moved to header.
2130 (reset_out_edges_aux): Likewise.
2131 (jump_table_cluster::is_beneficial): New.
2132 (bit_test_cluster::can_be_handled): Likewise.
2133 (add_case_node): Remove.
2134 (bit_test_cluster::is_beneficial): New.
2135 (case_bit_test::cmp): New.
2136 (bit_test_cluster::emit): New.
2137 (expand_switch_as_decision_tree_p): Remove.
2138 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2139 (fix_phi_operands_for_edge): Likewise.
2140 (switch_decision_tree::analyze_switch_statement): New.
2141 (compute_cases_per_edge): Move ...
2142 (switch_decision_tree::compute_cases_per_edge): ... here.
2143 (try_switch_expansion): Likewise.
2144 (switch_decision_tree::try_switch_expansion): Likewise.
2145 (record_phi_operand_mapping): Likewise.
2146 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2147 (emit_case_decision_tree): Likewise.
2148 (switch_decision_tree::emit): Likewise.
2149 (balance_case_nodes): Likewise.
2150 (switch_decision_tree::balance_case_nodes): Likewise.
2151 (dump_case_nodes): Likewise.
2152 (switch_decision_tree::dump_case_nodes): Likewise.
2153 (emit_jump): Likewise.
2154 (switch_decision_tree::emit_jump): Likewise.
2155 (emit_cmp_and_jump_insns): Likewise.
2156 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2157 (emit_case_nodes): Likewise.
2158 (switch_decision_tree::emit_case_nodes): Likewise.
2159 (conditional_probability): Remove.
2160 * tree-switch-conversion.h (enum cluster_type): New.
2162 (struct cluster): Likewise.
2163 (cluster::cluster): Likewise.
2164 (struct simple_cluster): Likewise.
2165 (simple_cluster::simple_cluster): Likewise.
2166 (struct group_cluster): Likewise.
2167 (struct jump_table_cluster): Likewise.
2168 (struct bit_test_cluster): Likewise.
2169 (struct min_cluster_item): Likewise.
2170 (struct case_tree_node): Likewise.
2171 (case_tree_node::case_tree_node): Likewise.
2172 (jump_table_cluster::case_values_threshold): Likewise.
2173 (struct case_bit_test): Likewise.
2174 (struct switch_decision_tree): Likewise.
2175 (struct switch_conversion): Likewise.
2176 (switch_decision_tree::reset_out_edges_aux): Likewise.
2178 2018-06-20 Martin Liska <mliska@suse.cz>
2180 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2181 (hoist_edge_and_branch_if_true): Likewise.
2182 (expand_switch_using_bit_tests_p): Likewise.
2183 (struct case_bit_test): Likewise.
2184 (case_bit_test_cmp): Likewise.
2185 (emit_case_bit_tests): Likewise.
2186 (switch_conversion::switch_conversion): New class.
2187 (struct switch_conv_info): Remove old struct.
2188 (collect_switch_conv_info): More to ...
2189 (switch_conversion::collect): ... this.
2190 (check_range): Likewise.
2191 (switch_conversion::check_range): Likewise.
2192 (check_all_empty_except_final): Likewise.
2193 (switch_conversion::check_all_empty_except_final): Likewise.
2194 (check_final_bb): Likewise.
2195 (switch_conversion::check_final_bb): Likewise.
2196 (create_temp_arrays): Likewise.
2197 (switch_conversion::create_temp_arrays): Likewise.
2198 (free_temp_arrays): Likewise.
2199 (gather_default_values): Likewise.
2200 (switch_conversion::gather_default_values): Likewise.
2201 (build_constructors): Likewise.
2202 (switch_conversion::build_constructors): Likewise.
2203 (constructor_contains_same_values_p): Likewise.
2204 (switch_conversion::contains_same_values_p): Likewise.
2205 (array_value_type): Likewise.
2206 (switch_conversion::array_value_type): Likewise.
2207 (build_one_array): Likewise.
2208 (switch_conversion::build_one_array): Likewise.
2209 (build_arrays): Likewise.
2210 (switch_conversion::build_arrays): Likewise.
2211 (gen_def_assigns): Likewise.
2212 (switch_conversion::gen_def_assigns): Likewise.
2213 (prune_bbs): Likewise.
2214 (switch_conversion::prune_bbs): Likewise.
2215 (fix_phi_nodes): Likewise.
2216 (switch_conversion::fix_phi_nodes): Likewise.
2217 (gen_inbound_check): Likewise.
2218 (switch_conversion::gen_inbound_check): Likewise.
2219 (process_switch): Use the newly created class.
2220 (switch_conversion::expand): New.
2221 (switch_conversion::~switch_conversion): New.
2222 * tree-switch-conversion.h: New file.
2224 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2226 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2227 tree-vect-patterns.c.
2228 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2229 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2230 (vect_recog_sad_pattern): Likewise.
2231 (vect_recog_widen_sum_pattern): Likewise.
2232 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2233 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2234 (vect_recog_rotate_pattern): Likewise.
2235 (vect_recog_mult_pattern): Likewise.
2236 (vect_recog_vector_vector_shift_pattern): Likewise.
2237 (vect_recog_divmod_pattern): Likewise.
2238 (vect_recog_mixed_size_cond_pattern): Likewise.
2239 (vect_recog_bool_pattern): Likewise.
2240 (vect_recog_mask_conversion_pattern): Likewise.
2241 (vect_try_gather_scatter_pattern): Likewise.
2242 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2243 (vect_recog_over_widening_pattern): Likewise.
2244 (vect_recog_gather_scatter_pattern): Likewise.
2245 (vect_recog_func_ptr): Move from tree-vectorizer.h
2246 (vect_vect_recog_func_ptrs): Move further down the file.
2247 (vect_recog_func): Likewise. Remove the third argument.
2248 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2249 (vect_pattern_recog_1): Expect the pattern function to do any
2250 necessary target tests. Also expect it to provide a vector type.
2251 Remove the type_in handling.
2253 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2255 * tree-vect-patterns.c (vect_pattern_detected): New function.
2256 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2257 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2258 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2259 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2260 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2261 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2262 (vect_recog_mask_conversion_pattern)
2263 (vect_try_gather_scatter_pattern): Likewise.
2265 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2267 * tree-vect-patterns.c (vect_get_internal_def): New function.
2268 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2269 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2270 (search_type_for_mask_1): Use it.
2272 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2274 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2275 redundant WIDEN_SUM_EXPR handling.
2276 (vect_recog_sad_pattern): Likewise.
2278 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2280 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2281 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2282 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2283 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2284 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2286 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2288 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2289 use the stmt_vec_info of the original bb statement for the
2290 new zero assignment, even if the call is part of a pattern.
2292 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2294 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2295 that the sequence is attached to the original statement rather
2296 than the pattern statement.
2297 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2298 PATTERN_DEF_SEQ from the original statement rather than
2299 the main pattern statement.
2300 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2301 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2302 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2304 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2306 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2307 definition statements before the early exit for statements that aren't
2309 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2311 (vect_transform_loop): ...here. Process pattern definition
2312 statements without first checking whether the main pattern
2313 statement is live or relevant.
2315 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2317 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2318 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2320 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2322 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2323 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2324 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2325 (expand_block_compare): Change select_block_compare_mode call.
2326 (expand_strncmp_align_check): Use new functions, fix comment.
2327 (emit_final_str_compare_gpr): New function.
2328 (expand_strn_compare): Refactor and clean up code.
2329 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2331 2018-06-19 Tony Reix <tony.reix@atos.com>
2332 Damien Bergamini <damien.bergamini@atos.com>
2333 David Edelsohn <dje.gcc@gmail.com>
2335 * collect2.c (static_obj): New variable.
2336 (static_libs): New variable.
2337 (is_in_list): Uncomment declaration.
2338 (main): Track AIX libraries linked statically.
2339 (is_in_list): Uncomment definition.
2340 (scan_prog_file): Don't add AIX shared libraries initializer
2341 to constructor list if linking statically.
2343 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2345 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2347 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2349 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2351 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2354 2018-06-19 Martin Liska <mliska@suse.cz>
2356 * config/i386/i386.c (ix86_can_inline_p): Do not use
2357 ipa_fn_summaries::get_create.
2358 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2360 (devirtualization_time_bonus): Likewise.
2361 (ipcp_propagate_stage): Likewise.
2362 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2363 (edge_set_predicate): Likewise.
2364 (evaluate_conditions_for_known_args): Likewise.
2365 (evaluate_properties_for_edge): Likewise.
2366 (ipa_call_summary::reset): Tranform to ...
2367 (ipa_call_summary::~ipa_call_summary): ... this.
2368 (ipa_fn_summary::reset): Transform to ...
2369 (ipa_fn_summary::~ipa_fn_summary): ... this.
2370 (ipa_fn_summary_t::remove): Rename to ...
2371 (ipa_fn_summary_t::remove_callees): ... this.
2372 (ipa_fn_summary_t::duplicate): Use placement new
2373 instead of memory copy.
2374 (ipa_call_summary_t::duplicate): Likewise.
2375 (ipa_call_summary_t::remove): Remove.
2376 (dump_ipa_call_summary): Change get_create to get.
2377 (ipa_dump_fn_summary): Dump only when summary exists.
2378 (analyze_function_body): Use symbol_summary::get instead
2380 (compute_fn_summary): Likewise.
2381 (estimate_edge_devirt_benefit): Likewise.
2382 (estimate_edge_size_and_time): Likewise.
2383 (inline_update_callee_summaries): Likewise.
2384 (remap_edge_change_prob): Likewise.
2385 (remap_edge_summaries): Likewise.
2386 (ipa_merge_fn_summary_after_inlining): Likewise.
2387 (write_ipa_call_summary): Likewise.
2388 (ipa_fn_summary_write): Likewise.
2389 (ipa_free_fn_summary): Likewise.
2390 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2391 (struct ipa_call_summary): Likewise.
2392 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2394 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2395 (estimate_size_after_inlining): Likewise.
2396 (estimate_growth): Likewise.
2397 (growth_likely_positive): Likewise.
2398 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2399 (inline_call): Likewise.
2400 * ipa-inline.c (caller_growth_limits): Likewise.
2401 (can_inline_edge_p): Likewise.
2402 (can_inline_edge_by_limits_p): Likewise.
2403 (compute_uninlined_call_time): Likewise.
2404 (compute_inlined_call_time): Likewise.
2405 (want_inline_small_function_p): Likewise.
2406 (edge_badness): Likewise.
2407 (update_caller_keys): Likewise.
2408 (update_callee_keys): Likewise.
2409 (inline_small_functions): Likewise.
2410 (inline_to_all_callers_1): Likewise.
2411 (dump_overall_stats): Likewise.
2412 (early_inline_small_functions): Likewise.
2413 (early_inliner): Likewise.
2414 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2415 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2416 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2417 * ipa-split.c (execute_split_functions): Likewise.
2418 * symbol-summary.h: Likewise.
2419 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2421 2018-06-19 Richard Biener <rguenther@suse.de>
2423 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2424 (vectorize_loops): ... here. Fix dbgcnt handling.
2425 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2427 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2430 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2431 ieee128 argument takes up only one (vector) register, not two (floating
2434 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2436 * gimplify.c (gimplify_init_constructor): Really never clear for an
2437 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2439 2018-06-19 Richard Biener <rguenther@suse.de>
2441 PR tree-optimization/86179
2442 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2443 after failed recognition.
2445 2018-06-18 Martin Sebor <msebor@redhat.com>
2448 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2449 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2450 Handle integer subtraction.
2451 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2452 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2454 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2456 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2457 param from rtx to rtx_insn *.
2458 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2460 (frv_ifcvt_modify_insn): Likwise.
2461 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2462 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2463 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2464 as_a <rtx_insn *> cast to local "unprotected_region" once
2465 it's been established that it's not NULL or pc_rtx.
2466 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2467 param "sethi" from rtx to rtx_insn *.
2468 (nds32_group_float_insns): Likewise for param "insn".
2469 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2471 (vax_output_int_subtract): Likewise.
2472 * config/vax/vax.c (vax_output_int_add): Likewise for param
2474 (vax_output_int_subtract): Likewise.
2475 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2476 (emit_pattern_after): Likewise for param "after".
2477 (emit_insn_after): Likewise.
2478 (emit_jump_insn_after): Likewise.
2479 (emit_call_insn_after): Likewise.
2480 (emit_debug_insn_after): Likewise.
2481 (emit_pattern_before): Likewise for param "before".
2482 (emit_insn_before): Likewise.
2483 (emit_jump_insn_before): Likewise.
2484 * final.c (get_insn_template): Likewise for param "insn", removing
2486 * output.h (get_insn_template): Likewise for 2nd param.
2487 * rtl.h (emit_insn_before): Likewise.
2488 (emit_jump_insn_before): Likewise.
2489 (emit_debug_insn_before_noloc): Likewise.
2490 (emit_insn_after): Likewise.
2491 (emit_jump_insn_after): Likewise.
2492 (emit_call_insn_after): Likewise.
2493 (emit_debug_insn_after): Likewise.
2494 (set_insn_deleted): Likewise for param.
2496 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2499 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2500 floating point modes, so that IFmode is numerically greater than
2501 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2502 to declare the ordering. This prevents IFmode from being
2503 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2504 machine. Include rs6000-modes.h to share the fractional values
2505 between genmodes* and the rest of the compiler.
2509 * config/rs6000/rs6000-modes.h: New file.
2510 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2511 meaning of rs6000_long_double_size so that 126..128 selects an
2512 appropriate 128-bit floating point type.
2513 (rs6000_option_override_internal): Likewise.
2514 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2515 (TARGET_LONG_DOUBLE_128): Change the meaning of
2516 rs6000_long_double_size so that 126..128 selects an appropriate
2517 128-bit floating point type.
2518 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2519 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2520 source and destination to match the standard usage.
2521 (truncifkf2): Likewise.
2522 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2523 ISA 2.07 to use an explicit clobber, instead of passing in a
2525 (copysign<mode>3_soft): Likewise.
2527 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2529 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2530 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2531 (vect_slp_analyze_instance_dependence): Likewise.
2532 (vect_enhance_data_refs_alignment): Likewise.
2533 (vect_analyze_data_refs_alignment): Likewise.
2534 (vect_slp_analyze_and_verify_instance_alignment
2535 (vect_analyze_data_ref_accesses): Likewise.
2536 (vect_prune_runtime_alias_test_list): Likewise.
2537 (vect_analyze_data_refs): Likewise.
2538 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2539 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2540 (vect_analyze_scalar_cycles_1): Likewise.
2541 (vect_get_loop_niters): Likewise.
2542 (vect_analyze_loop_form_1): Likewise.
2543 (vect_update_vf_for_slp): Likewise.
2544 (vect_analyze_loop_operations): Likewise.
2545 (vect_analyze_loop): Likewise.
2546 (vectorizable_induction): Likewise.
2547 (vect_transform_loop): Likewise.
2548 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2549 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2550 (vect_make_slp_decision): Likewise.
2551 (vect_detect_hybrid_slp): Likewise.
2552 (vect_slp_analyze_operations): Likewise.
2553 (vect_slp_bb): Likewise.
2554 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2555 (vectorizable_bswap): Likewise.
2556 (vectorizable_call): Likewise.
2557 (vectorizable_simd_clone_call): Likewise.
2558 (vectorizable_conversion): Likewise.
2559 (vectorizable_assignment): Likewise.
2560 (vectorizable_shift): Likewise.
2561 (vectorizable_operation): Likewise.
2562 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2564 2018-06-18 Martin Sebor <msebor@redhat.com>
2566 PR tree-optimization/81384
2567 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2568 * builtins.c (expand_builtin_strnlen): New function.
2569 (expand_builtin): Call it.
2570 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2571 * builtins.def (BUILT_IN_STRNLEN): New.
2572 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2573 Warn for bounds in excess of maximum object size.
2574 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2575 single-value ranges. Handle strnlen.
2576 (handle_builtin_strlen): Handle strnlen.
2577 (strlen_check_and_optimize_stmt): Same.
2578 * doc/extend.texi (Other Builtins): Document strnlen.
2580 2018-06-18 Maya Rashish <coypu@sdf.org>
2582 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2583 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2584 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2586 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2588 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2590 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2592 * tree.c (escaped_string::escape): Replace cast to char * by
2593 const_cast<char *> (unescaped).
2595 2018-06-18 Nick Clifton <nickc@redhat.com>
2598 * tree.c (escaped_string): New class. Converts an unescaped
2599 string into its escaped equivalent.
2600 (warn_deprecated_use): Use the new class to convert the
2601 deprecation message, if present.
2602 (test_escaped_strings): New self test.
2603 (test_c_tests): Add test_escaped_strings.
2604 * doc/extend.texi (deprecated): Add a note that the
2605 deprecation message is affected by the -fmessage-length
2606 option, and that control characters will be escaped.
2607 (#pragma GCC error): Document this pragma.
2608 (#pragma GCC warning): Likewise.
2609 * doc/invoke.texi (-fmessage-length): Document this option's
2610 effect on the #warning and #error preprocessor directives and
2611 the deprecated attribute.
2613 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2615 * tree.c (decl_value_expr_lookup): Revert latest change.
2616 (decl_value_expr_insert): Likewise.
2618 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2620 * gimplify.c (nonlocal_vlas): Delete.
2621 (nonlocal_vla_vars): Likewise.
2622 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2624 (gimplify_body): Do not create and destroy nonlocal_vlas.
2625 * tree-nested.c: Include diagnostic.h.
2626 (use_pointer_in_frame): Tweak.
2627 (lookup_field_for_decl): Add assertion and declare the transformation.
2628 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2629 internal error when the reference is in a wrong context. Do not
2630 create a debug decl by default.
2631 (note_nonlocal_block_vlas): Delete.
2632 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2633 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2634 create a debug decl by default.
2635 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2636 call is in a wrong context.
2637 (fixup_vla_decls): New function.
2638 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2639 debug variables were created.
2640 * tree.c (decl_value_expr_lookup): Add checking assertion.
2641 (decl_value_expr_insert): Likewise.
2643 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2646 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2647 * tree-scalar-evolution.c (interpret_expr): Likewise.
2648 (expression_expensive_p): Likewise.
2649 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2650 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2651 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2652 (ssa_defined_by_minus_one_stmt_p): New.
2654 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2657 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2658 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2659 * dojump.c (do_jump): Likewise.
2660 * expr.c (expand_expr_real_2): Check operand type's sign.
2661 * fold-const.c (const_unop): Handle ABSU_EXPR.
2662 (fold_abs_const): Likewise.
2663 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2664 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2665 (strip_sign_op_1): Likesise.
2666 * match.pd: Add new pattern to generate ABSU_EXPR.
2667 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2668 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2669 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2670 * tree-inline.c (estimate_operator_cost): Likewise.
2671 * tree-pretty-print.c (dump_generic_node): Likewise.
2672 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2673 * tree.def (ABSU_EXPR): New.
2675 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2678 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2679 documented as preserved for backward compatibility only.
2680 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2682 PR rtl-optimization/86108
2683 * bb-reorder.c (create_forwarder_block): Renamed to ...
2684 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2685 jump from new landing pad to the second part.
2686 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2689 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2692 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2693 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2694 Only call store_expr for halves if the mode is the same.
2697 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2698 Fix up comment formatting.
2700 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2702 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2703 (typed_splay_tree::closure,
2704 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2705 (typed_splay_tree::typed_splay_tree,
2706 typed_splay_tree::operator =): Declared private.
2707 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2708 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2709 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2710 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2711 typed_splay_tree::splay_tree_splay,
2712 typed_splay_tree::splay_tree_foreach_helper,
2713 typed_splay_tree::splay_tree_insert,
2714 typed_splay_tree::splay_tree_remove,
2715 typed_splay_tree::splay_tree_lookup,
2716 typed_splay_tree::splay_tree_predecessor,
2717 typed_splay_tree::splay_tree_successor,
2718 typed_splay_tree::splay_tree_min,
2719 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2720 (typed_splay_tree::root, typed_splay_tree::comp,
2721 typed_splay_tree::delete_key,
2722 typed_splay_tree::delete_value): New data members.
2723 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2724 typed_splay_tree::remove.
2726 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2728 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2729 -mginv and -mno-ginv to the assembler.
2730 * config/mips/mips.opt (-mcrc): New option.
2732 * doc/invoke.text (-mcrc): Document.
2735 2018-06-15 Nick Clifton <nickc@redhat.com>
2738 * tree.c (escaped_string): New class. Converts an unescaped
2739 string into its escaped equivalent.
2740 (warn_deprecated_use): Use the new class to convert the
2741 deprecation message, if present.
2742 (test_escaped_strings): New self test.
2743 (test_c_tests): Add test_escaped_strings.
2744 * doc/extend.texi (deprecated): Add a note that the
2745 deprecation message is affected by the -fmessage-length
2746 option, and that control characters will be escaped.
2747 (#pragma GCC error): Document this pragma.
2748 (#pragma GCC warning): Likewise.
2749 * doc/invoke.texi (-fmessage-length): Document this option's
2750 effect on the #warning and #error preprocessor directives and
2751 the deprecated attribute.
2753 2018-06-15 Richard Biener <rguenther@suse.de>
2755 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2756 here, also noting vector size used.
2757 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2758 size used in MSG_OPTIMIZED_LOCATIONS dump.
2759 (pass_slp_vectorize::execute): Adjust.
2761 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2764 * config/arc/arc.c (arc_return_address_register): Fix
2767 2018-06-15 Richard Biener <rguenther@suse.de>
2770 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2771 leave useless conversion stripping to force_gimple_operand_gsi.
2772 (gimplify_build2): Likewise.
2773 (gimplify_build1): Likewise.
2775 2018-06-15 Richard Biener <rguenther@suse.de>
2778 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2779 before adjusting their block.
2781 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2783 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2784 multilibs for *-*-rtems*.
2785 * config/riscv/t-rtems: New file.
2787 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2790 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2791 unsigned_type_for returns NULL.
2794 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2795 subregs of multi-word pseudos unless the float mode has word size.
2797 2018-06-14 Richard Biener <rguenther@suse.de>
2800 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2801 duplicate and harmful type_hash_canon.
2802 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2804 2018-06-14 Richard Biener <rguenther@suse.de>
2807 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2810 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2812 * config/rtems.h (STDINT_LONG32): Define.
2814 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2815 Prachi Godbole <prachi.godbole@imgtec.com>
2817 * config/mips/mips-cpus.def: Define P6600.
2818 * config/mips/mips-tables.opt: Regenerate.
2819 * config/mips/mips.c (mips_ucbranch_type): New enum.
2820 (mips_rtx_cost_data): Add support for P6600.
2821 (mips_issue_rate): Likewise.
2822 (mips_multipass_dfa_lookahead): Likewise.
2823 (mips_avoid_hazard): Likewise.
2824 (mips_reorg_process_insns): Likewise.
2825 (mips_classify_branch_p6600): New function.
2826 * config/mips/mips.h (TUNE_P6600): New define.
2827 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2828 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2829 * config/mips/mips.md: Include p6600.md.
2830 (processor): Add p6600.
2831 * config/mips/p6600.md: New file.
2832 * doc/invoke.texi: Add p6600 to supported architectures.
2834 2018-06-13 Martin Sebor <msebor@redhat.com>
2836 PR tree-optimization/86114
2837 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2839 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2841 2018-06-13 Richard Biener <rguenther@suse.de>
2843 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2844 Properly set vector type of the intermediate stmt.
2845 * tree-vect-stmts.c (vectorizable_operation): The destination
2846 var always has vectype_out type.
2848 2018-06-13 Jeff Law <law@redhat.com>
2850 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2851 integer 0 for argument to print_rtl_with_bb.
2852 (rl78_reorg): Likewise.
2854 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2856 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2857 from rtx to rtx_insn *.
2858 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2860 (add_sched_insns_for_speculation): Likewise for local "target",
2861 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2862 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2863 from rtx_insn ** to rtx_code_label **.
2864 (reorg_emit_nops): Likewise.
2865 (c6x_reorg): Likewise for local "call_labels".
2866 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2868 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2869 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2870 the loops over LABEL_REFS.
2871 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2873 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2874 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2875 (split_branches): Strengthen local "olabel" from rtx to
2876 rtx_insn *, adding a safe_as_a cast.
2877 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2879 (add_insn_after): Likewise for first two params.
2880 (add_insn_before): Likewise.
2881 (remove_insn): Likewise for param.
2882 (emit_pattern_before_noloc): Likewise for second and third params.
2883 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2884 (emit_call_insn_before_noloc): Likewise.
2885 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2887 (emit_barrier_before): Likewise.
2888 (emit_label_before): Strengthen "label" param from "rtx" to
2889 "rtx_code_label *". Strengthen "before" param from "rtx" to
2891 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2893 (emit_pattern_after_noloc): Likewise.
2894 (emit_insn_after_noloc): Likewise.
2895 (emit_jump_insn_after_noloc): Likewise.
2896 (emit_call_insn_after_noloc): Likewise.
2897 (emit_debug_insn_after_noloc): Likewise.
2898 (emit_barrier_after): Likewise.
2899 (emit_label_after): Likewise for both params.
2900 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2901 "loc" param from "int" to "location_t".
2902 (emit_insn_after_setloc): Likewise.
2903 (emit_jump_insn_after_setloc): Likewise.
2904 (emit_call_insn_after_setloc): Likewise.
2905 (emit_debug_insn_after_setloc): Likewise.
2906 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2907 "loc" param from "int" to "location_t".
2908 (emit_pattern_before): Convert NULL_RTX to NULL.
2909 (emit_insn_before_setloc): Convert "loc" param from "int" to
2911 (emit_jump_insn_before_setloc): Likewise.
2912 (emit_call_insn_before_setloc): Likewise.
2913 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2914 rtx_insn *. Convert "loc" param from "int" to "location_t".
2915 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2916 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2917 Convert 3rd param from "int" to "location_t".
2918 (emit_barrier_before, emit_barrier_after, next_real_insn):
2919 Strengthen param from rtx to rtx_insn *.
2920 (emit_label_before): Strengthen 1st param from "rtx" to
2921 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2923 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2924 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2925 Strengthen 2nd param from "rtx" to "rtx_insn *".
2926 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2927 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2928 Likewise. Convert 3rd param from "int" to "location_t".
2929 (emit_label_after): Strengthen 1st param from "rtx" to
2931 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2933 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2934 from "rtx" to "rtx_insn *".
2936 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2938 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2939 bodies streamed in with -Q.
2940 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2941 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2942 * gimple-streamer-out.c: Include gimple-pretty-print.h
2943 (output_bb): Dump stmts streamed.
2944 * lto-section-out.c: Include print-tree.h
2945 (lto_begin_section): Dump sections created.
2946 (lto_output_decl_index): Dump decl encoded.
2947 * lto-streamer-out.c: Include print-tree.h
2948 (create_output_block): Dump output block created.
2949 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2950 (output_function): Dump function output.
2951 (output_constructor): Dump constructor streamed.
2952 (write_global_stream): Output indexes encoded.
2953 (produce_asm_for_decls): Dump streams encoded.
2954 * lto-streamer.c (streamer_dump_file): New global var.
2955 * lto-streamer.h (streamer_dump_file): Declare.
2956 * passes.c (ipa_write_summaries): Initialize streamer dump.
2957 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2960 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2963 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2964 offsets for register save directives. Emit a second batch of save
2965 directives, if need be, when the function accesses prior frames.
2967 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2969 * config/arc/fpu.md (fmasf4): Force operand to register.
2970 (fnmasf4): Likewise.
2972 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2974 * config/arc/arc-protos.h (arc_pad_return): Remove.
2975 * config/arc/arc.c (machine_function): Remove force_short_suffix
2977 (arc_print_operand): Adjust printing of '&'.
2978 (arc_verify_short): Remove conditional printing of short suffix.
2979 (arc_final_prescan_insn): Remove reference to size_reason.
2980 (pad_return): New function.
2981 (arc_reorg): Call pad_return.
2982 (arc_pad_return): Remove.
2983 (arc_init_machine_status): Remove reference to force_short_suffix.
2984 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2985 (attr length): When attribute iscompact is true force to 2
2986 regardless; in the case of maybe check if we want to force the
2987 instruction to have 4 bytes length.
2988 (nopv): Change it to generate 4 byte long nop as well.
2989 (blockage): New pattern.
2990 (simple_return): Remove call to arc_pad_return.
2991 (p_return_i): Likewise.
2993 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2995 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2997 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2999 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3002 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3004 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3005 for ARC700 and ARCv2.
3007 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3010 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3011 operands[2] instead of operands[1].
3014 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3016 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3017 case, check whether the outer register overlaps an unallocatable
3018 register, not just whether it fits the required class.
3020 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3022 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3024 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3025 (memrefs_conflict_p): Likewise.
3026 (init_alias_analysis): Likewise.
3027 * cfgexpand.c (expand_debug_expr): Likewise.
3028 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3029 * cse.c (fold_rtx): Likewise.
3030 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3031 * expr.c (emit_block_move_hints): Likewise.
3032 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3033 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3034 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3035 (emit_group_store): Likewise.
3036 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3037 to read the PRE/POST_MODIFY increment.
3038 * calls.c (store_one_arg): Use strip_offset.
3039 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3041 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3043 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3044 (simplify_binary_operation_1): Extend CONST_INT handling to
3046 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3047 than a HOST_WIDE_INT.
3048 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3050 (adjust_mems, add_stores): Update accodingly.
3051 (vt_canonicalize_addr): Track polynomial offsets.
3052 (emit_note_insn_var_location): Likewise.
3053 (vt_add_function_parameter): Likewise.
3054 (vt_initialize): Likewise.
3056 2018-06-12 Jeff Law <law@redhat.com>
3058 * config.gcc (alpha*-*-freebsd*): Remove.
3059 * config/alpha/freebsd.h: Remove.
3061 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3064 * spellcheck-tree.c (levenshtein_distance): Rename to...
3065 (get_edit_distance): ...this, and update for underlying renaming.
3066 * spellcheck-tree.h (levenshtein_distance): Rename to...
3067 (get_edit_distance): ...this.
3068 * spellcheck.c (levenshtein_distance): Rename to...
3069 (get_edit_distance): ...this. Convert from Levenshtein distance
3070 to Damerau-Levenshtein distance by supporting transpositions of
3071 adjacent characters. Rename "v1" to "v_next" and "v0" to
3073 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3074 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3075 update for underlying renaming.
3076 (selftest::levenshtein_distance_unit_test): Rename to...
3077 (selftest::test_get_edit_distance_unit): ...this, and update for
3078 underlying renaming.
3079 (selftest::test_find_closest_string): Add example from PR 69968
3080 where transposition helps
3081 (selftest::test_metric_conditions): Update for renaming.
3082 (selftest::test_metric_conditions): Likewise.
3083 (selftest::spellcheck_c_tests): Likewise.
3084 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3085 (get_edit_distance): ...this.
3086 (best_match::consider): Update for renaming.
3088 2018-06-12 Martin Sebor <msebor@redhat.com>
3090 PR tree-optimization/85259
3091 * builtins.c (compute_objsize): Handle constant offsets.
3092 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3093 true iff a warning has been issued.
3094 * gimple.h (gimple_nonartificial_location): New function.
3095 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3096 gimple_nonartificial_location and handle -Wno-system-headers.
3097 (handle_builtin_stxncpy): Same.
3099 2018-06-12 Martin Sebor <msebor@redhat.com>
3102 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3104 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3106 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3107 BUILTIN_VEC_XST entries for pointer to double and long long.
3109 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3112 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3114 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3117 2018-06-12 Martin Liska <mliska@suse.cz>
3119 * doc/options.texi: Document IntegerRange.
3121 2018-06-12 Martin Liska <mliska@suse.cz>
3123 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3124 * opt-functions.awk: Handle Deprecated flag.
3125 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3127 (read_cmdline_option): Report warning for a deprecated option.
3128 * opts.h (struct cl_option): Add new field cl_deprecated.
3129 (CL_ERR_DEPRECATED): New.
3131 2018-06-12 Martin Liska <mliska@suse.cz>
3133 * doc/options.texi: Document Deprecated option flag.
3135 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3137 * config/arc/arc-arch.h (arc_extras): New enum.
3138 (arc_cpu_t):Add field extra.
3139 (arc_cpu_types): Consider the extras.
3140 * config/arc/arc-cpus.def: Add extras info.
3141 * config/arc/arc-opts.h (processor_type): Consider extra field.
3142 * config/arc/arc.c (arc_override_options): Handle extra field.
3144 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3146 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3147 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3149 * config/arc/arc.c (arc_init): Update pic warning.
3150 (irq_range): Update irq range parsing warnings.
3151 (arc_override_options): Update various warning messages.
3152 (arc_handle_aux_attribute): Likewise.
3154 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3156 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3158 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3160 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3161 compiler messages directives.
3163 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3165 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3166 * config/mips/mips-tables.opt: Regenerate.
3167 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3169 * doc/invoke.texi: Document -march=i6500.
3171 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3173 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3174 (i6400_gpmul): Add cpu_unit.
3175 (i6400_gpdiv): Likewise.
3176 (i6400_msa_add_d): Update reservations.
3177 (i6400_msa_int_add) Likewise.
3178 (i6400_msa_short_logic3) Likewise.
3179 (i6400_msa_short_logic2) Likewise.
3180 (i6400_msa_short_logic) Likewise.
3181 (i6400_msa_move) Likewise.
3182 (i6400_msa_cmp) Likewise.
3183 (i6400_msa_short_float2) Likewise.
3184 (i6400_msa_div_d) Likewise.
3185 (i6400_msa_long_logic1) Likewise.
3186 (i6400_msa_long_logic2) Likewise.
3187 (i6400_msa_mult) Likewise.
3188 (i6400_msa_long_float2) Likewise.
3189 (i6400_msa_long_float4) Likewise.
3190 (i6400_msa_long_float5) Likewise.
3191 (i6400_msa_long_float8) Likewise.
3192 (i6400_fpu_fadd): Include frint type.
3193 (i6400_fpu_store): New define_insn_reservation.
3194 (i6400_fpu_load): Likewise.
3195 (i6400_fpu_move): Likewise.
3196 (i6400_fpu_fcmp): Likewise.
3197 (i6400_fpu_fmadd): Likewise.
3198 (i6400_int_mult): Include imul3nc type and update reservation.
3199 (i6400_int_div): Include idiv3 type and update reservation.
3200 (i6400_int_load): Update to check type not move_type.
3201 (i6400_int_store): Likewise.
3202 (i6400_int_prefetch): Set zero latency.
3204 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3206 * gcc.c: Document new %@{...} sequence.
3207 (LINK_COMMAND_SPEC): Use it for the -L switches.
3208 (cpp_unique_options): Use it for the -I switches.
3209 (at_file_argbuf): New global variable.
3210 (in_at_file): Likewise.
3211 (alloc_args): Create at_file_argbuf.
3212 (clear_args): Truncate at_file_argbuf.
3213 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3214 (open_at_file): New function.
3215 (close_at_file): Likewise.
3216 (create_at_file): Delete.
3217 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3220 (validate_switches_from_spec): Deal with %@{...} sequence.
3221 (validate_switches): Likewise.
3222 (driver::finalize): Call clear_args.
3224 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3226 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3228 2018-06-11 Martin Sebor <msebor@redhat.com>
3230 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3231 (Wno-class-memaccess): Add @opindex.
3232 (Wno-templates, Wno-multiple-inheritance): Same.
3233 (Wno-virtual-inheritance, Wno-namespaces): Same.
3234 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3235 (Wno-format-overflow, Wno-format-truncation): Same.
3236 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3237 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3238 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3239 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3240 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3241 (Wno-misspelled-isr): Same.
3243 2018-06-11 Martin Sebor <msebor@redhat.com>
3245 * PR tree-optimization/86083
3246 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3248 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3250 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3252 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3255 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3256 on the correct operand.
3257 (*movdi_internal64): Ditto.
3259 2018-06-11 Martin Liska <mliska@suse.cz>
3261 PR tree-optimization/86089
3262 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3264 2018-06-11 Julia Koval <julia.koval@intel.com>
3266 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3267 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3268 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3270 2018-06-11 Olivier Hainque <hainque@adacore.com>
3272 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3273 for Ada with strict dwarf2.
3275 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3278 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3281 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3283 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3285 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3287 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3290 2018-06-08 Martin Liska <mliska@suse.cz>
3292 * tree-cfg.h (debug_function): Fix argument type to match
3295 2018-06-08 Martin Liska <mliska@suse.cz>
3297 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3298 Remove usage of MPX-related (and removed) fields.
3299 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3301 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3303 * cfg.c (debug): Use TDF_NONE rather than 0.
3304 * cfghooks.c (debug): Likewise.
3305 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3306 (struct dump_option_value_info): Convert to...
3307 (struct kv_pair): ...this template type.
3308 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3310 (optinfo_verbosity_options): Likewise.
3311 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3313 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3314 than int for "optgroup_flags" param.
3315 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3317 (dump_dec): Likewise.
3318 (dump_finish): Use TDF_NONE rather than 0.
3319 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3320 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3321 than 0. Update for change to option_ptr.
3322 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3323 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3324 0. Update for changes to optinfo_verbosity_options and
3326 (opt_info_switch_p): Convert optgroup_flags from int to
3328 (dump_basic_block): Use dump_flags_t rather than int
3329 for "dump_kind" param.
3330 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3331 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3332 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3333 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3334 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3335 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3336 TDF_NONE): Convert from macros to...
3337 (enum dump_flag): ...this new enum.
3338 (dump_flags_t): Update to use enum.
3339 (operator|, operator&, operator~, operator|=, operator&=):
3340 Implement for dump_flags_t.
3341 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3342 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3343 Convert from macros to...
3344 (enum optgroup_flag): ...this new enum.
3345 (optgroup_flags_t): New typedef.
3346 (operator|, operator|=): Implement for optgroup_flags_t.
3347 (struct dump_file_info): Convert field "alt_flags" to
3348 dump_flags_t. Convert field "optgroup_flags" to
3350 (dump_basic_block): Use dump_flags_t rather than int for param.
3351 (dump_generic_expr_loc): Likewise.
3352 (dump_dec): Likewise.
3353 (dump_register): Convert param "optgroup_flags" to
3355 (opt_info_enable_passes): Likewise.
3356 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3358 * gimple-pretty-print.c (debug): Likewise.
3359 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3360 (merged_store_group::apply_stores): Likewise.
3361 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3362 * gimple.c (verify_gimple_pp): Likewise.
3363 * graphite-poly.c (print_pbb_body): Likewise.
3364 * passes.c (pass_manager::register_one_dump_file): Convert
3365 local "optgroup_flags" to optgroup_flags_t.
3366 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3368 (debug_body): Likewise.
3369 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3370 to optgroup_flags_t.
3371 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3373 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3374 (convert_mult_to_fma): Likewise.
3375 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3376 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3377 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3378 "dump_kind" to dump_flags_t.
3380 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3382 * config/rs6000/rs6000.c (min, max): Delete.
3384 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3386 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3387 -mabi=spe and -mabi=no-spe.
3389 2018-06-08 Martin Liska <mliska@suse.cz>
3391 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3392 where we expect an existing summary.
3394 2018-06-08 Martin Liska <mliska@suse.cz>
3396 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3397 * ipa-inline.h (estimate_edge_growth): Likewise.
3399 2018-06-08 Martin Liska <mliska@suse.cz>
3401 * cgraph.c (function_version_hasher::hash): Use
3402 cgraph_node::get_uid ().
3403 (function_version_hasher::equal):
3404 * cgraph.h (cgraph_node::get_uid): New method.
3405 * ipa-inline.c (update_caller_keys): Use
3406 cgraph_node::get_uid ().
3407 (update_callee_keys): Likewise.
3408 * ipa-utils.c (searchc): Likewise.
3409 (ipa_reduced_postorder): Likewise.
3410 * lto-cgraph.c (input_node): Likewise.
3411 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3412 * symbol-summary.h (symtab_insertion): Likewise.
3413 (symtab_removal): Likewise.
3414 (symtab_duplication): Likewise.
3415 * tree-pretty-print.c (dump_function_header): Likewise.
3416 * tree-sra.c (convert_callers_for_node): Likewise.
3418 2018-06-08 Martin Liska <mliska@suse.cz>
3420 * cgraph.c (symbol_table::create_edge): Always assign a new
3422 (symbol_table::free_edge): Do not recycle numbers.
3423 * cgraph.h (cgraph_edge::get): New method.
3424 * symbol-summary.h (symtab_removal): Use it.
3425 (symtab_duplication): Likewise.
3426 (call_summary::hashable_uid): Remove.
3428 2018-06-08 Martin Liska <mliska@suse.cz>
3430 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3431 (initialize_growth_caches): Remove.
3432 (free_growth_caches): Likewise.
3433 (do_estimate_edge_time): Use edge_growth_cache.
3434 (do_estimate_edge_size): Likewise.
3435 (do_estimate_edge_hints): Likewise.
3436 * ipa-inline.c (reset_edge_caches): Likewise.
3437 (recursive_inlining): Likewise.
3438 (inline_small_functions): Likewise.
3439 * ipa-inline.h (initialize_growth_caches): Remove.
3440 (estimate_edge_size): Likewise.
3441 (estimate_edge_time): Likewise.
3442 (estimate_edge_hints): Likewise.
3443 (reset_edge_growth_cache): Likewise.
3444 * symbol-summary.h (call_summary::remove): New method.
3446 2018-06-08 Martin Liska <mliska@suse.cz>
3448 * ipa-cp.c (class edge_clone_summary): New summary.
3449 (grow_edge_clone_vectors): Remove.
3450 (ipcp_edge_duplication_hook): Remove.
3451 (class edge_clone_summary_t): New call_summary class.
3452 (ipcp_edge_removal_hook): Remove.
3453 (edge_clone_summary_t::duplicate): New function.
3454 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3455 (create_specialized_node): Likewise.
3456 (ipcp_driver): Initialize edge_clone_summaries and do not
3459 2018-06-08 Martin Liska <mliska@suse.cz>
3461 * symbol-summary.h (get): New function.
3462 (call_summary::m_initialize_when_cloning): New class member.
3464 2018-06-08 Martin Liska <mliska@suse.cz>
3466 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3467 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3468 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3469 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3470 (remove_cgraph_node_from_order): Use the removed_nodes set.
3471 (do_per_function_toporder): Likwise.
3472 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3473 instead of summary_uid.
3474 (symtab_removal): Likewise.
3475 (symtab_duplication): Likewise.
3477 2018-06-08 Martin Liska <mliska@suse.cz>
3479 * ipa-cp.c (ipcp_store_bits_results): Use
3480 ipcp_transformation_sum.
3481 (ipcp_store_vr_results): Likewise.
3482 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3484 (ipcp_transformation_initialize): ... this.
3485 (ipa_set_node_agg_value_chain):
3486 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3487 (write_ipcp_transformation_info): Likewise.
3488 (read_ipcp_transformation_info): Likewise.
3489 (ipcp_update_bits): Likewise.
3490 (ipcp_update_vr): Likewise.
3491 (ipcp_transform_function): Likewise.
3492 * ipa-prop.h: Rename ipcp_transformation_summary to
3493 ipcp_transformation.
3494 (class ipcp_transformation_t): New function summary.
3495 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3496 (ipa_get_agg_replacements_for_node): Likewise.
3498 2018-06-08 Martin Liska <mliska@suse.cz>
3500 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3502 (class funct_state_summary_t): New function_summary class.
3503 (has_function_state): Remove.
3504 (get_function_state): Likewise.
3505 (set_function_state): Likewise.
3506 (add_new_function): Likewise.
3507 (funct_state_summary_t::insert): New function.
3508 (duplicate_node_data): Remove.
3509 (remove_node_data): Remove.
3510 (funct_state_summary_t::duplicate): New function.
3511 (register_hooks): Create new funct_state_summaries.
3512 (pure_const_generate_summary): Use it.
3513 (pure_const_write_summary): Likewise.
3514 (pure_const_read_summary): Likewise.
3515 (propagate_pure_const): Likewise.
3516 (propagate_nothrow): Likewise.
3517 (dump_malloc_lattice): Likewise.
3518 (propagate_malloc): Likewise.
3519 (execute): Do not register hooks, just remove summary
3521 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3524 2018-06-08 Martin Liska <mliska@suse.cz>
3526 * ipa-reference.c (remove_node_data): Remove.
3527 (duplicate_node_data): Likewise.
3528 (class ipa_ref_var_info_summary_t): New class.
3529 (class ipa_ref_opt_summary_t): Likewise.
3530 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3531 (get_reference_optimization_summary): Use
3532 ipa_ref_opt_sum_summaries.
3533 (set_reference_vars_info): Remove.
3534 (set_reference_optimization_summary): Likewise.
3535 (ipa_init): Create summaries.
3536 (init_function_info): Use function summary.
3537 (ipa_ref_opt_summary_t::duplicate): New function.
3538 (ipa_ref_opt_summary_t::remove): New function.
3539 (get_read_write_all_from_node): Fix GNU coding style.
3540 (propagate): Use function summary.
3541 (write_node_summary_p): Fix GNU coding style.
3542 (stream_out_bitmap): Likewise.
3543 (ipa_reference_read_optimization_summary): Use function summary.
3544 (ipa_reference_c_finalize): Do not release hooks.
3546 2018-06-08 Martin Liska <mliska@suse.cz>
3548 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3549 (analyze_function_body): Extract multiple calls of get_create.
3550 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3551 * ipa-inline.c (recursive_inlining): Use ::get method.
3552 * ipa-inline.h (estimate_edge_growth): Likewise.
3554 2018-06-08 Martin Liska <mliska@suse.cz>
3556 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3558 (hsa_function_summary::hsa_function_summary): Use the new enum
3560 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3561 * hsa-gen.c (hsa_get_host_function): Likewise.
3562 (get_brig_function_name): Likewise.
3563 * ipa-hsa.c (process_hsa_functions): Likewise.
3564 (ipa_hsa_write_summary): Likewise.
3565 * symbol-summary.h (symtab_duplication): Use ::get function/
3566 (get): New function.
3568 2018-06-08 Martin Liska <mliska@suse.cz>
3570 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3572 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3573 (hsa_register_kernel): Likewise.
3574 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3575 * hsa-gen.c (hsa_get_host_function): Likewise.
3576 (get_brig_function_name): Likewise.
3577 (generate_hsa): Likewise.
3578 (pass_gen_hsail::execute): Likewise.
3579 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3580 (devirtualization_time_bonus): Likewise.
3581 (ipcp_propagate_stage): Likewise.
3582 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3583 (edge_set_predicate): Likewise.
3584 (evaluate_conditions_for_known_args): Likewise.
3585 (evaluate_properties_for_edge): Likewise.
3586 (ipa_fn_summary::reset): Likewise.
3587 (ipa_fn_summary_t::duplicate): Likewise.
3588 (dump_ipa_call_summary): Likewise.
3589 (ipa_dump_fn_summary): Likewise.
3590 (analyze_function_body): Likewise.
3591 (compute_fn_summary): Likewise.
3592 (estimate_edge_devirt_benefit): Likewise.
3593 (estimate_edge_size_and_time): Likewise.
3594 (estimate_calls_size_and_time): Likewise.
3595 (estimate_node_size_and_time): Likewise.
3596 (inline_update_callee_summaries): Likewise.
3597 (remap_edge_change_prob): Likewise.
3598 (remap_edge_summaries): Likewise.
3599 (ipa_merge_fn_summary_after_inlining): Likewise.
3600 (ipa_update_overall_fn_summary): Likewise.
3601 (read_ipa_call_summary): Likewise.
3602 (inline_read_section): Likewise.
3603 (write_ipa_call_summary): Likewise.
3604 (ipa_fn_summary_write): Likewise.
3605 (ipa_free_fn_summary): Likewise.
3606 * ipa-hsa.c (process_hsa_functions): Likewise.
3607 (ipa_hsa_write_summary): Likewise.
3608 (ipa_hsa_read_section): Likewise.
3609 * ipa-icf.c (sem_function::merge): Likewise.
3610 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3611 (do_estimate_edge_time): Likewise.
3612 (estimate_size_after_inlining): Likewise.
3613 (estimate_growth): Likewise.
3614 (growth_likely_positive): Likewise.
3615 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3616 (inline_call): Likewise.
3617 * ipa-inline.c (caller_growth_limits): Likewise.
3618 (can_inline_edge_p): Likewise.
3619 (can_inline_edge_by_limits_p): Likewise.
3620 (compute_uninlined_call_time): Likewise.
3621 (compute_inlined_call_time): Likewise.
3622 (want_inline_small_function_p): Likewise.
3623 (edge_badness): Likewise.
3624 (update_caller_keys): Likewise.
3625 (update_callee_keys): Likewise.
3626 (recursive_inlining): Likewise.
3627 (inline_small_functions): Likewise.
3628 (inline_to_all_callers_1): Likewise.
3629 (dump_overall_stats): Likewise.
3630 (early_inline_small_functions): Likewise.
3631 (early_inliner): Likewise.
3632 * ipa-inline.h (estimate_edge_growth): Likewise.
3633 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3634 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3635 * ipa-prop.h (IPA_NODE_REF): Likewise.
3636 (IPA_EDGE_REF): Likewise.
3637 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3638 (propagate_malloc): Likewise.
3639 * ipa-split.c (execute_split_functions): Likewise.
3640 * symbol-summary.h: Rename get to get_create.
3642 (get_create): Likewise.
3643 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3645 2018-06-08 Martin Liska <mliska@suse.cz>
3647 * symbol-summary.h (release): Move definition out of class
3649 (symtab_removal): Likewise.
3650 (symtab_duplication): Likewise.
3652 2018-06-08 Martin Liska <mliska@suse.cz>
3654 * symbol-summary.h (function_summary): Move constructor
3655 implementation out of class declaration.
3656 (release): Likewise.
3657 (symtab_insertion): Likewise.
3658 (symtab_removal): Likewise.
3659 (symtab_duplication): Likewise.
3662 2018-06-08 Martin Liska <mliska@suse.cz>
3664 * Makefile.in: Remove support for MPX (macros, related functions,
3665 fields in cgraph_node, ...).
3666 * builtin-types.def (BT_BND): Likewise.
3667 (BT_FN_BND_CONST_PTR): Likewise.
3668 (BT_FN_CONST_PTR_BND): Likewise.
3669 (BT_FN_VOID_PTR_BND): Likewise.
3670 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3671 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3672 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3673 (expand_builtin_mempcpy_with_bounds): Likewise.
3674 (expand_builtin_memset_with_bounds): Likewise.
3675 (expand_builtin_memset_args): Likewise.
3676 (std_expand_builtin_va_start): Likewise.
3677 (expand_builtin): Likewise.
3678 (expand_builtin_with_bounds): Likewise.
3679 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3680 (DEF_LIB_BUILTIN_CHKP): Likewise.
3681 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3682 (DEF_CHKP_BUILTIN): Likewise.
3683 (BUILT_IN_MEMCPY): Likewise.
3684 (BUILT_IN_MEMMOVE): Likewise.
3685 (BUILT_IN_MEMPCPY): Likewise.
3686 (BUILT_IN_MEMSET): Likewise.
3687 (BUILT_IN_STPCPY): Likewise.
3688 (BUILT_IN_STRCAT): Likewise.
3689 (BUILT_IN_STRCHR): Likewise.
3690 (BUILT_IN_STRCPY): Likewise.
3691 (BUILT_IN_STRLEN): Likewise.
3692 (BUILT_IN_MEMCPY_CHK): Likewise.
3693 (BUILT_IN_MEMMOVE_CHK): Likewise.
3694 (BUILT_IN_MEMPCPY_CHK): Likewise.
3695 (BUILT_IN_MEMSET_CHK): Likewise.
3696 (BUILT_IN_STPCPY_CHK): Likewise.
3697 (BUILT_IN_STRCAT_CHK): Likewise.
3698 (BUILT_IN_STRCPY_CHK): Likewise.
3699 * calls.c (store_bounds): Likewise.
3700 (emit_call_1): Likewise.
3701 (special_function_p): Likewise.
3702 (maybe_warn_nonstring_arg): Likewise.
3703 (initialize_argument_information): Likewise.
3704 (finalize_must_preallocate): Likewise.
3705 (compute_argument_addresses): Likewise.
3706 (expand_call): Likewise.
3707 * cfgexpand.c (expand_call_stmt): Likewise.
3708 (expand_return): Likewise.
3709 (expand_gimple_stmt_1): Likewise.
3710 (pass_expand::execute): Likewise.
3711 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3712 (cgraph_node::remove): Likewise.
3713 (cgraph_node::dump): Likewise.
3714 (cgraph_node::verify_node): Likewise.
3715 * cgraph.h (chkp_function_instrumented_p): Likewise.
3716 (symtab_node::get_alias_target): Likewise.
3717 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3718 (cgraph_local_p): Likewise.
3719 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3720 (cgraph_edge::rebuild_references): Likewise.
3721 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3722 (walk_polymorphic_call_targets): Likewise.
3723 (cgraph_node::expand_thunk): Likewise.
3724 (symbol_table::output_weakrefs): Likewise.
3725 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3726 (ix86_handle_option): Likewise.
3727 * config/i386/constraints.md: Likewise.
3728 * config/i386/i386-builtin-types.def (BND): Likewise.
3732 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3733 (BDESC_FIRST): Likewise.
3735 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3736 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3737 * config/i386/i386.c (enum reg_class): Likewise.
3738 (ix86_target_string): Likewise.
3739 (ix86_option_override_internal): Likewise.
3740 (ix86_conditional_register_usage): Likewise.
3741 (ix86_valid_target_attribute_inner_p): Likewise.
3742 (ix86_set_indirect_branch_type): Likewise.
3743 (ix86_set_current_function): Likewise.
3744 (ix86_function_arg_regno_p): Likewise.
3745 (init_cumulative_args): Likewise.
3746 (ix86_function_arg_advance): Likewise.
3747 (ix86_function_arg): Likewise.
3748 (ix86_pass_by_reference): Likewise.
3749 (ix86_function_value_regno_p): Likewise.
3750 (ix86_function_value_1): Likewise.
3751 (ix86_function_value_bounds): Likewise.
3752 (ix86_return_in_memory): Likewise.
3753 (ix86_setup_incoming_vararg_bounds): Likewise.
3754 (ix86_va_start): Likewise.
3755 (indirect_thunk_need_prefix): Likewise.
3756 (print_reg): Likewise.
3757 (ix86_print_operand): Likewise.
3758 (ix86_expand_call): Likewise.
3759 (ix86_output_function_return): Likewise.
3760 (reg_encoded_number): Likewise.
3761 (BDESC_VERIFYS): Likewise.
3762 (ix86_init_mpx_builtins): Likewise.
3763 (ix86_init_builtins): Likewise.
3764 (ix86_emit_cmove): Likewise.
3765 (ix86_emit_move_max): Likewise.
3766 (ix86_expand_builtin): Likewise.
3767 (ix86_builtin_mpx_function): Likewise.
3768 (ix86_get_arg_address_for_bt): Likewise.
3769 (ix86_load_bounds): Likewise.
3770 (ix86_store_bounds): Likewise.
3771 (ix86_load_returned_bounds): Likewise.
3772 (ix86_store_returned_bounds): Likewise.
3773 (ix86_class_likely_spilled_p): Likewise.
3774 (ix86_hard_regno_mode_ok): Likewise.
3775 (x86_order_regs_for_local_alloc): Likewise.
3776 (ix86_mitigate_rop): Likewise.
3777 (ix86_bnd_prefixed_insn_p): Likewise.
3778 (ix86_mpx_bound_mode): Likewise.
3779 (ix86_make_bounds_constant): Likewise.
3780 (ix86_initialize_bounds): Likewise.
3781 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3782 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3783 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3784 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3785 (TARGET_CHKP_BOUND_MODE): Likewise.
3786 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3787 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3788 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3789 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3790 * config/i386/i386.h (TARGET_MPX): Likewise.
3791 (TARGET_MPX_P): Likewise.
3792 (VALID_BND_REG_MODE): Likewise.
3793 (FIRST_BND_REG): Likewise.
3794 (LAST_BND_REG): Likewise.
3795 (enum reg_class): Likewise.
3796 (BND_REG_P): Likewise.
3797 (BND_REGNO_P): Likewise.
3798 (BNDmode): Likewise.
3799 (ADJUST_INSN_LENGTH): Likewise.
3800 * config/i386/i386.md: Likewise.
3801 * config/i386/i386.opt: Likewise.
3802 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3803 (defined): Likewise.
3804 (LINK_MPX): Likewise.
3805 (MPX_SPEC): Likewise.
3806 (LIBMPX_SPEC): Likewise.
3807 (LIBMPXWRAPPERS_SPEC): Likewise.
3808 (CHKP_SPEC): Likewise.
3809 * config/i386/predicates.md: Likewise.
3810 * dbxout.c (dbxout_type): Likewise.
3811 * doc/extend.texi: Likewise.
3812 * doc/invoke.texi: Likewise.
3813 * doc/md.texi: Likewise.
3814 * doc/tm.texi: Likewise.
3815 * doc/tm.texi.in: Likewise.
3816 * dwarf2out.c (is_base_type): Likewise.
3817 (gen_formal_types_die): Likewise.
3818 (gen_subprogram_die): Likewise.
3819 (gen_type_die_with_usage): Likewise.
3820 (gen_decl_die): Likewise.
3821 (dwarf2out_late_global_decl): Likewise.
3822 * expr.c (expand_assignment): Likewise.
3823 (emit_storent_insn): Likewise.
3824 (store_expr_with_bounds): Likewise.
3825 (store_expr): Likewise.
3826 (expand_expr_real_1): Likewise.
3827 * expr.h (store_expr_with_bounds): Likewise.
3828 * function.c (use_register_for_decl): Likewise.
3829 (struct bounds_parm_data): Likewise.
3830 (assign_parms_augmented_arg_list): Likewise.
3831 (assign_parm_find_entry_rtl): Likewise.
3832 (assign_parm_is_stack_parm): Likewise.
3833 (assign_parm_load_bounds): Likewise.
3834 (assign_bounds): Likewise.
3835 (assign_parms): Likewise.
3836 (expand_function_start): Likewise.
3837 * gcc.c (CHKP_SPEC): Likewise.
3838 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3839 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3840 (wrestrict_dom_walker::check_call): Likewise.
3841 * gimple.c (gimple_build_call_from_tree): Likewise.
3842 * gimple.h (enum gf_mask): Likewise.
3843 (gimple_call_with_bounds_p): Likewise.
3844 (gimple_call_set_with_bounds): Likewise.
3845 * gimplify.c (gimplify_init_constructor): Likewise.
3846 * ipa-cp.c (initialize_node_lattices): Likewise.
3847 (propagate_constants_across_call): Likewise.
3848 (find_more_scalar_values_for_callers_subset): Likewise.
3849 * ipa-hsa.c (process_hsa_functions): Likewise.
3850 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3851 * ipa-icf.c (sem_function::merge): Likewise.
3852 * ipa-inline.c (early_inliner): Likewise.
3853 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3854 (warn_function_cold): Likewise.
3855 (propagate_pure_const): Likewise.
3856 * ipa-ref.h (enum GTY): Likewise.
3857 * ipa-split.c (find_retbnd): Likewise.
3858 (consider_split): Likewise.
3859 (split_function): Likewise.
3860 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3861 * ipa.c (walk_polymorphic_call_targets): Likewise.
3862 (symbol_table::remove_unreachable_nodes): Likewise.
3863 (process_references): Likewise.
3864 (cgraph_build_static_cdtor_1): Likewise.
3865 * lto-cgraph.c (lto_output_node): Likewise.
3866 (output_refs): Likewise.
3867 (compute_ltrans_boundary): Likewise.
3868 (input_overwrite_node): Likewise.
3869 (input_node): Likewise.
3870 (input_cgraph_1): Likewise.
3871 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3872 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3873 (class pass_chkp_instrumentation_passes): Likewise.
3874 (make_pass_chkp_instrumentation_passes): Likewise.
3875 * passes.def: Likewise.
3876 * rtl.h (struct GTY): Likewise.
3877 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3878 * stor-layout.c (layout_type): Likewise.
3879 * symtab.c: Likewise.
3880 * target.def: Likewise.
3881 * targhooks.c (default_chkp_bound_type): Likewise.
3882 (default_chkp_bound_mode): Likewise.
3883 (default_builtin_chkp_function): Likewise.
3884 (default_chkp_function_value_bounds): Likewise.
3885 (default_chkp_make_bounds_constant): Likewise.
3886 (default_chkp_initialize_bounds): Likewise.
3887 * targhooks.h (default_chkp_bound_type): Likewise.
3888 (default_chkp_bound_mode): Likewise.
3889 (default_builtin_chkp_function): Likewise.
3890 (default_chkp_function_value_bounds): Likewise.
3891 (default_chkp_make_bounds_constant): Likewise.
3892 (default_chkp_initialize_bounds): Likewise.
3893 * toplev.c (compile_file): Likewise.
3894 (process_options): Likewise.
3895 * tree-core.h (DEF_BUILTIN): Likewise.
3896 (DEF_BUILTIN_CHKP): Likewise.
3897 * tree-inline.c (declare_return_variable): Likewise.
3898 (remap_gimple_stmt): Likewise.
3899 (copy_bb): Likewise.
3900 (initialize_inlined_parameters): Likewise.
3901 (expand_call_inline): Likewise.
3902 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3903 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3904 (make_pass_ipa_chkp_produce_thunks): Likewise.
3905 (make_pass_chkp): Likewise.
3906 (make_pass_chkp_opt): Likewise.
3907 (make_pass_chkp_instrumentation_passes): Likewise.
3908 * tree-pretty-print.c (dump_generic_node): Likewise.
3909 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3910 * tree-ssa-dce.c (propagate_necessity): Likewise.
3911 (eliminate_unnecessary_stmts): Likewise.
3912 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3913 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3914 * tree-ssa-sccvn.h: Likewise.
3915 * tree-ssa-strlen.c (get_string_length): Likewise.
3916 (valid_builtin_call): Likewise.
3917 (adjust_last_stmt): Likewise.
3918 (handle_builtin_strchr): Likewise.
3919 (handle_builtin_strcpy): Likewise.
3920 (handle_builtin_stxncpy): Likewise.
3921 (handle_builtin_memcpy): Likewise.
3922 (handle_builtin_strcat): Likewise.
3923 (strlen_check_and_optimize_stmt): Likewise.
3924 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3925 * tree-streamer-in.c: Likewise.
3926 * tree-streamer.c (record_common_node): Likewise.
3927 * tree.c (tree_code_size): Likewise.
3928 (wide_int_to_tree_1): Likewise.
3929 (type_contains_placeholder_1): Likewise.
3930 (build_common_tree_nodes): Likewise.
3931 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3932 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3933 (POINTER_BOUNDS_P): Likewise.
3934 (BOUNDED_TYPE_P): Likewise.
3935 (BOUNDED_P): Likewise.
3936 (CALL_WITH_BOUNDS_P): Likewise.
3937 (pointer_bounds_type_node): Likewise.
3938 * value-prof.c (gimple_ic): Likewise.
3939 * var-tracking.c (vt_add_function_parameters): Likewise.
3940 * varasm.c (make_decl_rtl): Likewise.
3941 (assemble_start_function): Likewise.
3942 (output_constant): Likewise.
3943 (maybe_assemble_visibility): Likewise.
3944 * varpool.c (ctor_for_folding): Likewise.
3945 * chkp-builtins.def: Remove.
3946 * ipa-chkp.c: Remove.
3947 * ipa-chkp.h: Remove.
3948 * rtl-chkp.c: Remove.
3949 * rtl-chkp.h: Remove.
3950 * tree-chkp-opt.c: Remove.
3951 * tree-chkp.c: Remove.
3952 * tree-chkp.h: Remove.
3954 2018-06-07 Carl Love <cel@us.ibm.com>
3956 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3957 vextract_fp_from_shortl): Add BE support.
3959 2018-06-07 Paul Koning <ni1d@arrl.net>
3961 * compare-elim.c (try_merge_compare): Don't merge compare if
3962 address contains a side effect.
3963 (try_eliminate_compare): Likewise.
3965 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3967 * config.gcc: Support "tremont".
3968 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3969 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3971 * config/i386/i386.c (m_TREMONT): Define.
3972 (processor_target_table): Add "tremont".
3973 (PTA_TREMONT): Define.
3974 (ix86_lea_outperforms): Add TARGET_TREMONT.
3975 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3976 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3977 and M_INTEL_GOLDMONT_PLUS.
3978 (fold_builtin_cpu): Add "tremont".
3979 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3980 (ix86_option_override_internal): Add "tremont".
3981 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3982 (processor_type): Add PROCESSOR_TREMONT.
3983 * config/i386/x86-tune.def: Add m_TREMONT.
3984 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3986 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3988 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3989 symbol defined for msp430i* devices to be lower case.
3991 gcc/testsuite/gcc.target/msp430/
3992 * msp430i-device-symbol.c: New test.
3993 * msp430f-device-symbol.c: New test.
3994 * msp430.h: New test header file.
3996 2018-06-07 Richard Biener <rguenther@suse.de>
3998 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3999 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4000 Properly wrap signed arithmetic if overflow wraps.
4002 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4004 PR tree-optimization/69615
4005 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4006 of a cast from a same precision integral SSA_NAME in a bb dominated
4007 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4008 cast to utype if rhs2 has already a compatible type.
4010 2018-06-07 Richard Biener <rguenther@suse.de>
4012 PR tree-optimization/85935
4013 * graphite-scop-detection.c (find_params_in_bb): Analyze
4014 condition operands with respect to the correct loop. Assert
4015 the analysis doesn't fail.
4017 2018-06-04 Carl Love <cel@us.ibm.com>
4019 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4020 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4021 as it is slightly cheaper.
4022 (first_match_or_eos_index_<mode>):
4023 Calculate index using natural element order.
4024 (first_match_index_<mode>):
4025 Calculate index using natural element order.
4026 (first_match_or_eos_index_<mode>):
4027 Calculate index using natural order.
4028 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4029 for BE and LE modes.
4030 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4031 P9V_BUILTIN_VCLZLSBB_V16QI.
4032 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4035 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4037 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4038 indentation and line wrap for many prototypes. Add missing
4039 @smallexample directives around block of prototypes for vec_xl and
4042 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4044 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4045 track if we pass or return IEEE 128-bit floating point.
4046 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4047 C++ mangling that is compatible with GCC 8.1.
4048 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4049 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4050 floating point types.
4051 (rs6000_function_arg_advance_1): Likewise.
4052 (rs6000_mangle_type): Optionally generate mangled names that match
4053 what GCC 8.1 generated for IEEE 128-bit floating point types.
4054 (rs6000_globalize_decl_name): If we have an external function that
4055 passes or returns IEEE 128-bit types, generate a weak reference
4056 from the mangled name used in GCC 8.1 to the current mangled
4058 (rs6000_init_builtins): Make __ibm128 use the long double type if
4059 long double is IBM extended double. Make __float128 use the long
4060 double type if long double is IEEE 128-bit.
4063 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4064 macro for __ibm128 built-in functions.
4065 (PACK_IF): Add __ibm128 pack/unpack functions.
4066 (UNPACK_IF): Likewise.
4067 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4068 enable long double built-in functions if long double is IEEE
4069 128-bit floating point.
4070 (rs6000_invalid_builtin): Update long double built-in function
4072 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4073 functions, adjust the built-in function to use the long double
4074 built-in function if __ibm128 and long double are the same type.
4075 * doc/extend.texi (PowerPC builtins): Update documention for
4076 __builtin_{,un}pack_longdouble. Add documentation for
4077 __builtin_{,un}pack_ibm128.
4079 2018-06-06 Jim Wilson <jimw@sifive.com>
4081 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4082 (struct machine_function): New field interrupt_mode.
4083 (riscv_handle_type_attribute): New function. Add forward declaration.
4084 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4085 (riscv_expand_epilogue): Check interrupt_mode field.
4086 (riscv_set_current_function): Check interrupt attribute args and
4087 set interrupt_mode field.
4088 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4089 (riscv_sret, riscv_uret): New.
4090 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4091 new arguments to interrupt attribute.
4093 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4096 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4097 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4099 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4101 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4102 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4103 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4104 several redundant entries.
4106 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4108 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4109 type from "rtx" to "rtx_insn *".
4110 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4111 for local "call_insn", removing cast.
4112 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4114 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4116 PR tree-optimization/86066
4117 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4118 for BIT_INSERT_EXPR stores.
4120 2018-06-06 Richard Biener <rguenther@suse.de>
4122 PR tree-optimization/86062
4123 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4124 component refs ontop
4125 of to be offsetted base.
4127 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4129 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4130 to be static and remove check on interrupt attribute name.
4132 gcc/testsuite/gcc.target/msp430/
4133 * function-attributes-4.c: New test.
4134 * static-interrupts.c: New test.
4136 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4138 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4139 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4141 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4144 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4146 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4147 Remove second argument, change how error is called.
4148 (aarch64_layout_arg): Remove second argument from
4149 aarch64_err_no_fpadvsimd call.
4150 (aarch64_init_cumulative_args): Ditto.
4151 (aarch64_gimplify_va_arg_expr): Ditto.
4152 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4154 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4156 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4157 (*simple_return_indirect_internal<mode>): Rename from
4158 simple_return_indirect_internal. Use W mode iterator.
4159 (rstorssp): New expander.
4160 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4161 (clrssbsy): New expander.
4162 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4164 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4166 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4168 (cmse_check_pointed_object): Likewise.
4170 2018-06-05 Martin Liska <mliska@suse.cz>
4172 PR gcov-profile/47618
4173 * doc/invoke.texi: Document how -fprofile-dir format
4176 2018-06-05 Richard Biener <rguenther@suse.de>
4178 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4179 removal pretend DOM info isn't available so we do not update
4180 it and only remove edges, not dominated blocks. Actually free
4181 DOM info in case we removed something. Remove unreachable blocks.
4182 (mfb_keep_latches): Work with either DOM info or marked backedges.
4183 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4184 first. Mark backedges if DOM info isn't available.
4185 (Re-)compute DOM info after cleanup_control_flow_pre.
4187 2018-06-05 Richard Biener <rguenther@suse.de>
4189 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4190 (locus_discrim_hasher::hash): Adjust.
4191 (locus_discrim_hasher::equal): Likewise.
4192 (next_discriminator_for_locus): Work on line directly.
4193 (same_line_p): Pass in expanded locus1 as well.
4194 (assign_discriminators): Avoid redundant location expansions.
4196 2018-06-05 Richard Biener <rguenther@suse.de>
4198 PR tree-optimization/86046
4199 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4200 if required after clearing TREE_ADDRESSABLE.
4202 2018-06-05 Richard Biener <rguenther@suse.de>
4204 PR tree-optimization/86047
4205 * tree-ssa-loop.c (for_each_index): Glob handling of all
4206 decls and constants and really handle all of them.
4208 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4211 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4212 qualifier_void_pointer and qualifier_const_void_pointer.
4213 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4214 (arm_init_builtins): Handle the above.
4215 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4216 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4217 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4220 2018-06-05 Martin Liska <mliska@suse.cz>
4222 * auto-profile.c (read_autofdo_file): Do not use
4223 gcov_ctr_summary struct.
4224 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4225 * coverage.c (struct counts_entry): Likewise.
4226 (read_counts_file): Read just single summary entry.
4227 (get_coverage_counts): Use gcov_summary struct.
4228 * coverage.h (get_coverage_counts): Likewise.
4229 * gcov-dump.c (dump_working_sets): Likewise.
4230 (tag_summary): Dump just single summary.
4231 * gcov-io.c (gcov_write_summary): Write just histogram
4233 (gcov_read_summary): Read just single summary.
4234 (compute_working_sets): Use gcov_summary struct.
4235 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4236 of GCOV_COUNTERS_SUMMABLE.
4237 (GCOV_COUNTERS_SUMMABLE): Remove.
4238 (GCOV_FIRST_VALUE_COUNTER): Replace with
4239 GCOV_COUNTER_V_INTERVAL.
4240 (struct gcov_ctr_summary): Remove.
4241 (struct gcov_summary): Directly use fields of former
4243 (compute_working_sets): Use gcov_summary struct.
4244 * gcov.c (read_count_file): Do not use ctrs fields.
4245 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4247 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4249 * profile.c: Likewise.
4250 * profile.h: Likewise.
4252 2018-06-05 Martin Liska <mliska@suse.cz>
4254 PR gcov-profile/84846
4255 * gcov.c (output_lines): Print working directory only
4256 in intermediate format.
4258 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4260 * config/s390/s390-builtin-types.def: Add void function type.
4261 * config/s390/s390-builtins.def: Use the function type for the
4264 2018-06-04 Jim Wilson <jimw@sifive.com>
4266 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4268 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4269 and maybe_eh_return. Change regno to unsigned int. Use new args to
4270 handle EH_RETURN_DATA_REGNO registers properly.
4271 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4272 (riscv_expand_epilogue): Update comment. Change argument name and
4273 type. Update code to use new name and type. Pass new args to
4274 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4276 * config/riscv/riscv.md (NORMAL_RETURN): New.
4277 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4278 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4279 (eh_return): Call gen_eh_return_internal and emit barrier.
4280 (eh_return_internal): Call riscv_expand_epilogue.
4282 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4284 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4285 bit_insertion field and declare can_be_merged_into method.
4286 (merged_store_group::can_be_merged_into): New method.
4287 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4288 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4289 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4291 2018-06-04 Richard Biener <rguenther@suse.de>
4293 PR tree-optimization/85955
4294 * builtins.c (fold_builtin_sincos): Convert pointers to
4295 destination to appropriate type before dereferencing.
4297 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4299 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4301 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4303 * expr.c (expand_expr_real_1): Force the operand into memory if
4304 its TYPE_MODE is BLKmode and if there is no integer mode for
4305 the number of bits being extracted.
4307 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4311 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4312 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4314 2018-06-04 Richard Biener <rguenther@suse.de>
4316 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4317 (cleanup_tree_cfg_noloop): ... single caller. Do
4318 start_recording_case_labels later.
4320 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4322 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4323 to _IMMINTRIN_H_INCLUDED.
4324 * config/i386/pconfigintrin.h: Ditto.
4325 * config/i386/waitpkgintrin.h: Ditto.
4326 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4327 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4328 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4329 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4330 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4331 waitpkgintrin.h and cldemoteintrin.h.
4333 2018-06-04 Richard Biener <rguenther@suse.de>
4335 PR tree-optimization/86038
4336 * tracer.c (find_best_successor): Check probability for
4337 being initialized, bail out if not.
4339 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4342 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4343 of bits to ignore when comparing architectures.
4345 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4347 PR tree-optimization/69615
4348 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4349 maximum or minimum of the type, try to merge it also as if
4350 range1 is + [-, x - 1] or + [x + 1, -].
4353 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4355 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4357 PR tree-optimization/86034
4358 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4359 the unsigned bitfield type in a bit insertion sequence if it does not
4360 have a larger precision than the bitfield size.
4361 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4363 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4365 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4367 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4369 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4370 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4371 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4372 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4374 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4376 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4377 Disable -fdelete-null-pointer-checks for ELF toolchain.
4379 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4380 Kito Cheng <kito.cheng@gmail.com>
4382 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4383 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4384 (nds32*-*-*): Add glibc and uclibc conditions.
4385 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4386 (TARGET_EXCEPT_UNWIND_INFO): Define.
4387 * config/nds32/elf.h: New file.
4388 * config/nds32/linux.h: New file.
4389 * config/nds32/nds32-elf.opt: New file.
4390 * config/nds32/nds32-linux.opt: New file.
4391 * config/nds32/nds32-fp-as-gp.c
4392 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4393 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4395 (nds32_asm_file_end): Ditto.
4396 (nds32_print_operand): Ditto.
4397 (nds32_insert_attributes): Ditto.
4398 (nds32_init_libfuncs): New function.
4399 (TARGET_HAVE_TLS): Define.
4400 (TARGET_INIT_LIBFUNCS): Define.
4401 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4403 (TARGET_ELF): Apply different mcmodel setting.
4404 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4405 been migrated into elf.h and linux.h files.
4406 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4407 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4408 (mcmodel): The content has been migrated into nds32-elf.opt and
4409 nds32-linux.opt files.
4410 * config/nds32/t-elf: New file.
4411 * config/nds32/t-linux: New file.
4413 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4414 Shiva Chen <shiva0217@gmail.com>
4416 * config/nds32/constants.md (unspec_volatile_element): Add
4417 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4418 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4420 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4421 (make_pass_nds32_fp_as_gp): Declare.
4422 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4424 (nds32_asm_function_end_prologue): Remove unused asm output.
4425 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4426 (nds32_asm_file_start): Output necessary fp_as_gp information.
4427 (nds32_option_override): Adjust register usage.
4428 (nds32_expand_prologue): Consider fp_as_gp situation.
4429 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4430 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4433 (simple_return): Ditto.
4434 (omit_fp_begin): Output special directive for fp_as_gp.
4435 (omit_fp_end): Output special directive for fp_as_gp.
4436 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4437 mforbid-fp-as-gp): New options.
4439 2018-06-01 Mark Wielaard <mark@klomp.org>
4441 * dwarf2out.c (dwarf2out_finish): Remove generation of
4442 DW_AT_loclists_base.
4444 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4446 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4447 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4448 (struct merged_store_group): Add bit_insertion field.
4449 (dump_char_array): Use standard hexadecimal format.
4450 (merged_store_group::merged_store_group): Set bit_insertion to false.
4451 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4452 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4453 also print the mask in the dump file.
4454 (pass_store_merging::gate): Minor tweak.
4455 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4456 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4457 stores with INTEGER_CST stores.
4458 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4459 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4460 and use it throughout. Generate bit insertion sequences if need be.
4461 (pass_store_merging::process_store): Remove redundant condition.
4462 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4464 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4466 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4467 the 128-bit floating point types. Fix function comment.
4469 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4471 * config/aarch64/aarch64-simd.md
4472 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4474 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4477 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4479 PR tree-optimization/85989
4480 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4482 (backprop::intersect_uses): Check it when deciding whether this
4483 is a backedge reference.
4484 (backprop::process_block): Add each phi to m_visited_phis
4485 after visiting it, then clear it at the end.
4487 2018-06-01 Richard Biener <rguenther@suse.de>
4489 * tree-vectorizer.h (vect_dr_stmt): New function.
4490 (vect_get_load_cost): Adjust.
4491 (vect_get_store_cost): Likewise.
4492 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4493 Use vect_dr_stmt instead of DR_SMTT.
4494 (vect_record_base_alignments): Likewise.
4495 (vect_calculate_target_alignment): Likewise.
4496 (vect_compute_data_ref_alignment): Likewise and make static.
4497 (vect_update_misalignment_for_peel): Likewise.
4498 (vect_verify_datarefs_alignment): Likewise.
4499 (vector_alignment_reachable_p): Likewise.
4500 (vect_get_data_access_cost): Likewise. Pass down
4501 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4502 (vect_get_peeling_costs_all_drs): Likewise.
4503 (vect_peeling_hash_get_lowest_cost): Likewise.
4504 (vect_enhance_data_refs_alignment): Likewise.
4505 (vect_find_same_alignment_drs): Likewise.
4506 (vect_analyze_data_refs_alignment): Likewise.
4507 (vect_analyze_group_access_1): Likewise.
4508 (vect_analyze_group_access): Likewise.
4509 (vect_analyze_data_ref_access): Likewise.
4510 (vect_analyze_data_ref_accesses): Likewise.
4511 (vect_vfa_segment_size): Likewise.
4512 (vect_small_gap_p): Likewise.
4513 (vectorizable_with_step_bound_p): Likewise.
4514 (vect_prune_runtime_alias_test_list): Likewise.
4515 (vect_analyze_data_refs): Likewise.
4516 (vect_supportable_dr_alignment): Likewise.
4517 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4518 (vect_gen_prolog_loop_niters): Likewise.
4519 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4520 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4522 (vect_recog_mask_conversion_pattern): Likewise.
4523 (vect_try_gather_scatter_pattern): Likewise.
4524 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4525 to vect_get_store_cost.
4526 (vect_get_store_cost): Get stmt_info instead of DR.
4527 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4528 (vect_get_load_cost): Get stmt_info instead of DR.
4530 2018-06-01 Richard Biener <rguenther@suse.de>
4533 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4534 vars wrapped in MEM_REFs.
4536 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4538 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4539 Fix subreg tests so that we only return a choice between
4540 GENERAL_REGS and FP_REGS if the original classes included both.
4542 2018-06-01 Richard Biener <rguenther@suse.de>
4545 * tree-ssa-structalias.c (get_function_part_constraint):
4546 Handle NULL fi->decl.
4547 (find_func_aliases_for_call): Properly handle indirect
4548 fi from direct call.
4549 (find_func_clobbers): Likewise.
4550 (ipa_pta_execute): Likewise.
4551 (create_variable_info_for): For functions that are ifunc_resolver
4552 resolve to a varinfo that contains the result of the resolver call.
4553 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4556 2018-05-31 Michael Collison <michael.collison@arm.com>
4558 * config/aarch64/aarch64.md:
4559 (*fix_to_zero_extenddfdi2): New pattern.
4560 * gcc.target/aarch64/fix_extend1.c: New testcase.
4562 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4566 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4567 and BUILT_IN_STRNCMP_EQ.
4568 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4569 BUILT_IN_STRNCMP_EQ.
4570 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4571 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4572 (gimple_fold_builtin): Likewise.
4573 * tree-ssa-strlen.c (compute_string_length): New function.
4574 (determine_min_obsize): New function.
4575 (handle_builtin_string_cmp): New function to handle calls to
4576 string compare functions.
4577 (strlen_optimize_stmt): Add handling to builtin string compare
4579 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4580 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4581 * tree.c (build_common_builtin_nodes): Add new defines of
4582 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4584 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4587 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4588 functions with naked attribute.
4590 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4592 * config/i386/sse.md (avx_vec_concat<mode>):
4593 Substitute concat_tg_mode mode attribute with xtg_mode.
4594 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4595 (concat_tg_mode): Remove mode attribute.
4597 2018-05-31 Martin Sebor <msebor@redhat.com>
4600 * calls.c (alloc_max_size): Correct a logic error/typo.
4601 Treat excessive arguments as infinite. Warn for invalid arguments.
4602 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4604 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4607 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4608 and movx for Haswell.
4610 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4611 Cesar Philippidis <cesar@codesourcery.com>
4614 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4615 when emitting error on private/firstprivate reductions.
4616 * omp-low.c (lower_omp_target): Avoid reference-type processing
4617 on pointers for firstprivate clause.
4619 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4621 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4624 * config/aarch64/aarch64-simd.md
4625 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4626 (aarch64_ld1_x3_<mode>): Likewise
4627 (aarch64_st1x2<VALLDIF:mode>): Likewise
4628 (aarch64_st1_x2_<mode>): Likewise
4629 (aarch64_st1x3<VALLDIF:mode>): Likewise
4630 (aarch64_st1_x3_<mode>): Likewise
4631 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4632 (vld1_s8_x3): Likewise.
4633 (vld1_u16_x3): Likewise.
4634 (vld1_s16_x3): Likewise.
4635 (vld1_u32_x3): Likewise.
4636 (vld1_s32_x3): Likewise.
4637 (vld1_u64_x3): Likewise.
4638 (vld1_s64_x3): Likewise.
4639 (vld1_f16_x3): Likewise.
4640 (vld1_f32_x3): Likewise.
4641 (vld1_f64_x3): Likewise.
4642 (vld1_p8_x3): Likewise.
4643 (vld1_p16_x3): Likewise.
4644 (vld1_p64_x3): Likewise.
4645 (vld1q_u8_x3): Likewise.
4646 (vld1q_s8_x3): Likewise.
4647 (vld1q_u16_x3): Likewise.
4648 (vld1q_s16_x3): Likewise.
4649 (vld1q_u32_x3): Likewise.
4650 (vld1q_s32_x3): Likewise.
4651 (vld1q_u64_x3): Likewise.
4652 (vld1q_s64_x3): Likewise.
4653 (vld1q_f16_x3): Likewise.
4654 (vld1q_f32_x3): Likewise.
4655 (vld1q_f64_x3): Likewise.
4656 (vld1q_p8_x3): Likewise.
4657 (vld1q_p16_x3): Likewise.
4658 (vld1q_p64_x3): Likewise.
4659 (vst1_s64_x2): Likewise.
4660 (vst1_u64_x2): Likewise.
4661 (vst1_f64_x2): Likewise.
4662 (vst1_s8_x2): Likewise.
4663 (vst1_p8_x2): Likewise.
4664 (vst1_s16_x2): Likewise.
4665 (vst1_p16_x2): Likewise.
4666 (vst1_s32_x2): Likewise.
4667 (vst1_u8_x2): Likewise.
4668 (vst1_u16_x2): Likewise.
4669 (vst1_u32_x2): Likewise.
4670 (vst1_f16_x2): Likewise.
4671 (vst1_f32_x2): Likewise.
4672 (vst1_p64_x2): Likewise.
4673 (vst1q_s8_x2): Likewise.
4674 (vst1q_p8_x2): Likewise.
4675 (vst1q_s16_x2): Likewise.
4676 (vst1q_p16_x2): Likewise.
4677 (vst1q_s32_x2): Likewise.
4678 (vst1q_s64_x2): Likewise.
4679 (vst1q_u8_x2): Likewise.
4680 (vst1q_u16_x2): Likewise.
4681 (vst1q_u32_x2): Likewise.
4682 (vst1q_u64_x2): Likewise.
4683 (vst1q_f16_x2): Likewise.
4684 (vst1q_f32_x2): Likewise.
4685 (vst1q_f64_x2): Likewise.
4686 (vst1q_p64_x2): Likewise.
4687 (vst1_s64_x3): Likewise.
4688 (vst1_u64_x3): Likewise.
4689 (vst1_f64_x3): Likewise.
4690 (vst1_s8_x3): Likewise.
4691 (vst1_p8_x3): Likewise.
4692 (vst1_s16_x3): Likewise.
4693 (vst1_p16_x3): Likewise.
4694 (vst1_s32_x3): Likewise.
4695 (vst1_u8_x3): Likewise.
4696 (vst1_u16_x3): Likewise.
4697 (vst1_u32_x3): Likewise.
4698 (vst1_f16_x3): Likewise.
4699 (vst1_f32_x3): Likewise.
4700 (vst1_p64_x3): Likewise.
4701 (vst1q_s8_x3): Likewise.
4702 (vst1q_p8_x3): Likewise.
4703 (vst1q_s16_x3): Likewise.
4704 (vst1q_p16_x3): Likewise.
4705 (vst1q_s32_x3): Likewise.
4706 (vst1q_s64_x3): Likewise.
4707 (vst1q_u8_x3): Likewise.
4708 (vst1q_u16_x3): Likewise.
4709 (vst1q_u32_x3): Likewise.
4710 (vst1q_u64_x3): Likewise.
4711 (vst1q_f16_x3): Likewise.
4712 (vst1q_f32_x3): Likewise.
4713 (vst1q_f64_x3): Likewise.
4714 (vst1q_p64_x3): Likewise.
4716 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4718 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4719 user_label_prefix to name.
4721 * tree-core.h: Update comment about the format of NAME string
4722 passed to handler in attribute_spec.
4724 * config/msp430/msp430.md: Remove erroneous subreg expression from
4725 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4726 zero_extend{q,h}isi2.
4728 2018-05-30 Borislav Petkov <bp@suse.de>
4730 * doc/extend.texi: Document some architecture specific
4731 constraints and sort entries.
4733 2018-05-30 Martin Sebor <msebor@redhat.com>
4736 * builtins.c (expand_builtin_stpcpy_1): New function.
4737 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4738 only if the former succeeds.
4740 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4742 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4745 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4747 * doc/invoke.texi (-flinker-output): Document
4749 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4751 * passes.c (ipa_write_summaries): Only modify statements if body
4753 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4754 incrementally linking.
4755 (ipa_passes): Likewise.
4756 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4757 pass down resolution info.
4758 * common.opt (flag_incremental_link): Update info.
4759 * gcc.c (plugin specs): Turn flinker-output=* to
4760 -plugin-opt=-linker-output-known
4761 * toplev.c (compile_file): Also cut compilation when doing incremental
4763 * flag-types. (enum lto_partition_model): Add
4764 LTO_LINKER_OUTPUT_NOLTOREL.
4765 (invoke.texi): Add -flinker-output docs.
4766 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4767 link same way as WPA; do not stream in dead initializers.
4769 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4770 darf2out_register_external_decl): Support incremental link.
4772 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4774 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4776 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4778 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4779 it down to simple_object_copy_lto_debug_sections.
4780 (run_gcc): Determine incremental LTO link time and configure
4781 lto1 into non-wpa mode, disable renaming of debug sections.
4783 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4785 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4786 descriptions of various incorrectly documented functions.
4788 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4791 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4792 address check not strict.
4794 2018-05-30 Richard Biener <rguenther@suse.de>
4796 PR tree-optimization/85964
4797 * tracer.c (better_p): Drop initialized count check, we only
4798 call the function with initialized counts now.
4799 (find_best_successor): Do find a best edge if one
4800 has uninitialized count.
4801 (find_best_predecessor): Likewise. Do BB frequency check only
4802 if count is initialized.
4804 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4806 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4807 (aarch64_ldrstr_offset_compare): New.
4808 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4809 load/store orderings.
4810 (aarch64_gen_adjusted_ldpstp): Likewise.
4812 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4814 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4815 Check for subset of GENERAL_REGS and FP_REGS.
4816 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4819 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4821 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4822 and wi::to_poly_offset. Add the current offset and then check
4823 whether the sum fits, rather than using an unchecked addition of
4824 a checked term. Check for a shwi rather than a uhwi.
4825 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4826 (store_constructor): Use poly_int_tree_p.
4827 (expand_expr_real_1): Likewise.
4828 * function.c (assign_temp): Likewise.
4829 * fold-const.c (const_binop): Use poly_int_tree_p and
4831 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4833 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4834 to_poly_offset for MEM offsets.
4835 * ipa-icf.c (sem_variable::equals): Likewise.
4836 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4837 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4838 wi::to_poly_offset for BIT_FIELD_REF offsets.
4839 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4841 * var-tracking.c (emit_note_insn_var_location): Use
4842 tree_to_poly_uint64.
4844 2018-05-29 Jim Wilson <jimw@sifive.com>
4846 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4848 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4851 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4852 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4854 (sse4_1_round<mode>2): Use nonimmediate_operand
4855 for operand 1 predicate.
4857 2018-05-29 Martin Sebor <msebor@redhat.com>
4858 Richard Biener <rguenther@suse.de>
4861 * calls.c (get_size_range): Call determine_value_range instead
4862 of get_value_range..
4863 * tree-vrp.h (determine_value_range): Declared new function.
4864 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4866 2018-05-29 Richard Biener <rguenther@suse.de>
4868 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4869 sure to use non-pattern stmts for get_earlier_stmt arguments.
4870 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4871 called on pattern stmts.
4872 (get_later_stmt): Likewise.
4874 2018-05-29 Martin Liska <mliska@suse.cz>
4876 PR gcov-profile/85759
4877 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4880 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4882 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4884 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4888 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4889 VEC_PACK_FLOAT_EXPR): New tree codes.
4890 * tree-pretty-print.c (op_code_prio): Handle
4891 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4892 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4893 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4894 * tree-inline.c (estimate_operator_cost): Likewise.
4895 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4896 * fold-const.c (const_binop): Likewise.
4897 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4898 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4899 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4900 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4901 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4902 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4903 * expr.c (expand_expr_real_2): Likewise.
4904 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4905 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4906 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4908 * optabs.c (expand_widen_pattern_expr): For
4909 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4910 sign from result type rather than operand's type.
4911 (expand_binop_directly): For vec_packu_float_optab and
4912 vec_packs_float_optab allow result type to be different from operand's
4914 * optabs-tree.c (optab_for_tree_code): Handle
4915 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4916 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4917 * tree-vect-generic.c (expand_vector_operations_1): Handle
4918 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4919 VEC_PACK_FLOAT_EXPR.
4920 * tree-vect-stmts.c (supportable_widening_operation): Handle
4922 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4923 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4924 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4925 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4926 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4928 (vec_pack<floatprefix>_float_<mode>): New expander.
4929 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4931 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4932 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4933 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4934 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4935 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4937 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4938 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4939 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4940 VEC_PACK_FLOAT_EXPR): Document.
4942 2018-05-29 Richard Biener <rguenther@suse.de>
4944 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4946 (stmt_vec_info_vec): Make pointer.
4947 (init_stmt_vec_info_vec): Remove.
4948 (free_stmt_vec_info_vec): Likewise.
4949 (set_stmt_vec_info_vec): New function.
4950 (free_stmt_vec_infos): Likewise.
4951 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4952 (set_vinfo_for_stmt): Likewise.
4953 (get_earlier_stmt): Likewise.
4954 (get_later_stmt): Likewise.
4955 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4956 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4957 (vec_info::~vec_info): Free stmt_vec_infos.
4958 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4959 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4960 (pass_slp_vectorize::execute): Likewise.
4961 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4962 (free_stmt_vec_info_vec): Likewise.
4963 (set_stmt_vec_info_vec): New function.
4964 (free_stmt_vec_infos): Likewise.
4965 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4966 the global stmt_vec_info_vec.
4967 * tree-parloops.c (gather_scalar_reductions): Use
4968 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4971 2018-05-29 Richard Biener <rguenther@suse.de>
4973 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4975 2018-05-29 Martin Liska <mliska@suse.cz>
4976 David Malcolm <dmalcolm@redhat.com>
4978 * vec.c (test_reverse): New.
4979 (vec_c_tests): Add new test.
4980 * vec.h (vl_ptr>::reverse): New function.
4982 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4984 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4986 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4989 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4991 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4993 2018-05-28 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/85933
4996 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4997 look at stmts marked as vectorizable.
4999 2018-05-28 Richard Biener <rguenther@suse.de>
5001 PR tree-optimization/85934
5002 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5003 vector boolean check before scalar optimization.
5005 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5007 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5010 2018-05-28 Mark Wielaard <mark@klomp.org>
5012 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5013 if it is an expression containing a minus sign.
5015 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5017 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5019 2018-05-27 Paul Koning <ni1d@arrl.net>
5021 * config/pdp11/pdp11.md (truncsihi2): Remove.
5023 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5024 Chung-Ju Wu <jasonwucj@gmail.com>
5026 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5028 (unaligned_store_dw): Ditto.
5029 * config/nds32/nds32-memory-manipulation.c
5030 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5031 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5032 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5033 (emit_setmem_word_loop): Rename to ...
5034 (emit_setmem_doubleword_loop): ... this.
5035 (nds32_gen_dup_4_byte_to_word_value): New function.
5036 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5037 (nds32_expand_setmem_loop): Refine implementation.
5038 (nds32_expand_setmem_loop_v3m): Ditto.
5039 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5042 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5046 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5048 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5049 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5050 machine->attr_no_prologue_p.
5051 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5052 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5053 (nds32_expand_epilogue): Consider attr_naked_p.
5054 (nds32_expand_epilogue_v3pop): Likewise.
5055 (nds32_can_use_return_insn): Likewise.
5056 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5057 attr_no_prologue_p fields.
5058 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5060 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5063 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5065 * config/i386/sse.md
5066 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5068 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5070 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5072 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5074 (*<floatsuffix>floatv2div2sf2): Rename to ...
5075 (*float<floatunssuffix>v2div2sf2): ... this.
5076 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5077 (float<floatunssuffix>v2div2sf2_mask): ... this.
5078 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5079 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5080 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5082 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5084 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5086 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5088 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5090 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5092 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5093 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5094 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5095 gen_ufix_truncv8dfv8si2.
5096 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5097 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5098 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5099 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5100 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5101 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5102 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5103 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5105 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5109 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5111 2018-05-25 Jim Wilson <jimw@sifive.com>
5113 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5114 * config/riscv/riscv.c (struct machine_function): Add
5115 interrupt_handler_p and attribute_checked_p fields.
5116 (riscv_attribute_table): Add interrupt.
5117 (riscv_interrupt_type_p): New.
5118 (riscv_save_reg_p): Save extra regs for interrupt handler.
5119 (riscv_use_save_libcall): Return false for interrupt handler.
5120 (riscv_first_stack_step): Add forward declaration.
5121 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5122 for interrupt handler with large frame. Use it for saved reg list.
5123 (riscv_expand_prologue): Move flag_stack_usage_info support to
5124 eliminate duplication.
5125 (riscv_expand_epilogue): Generate mret for interrupt handler.
5126 (riscv_epilogue_uses): New.
5127 (riscv_can_use_return_insn): Return false for interrupt handler.
5128 (riscv_function_ok_for_sibcall): Likewise.
5129 (riscv_set_current_function): Add interrupt handler support.
5130 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5131 * config/riscv/riscv.md (UNSPECV_MRET): New.
5133 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5135 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5137 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5139 PR tree-optimization/85712
5140 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5141 this candidate has already been replaced in-situ by a copy.
5143 2018-05-25 Jason Merrill <jason@redhat.com>
5145 PR c++/80485 - inline function non-zero address.
5146 * symtab.c (nonzero_address): Check DECL_COMDAT.
5148 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5151 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5152 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5153 extension of SImode operation. Use const123_operand predicate.
5154 (*saddsi_1): Remove.
5155 (*saddl_se_1): Ditto.
5157 (*ssubl_se_1): Ditto.
5158 * config/alpha/predicates.md (const123_operand): New predicate.
5159 * config/alpha/constraints.md (P): Use IN_RANGE.
5161 2018-05-25 Richard Biener <rguenther@suse.de>
5163 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5165 (ref_maybe_used_by_stmt_p): Likewise.
5166 (stmt_may_clobber_ref_p): Likewise.
5167 (stmt_may_clobber_ref_p_1): Likewise.
5168 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5170 (ref_maybe_used_by_stmt_p): Likewise.
5171 (stmt_may_clobber_ref_p): Likewise.
5172 (stmt_may_clobber_ref_p_1): Likewise.
5173 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5174 the alias oracle to disambiguate DRs with stmts DR analysis
5176 (vect_analyze_data_refs): Do not give up on not analyzable
5177 DRs for BB vectorization. Remove code truncating the dataref
5180 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5183 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5184 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5185 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5187 2018-05-25 Richard Biener <rguenther@suse.de>
5189 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5190 function, combining stmt data ref gathering and fatal analysis
5192 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5193 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5194 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5195 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5196 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5198 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5200 PR tree-optimization/85720
5201 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5202 SCC if all partitions are builtins.
5203 (version_loop_by_alias_check): New parameter. Generate cancelable
5204 runtime alias check if all partitions are builtins.
5205 (distribute_loop): Update call to above function.
5207 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5209 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5210 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5211 (parm_default_def_partition_arg): Ditto.
5212 (set_parm_default_def_partition): Ditto.
5213 (get_parm_default_def_partitions): Ditto and make it static.
5214 (get_undefined_value_partitions): Ditto and make it static.
5215 (remove_ssa_form): Refactor call to init_var_map here.
5216 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5217 computation for loop region.
5218 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5219 (register_default_def): Delete.
5220 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5221 (parm_default_def_partition_arg): Ditto.
5222 (set_parm_default_def_partition): Ditto.
5223 (get_parm_default_def_partitions): Ditto and make it static.
5224 (get_undefined_value_partitions): Ditto and make it static.
5225 (coalesce_with_default, coalesce_with_default): Update comment.
5226 (create_coalesce_list_for_region): New func factored out from
5227 create_outofssa_var_map.
5228 (populate_coalesce_list_for_outofssa): New func factored out from
5229 create_outofssa_var_map and coalesce_ssa_name.
5230 (create_outofssa_var_map): Delete.
5231 (coalesce_ssa_name): Refactor to support live range computation.
5232 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5233 (get_parm_default_def_partitions): Delete.
5234 (get_undefined_value_partitions): Ditto.
5235 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5236 computation for loop region.
5237 (new_tree_live_info, loe_visit_block): Ditto.
5238 (live_worklist, set_var_live_on_entry): Ditto.
5239 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5240 * tree-ssa-live.h (struct _var_map): New fields.
5241 (init_var_map): Change decl.
5242 (region_contains_p): New.
5244 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5246 * tree-ssa-live.h (live_merge_and_clear): Delete.
5248 2018-05-25 Richard Biener <rguenther@suse.de>
5251 * tree-dump.c (dequeue_and_dump): Remove access to removed
5252 operand 2 of a SWITCH_EXPR.
5254 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5256 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5257 multiplication and division.
5258 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5259 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5260 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5261 (cond_udiv_optab, cond_umod_optab): New optabs.
5262 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5263 (IFN_COND_RDIV): New internal functions.
5264 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5265 TRUNC_MOD_EXPR and RDIV_EXPR.
5266 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5267 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5269 (SVE_INT_BINARY): Include mult.
5270 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5271 (optab, sve_int_op): Handle mult.
5272 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5274 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5275 for SVE_INT_BINARY_SD.
5277 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5279 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5280 (optab, sve_int_op): Handle div and udiv.
5281 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5282 for SVE_INT_BINARY_SD.
5283 (*<optab><mode>3): New insn for the same.
5285 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5287 * tree-vect-patterns.c: Include predict.h.
5288 (vect_recog_divmod_pattern): Restrict check for division support
5289 to when optimizing for size.
5291 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5293 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5294 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5295 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5296 (gimple_match_op::set_op): Likewise.
5297 (gimple_resimplify4): Declare.
5298 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5299 (expr::gen_transform): Likewise.
5300 (decision_tree::gen): Generate a simplification routine for 4 operands.
5301 * gimple-match-head.c (gimple_simplify): Add an overload for
5302 4 operands. In the top-level function, handle up to 4 call
5303 arguments and call gimple_resimplify4.
5304 (gimple_resimplify4): New function.
5305 (build_call_internal): Pass a fourth operand.
5306 (maybe_push_to_seq): Likewise.
5307 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5308 Fold VEC_COND_EXPRs of an operation and a default value into
5309 an IFN_COND_* function if possible.
5310 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5312 (SVE_COND_FP_BINARY): Include them.
5313 (optab, sve_fp_op): Handle them.
5314 (SVE_INT_BINARY_REV): New code iterator.
5315 (SVE_COND_FP_BINARY_REV): New int iterator.
5316 (commutative): New int attribute.
5317 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5319 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5321 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5322 (*cond_<optab><mode>): New patterns for reversed operands.
5324 2018-05-25 Richard Biener <rguenther@suse.de>
5326 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5327 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5328 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5329 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5330 * tree-vect-data-refs.c (everywhere): Adjust users.
5331 * tree-vect-loop.c (everywhere): Likewise.
5332 * tree-vect-slp.c (everywhere): Likewise.
5333 * tree-vect-stmts.c (everywhere): Likewise.
5334 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5336 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5338 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5340 (gcc_cv_as_section_exclude): ... this.
5341 Try Solaris as #exclude syntax.
5342 * configure: Regenerate.
5343 * config.in: Regenerate.
5344 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5346 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5347 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5349 * varasm.c (default_elf_asm_named_section): Don't check if
5350 HAVE_GAS_SECTION_EXCLUDE is defined.
5352 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5354 * doc/md.texi: Update the documentation of the cond_* optabs
5355 to mention the new final operand. Fix GET_MODE_NUNITS call.
5356 Describe the scalar case too.
5357 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5358 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5360 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5361 (get_conditional_internal_fn): Update comment.
5362 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5363 accumulator value as a final argument to conditional functions.
5364 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5365 a define_expand and add an "else" operand. Assert for now that
5366 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5367 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5368 (*cond_<optab><mode>): New patterns.
5369 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5370 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5371 (UNSPEC_COND_EOR): Delete.
5372 (optab): Remove associated mappings.
5373 (SVE_INT_BINARY): New code iterator.
5374 (sve_int_op): Remove int attribute and add "minus" to the code
5376 (SVE_COND_INT_OP): Delete.
5377 (SVE_COND_FP_OP): Rename to...
5378 (SVE_COND_FP_BINARY): ...this.
5380 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5382 * optabs.c (can_reuse_operands_p): New function.
5383 (maybe_legitimize_operands): Try to reuse the results for
5386 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5388 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5389 Add {q} suffix to insn mnemonic.
5391 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5393 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5394 (msp430_warn_func_return): New.
5396 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5398 * fold-const.c (tree_nonzero_bits): New function.
5399 * fold-const.h (tree_nonzero_bits): Likewise.
5400 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5401 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5403 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5407 * varasm.c (assemble_alias): Check ifunc_resolver only on
5410 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5413 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5414 when memory input operand is handled.
5416 2018-05-24 Luis Machado <luis.machado@linaro.org>
5418 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5420 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5422 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5424 * match.pd: Delay FMA folds until after vectorization.
5426 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5429 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5430 address check not strict.
5432 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5434 * gimple-match.h (gimple_match_op): New class.
5435 (mprts_hook): Replace parameters with a gimple_match_op *.
5436 (maybe_build_generic_op): Likewise.
5437 (gimple_simplified_result_is_gimple_val): Replace parameters with
5438 a const gimple_match_op *.
5439 (gimple_simplify): Replace code_helper * and tree * parameters with
5440 a gimple_match_op * parameter.
5441 (gimple_resimplify1): Replace code_helper *, tree and tree *
5442 parameters with a gimple_match_op * parameter.
5443 (gimple_resimplify2): Likewise.
5444 (gimple_resimplify3): Likewise.
5445 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5446 parameters with a gimple_match_op * parameter.
5447 * gimple-match-head.c (gimple_simplify): Change prototypes of
5448 auto-generated functions to take a gimple_match_op * instead of
5449 separate code_helper * and tree * parameters. Make the same
5450 change in the top-level overload and update calls to the
5451 gimple_resimplify routines. Update calls to the auto-generated
5452 functions and to maybe_push_res_to_seq in the publicly-facing
5453 operation-specific gimple_simplify overloads.
5454 (gimple_match_op::MAX_NUM_OPS): Define.
5455 (gimple_resimplify1): Replace rcode and ops with a single res_op
5456 parameter. Update call to gimple_simplify.
5457 (gimple_resimplify2): Likewise.
5458 (gimple_resimplify3): Likewise.
5459 (mprts_hook): Replace parameters with a gimple_match_op *.
5460 (maybe_build_generic_op): Likewise.
5461 (build_call_internal): Replace type, nargs and ops with
5462 a gimple_match_op *.
5463 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5464 with a single gimple_match_op *. Update calls to mprts_hook,
5465 build_call_internal and gimple_simplified_result_is_gimple_val.
5466 Factor out code that is common to the tree_code and combined_fn cases.
5467 * genmatch.c (expr::gen_transform): Replace tem_code and
5468 tem_ops with a gimple_match_op called tem_op. Update calls
5469 to the gimple_resimplify functions and maybe_push_res_to_seq.
5470 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5471 res_ops. Update call to the gimple_resimplify functions.
5472 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5473 (decision_tree::gen): Make the functions take a gimple_match_op *
5474 called res_op instead of separate res_code and res_ops parameters.
5475 Update call accordingly.
5476 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5477 and ops with a single res_op parameter. Update calls to
5478 maybe_build_generic_op and maybe_push_res_to_seq.
5479 (fold_stmt_1): Update calls to gimple_simplify and
5480 replace_stmt_with_simplification.
5481 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5482 and gimple_simplified_result_is_gimple_val.
5483 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5485 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5486 with a gimple_match_op *.
5487 (vn_nary_build_or_lookup): Likewise. Update call to
5488 vn_nary_build_or_lookup_1.
5489 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5490 gimple_match_op *. Update calls to the gimple_resimplify routines
5491 and to gimple_simplified_result_is_gimple_val.
5492 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5493 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5494 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5495 (visit_nary_op): Likewise.
5496 (visit_reference_op_load): Likewise.
5498 2018-05-23 Luis Machado <luis.machado@linaro.org>
5500 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5501 modifier for printing the step amount.
5503 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5506 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5509 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5511 * doc/sourcebuild.texi (Endianness): New subsubsection.
5513 2018-05-23 Luis Machado <luis.machado@linaro.org>
5515 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5516 <prefetch_dynamic_strides>: New const bool field.
5517 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5518 prefetch_dynamic_strides.
5519 (exynosm1_prefetch_tune): Likewise.
5520 (thunderxt88_prefetch_tune): Likewise.
5521 (thunderx_prefetch_tune): Likewise.
5522 (thunderx2t99_prefetch_tune): Likewise.
5523 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5525 (aarch64_override_options_internal): Update to set
5526 PARAM_PREFETCH_DYNAMIC_STRIDES.
5527 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5528 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5529 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5530 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5531 prefetch-dynamic-strides setting.
5533 2018-05-23 Luis Machado <luis.machado@linaro.org>
5535 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5536 <minimum_stride>: New const int field.
5537 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5538 minimum_stride field defaulting to -1.
5539 (exynosm1_prefetch_tune): Likewise.
5540 (thunderxt88_prefetch_tune): Likewise.
5541 (thunderx_prefetch_tune): Likewise.
5542 (thunderx2t99_prefetch_tune): Likewise.
5543 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5544 <default_opt_level>: Set to 3.
5545 (aarch64_override_options_internal): Update to set
5546 PARAM_PREFETCH_MINIMUM_STRIDE.
5547 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5548 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5549 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5550 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5551 stride is constant and is below the minimum stride threshold.
5553 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5555 * config/arm/arm-cpus.in (mode26): Delete.
5556 (armv4): Delete mode26 reference.
5557 * config/arm/arm.c (arm_configure_build_target): Delete use of
5560 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5562 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5564 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5565 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5567 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5568 (floatunsdidf2): Ditto.
5570 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5572 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5573 (fixuns_trunc<mode>si2_avx512f): Ditto.
5574 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5575 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5576 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5578 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5580 PR rtl-optimization/79985
5581 * df-scan.c (df_insn_refs_collect): Remove special case for
5582 global registers and asm statements.
5584 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5586 * extend.texi (Global Register Variables): Rewrite the bullet list.
5587 Note that the register is available for allocation. Note that access
5588 via inline asm must use constraints. Add note about async-signal
5589 handlers. Remove paragraph about automagic register selection.
5591 2018-05-23 Richard Biener <rguenther@suse.de>
5593 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5594 of fixed offset from memset VN.
5596 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5598 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5600 (alloc_cand_and_find_basis): Initialize first_interp field.
5601 (slsr_process_mul): Modify first_interp field.
5602 (slsr_process_add): Likewise.
5603 (slsr_process_cast): Modify first_interp field for each new
5605 (slsr_process_copy): Likewise.
5606 (dump_candidate): Dump first_interp field.
5607 (replace_mult_candidate): Process all interpretations, not just
5609 (replace_rhs_if_not_dup): Likewise.
5610 (replace_one_candidate): Likewise.
5612 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5614 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5616 (aarch64_needs_frame_chain): New function.
5617 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5619 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5622 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5623 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5624 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5625 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5626 as true for strict-align.
5627 (aarch64_can_inline_p): Perform checks even when callee has no
5628 attributes to check for strict alignment.
5629 * doc/extend.texi (AArch64 Function Attributes): Document
5631 * doc/invoke.texi: (AArch64 Options): Likewise.
5633 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5635 PR tree-optimization/85853
5636 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5637 the handling of the root of the node to...
5638 (vect_slp_analyze_node_operations_1): ...this new function,
5639 and run the whole thing with the child nodes' def types
5640 set according to their SLP node's def type.
5642 2018-05-23 Richard Biener <rguenther@suse.de>
5645 * tree-data-ref.c (create_runtime_alias_checks): Defer
5646 and ignore overflow warnings.
5648 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5650 PR tree-optimization/85822
5651 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5654 2018-05-23 Richard Biener <rguenther@suse.de>
5656 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5657 memset constants via native_interpret_expr.
5659 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5662 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5664 (cgraph_node::create_alias): Likewise.
5665 (cgraph_node::get_availability): Check ifunc_resolver instead
5666 of looking up ifunc attribute.
5667 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5668 * varasm.c (do_assemble_alias): Likewise.
5669 (assemble_alias): Likewise.
5670 (default_binds_local_p_3): Likewise.
5671 * cgraph.h (cgraph_node): Add ifunc_resolver.
5672 (cgraph_node::only_called_directly_or_aliased_p): Return false
5674 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5676 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5677 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5678 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5679 instead of looking up ifunc attribute.
5681 2018-05-22 Luis Machado <luis.machado@linaro.org>
5683 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5685 2018-05-22 Martin Sebor <msebor@redhat.com>
5688 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5689 only when expasion succeeds.
5690 (expand_builtin_strcmp): Same.
5691 (expand_builtin_strncmp): Same.
5693 2018-05-22 Martin Sebor <msebor@redhat.com>
5695 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5697 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5698 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5700 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5701 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5703 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5704 Add check for legitimate_address.
5705 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5706 (aarch64_swap_ldrstr_operands): New.
5707 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5710 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5711 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5713 * config/aarch64/aarch64.md: New patterns to generate stp
5715 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5716 single words and double words.
5717 (load_pair_sw, load_pair_dw): Likewise.
5718 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5720 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5722 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5723 for different mode ldpstp and add peephole for merged zero stores.
5725 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5727 (aarch64_gen_store_pair): Rename calls to match new patterns.
5728 (aarch64_gen_load_pair): Rename calls to match new patterns.
5729 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5730 (load_pair<DREG:mode><DREG2:mode>): ... This.
5731 (store_pair<mode>): Rename to...
5732 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5733 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5735 (V_INT_EQUIV): Handle SImode.
5736 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5739 2018-05-22 Martin Sebor <msebor@redhat.com>
5742 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5743 or ajust the presumed bound on an operation to avoid unnecessary
5746 2018-05-22 Martin Sebor <msebor@redhat.com>
5748 PR tree-optimization/85826
5749 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5750 assuming that a DECL necesarily has a constant size.
5752 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5755 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5757 2018-05-22 Richard Biener <rguenther@suse.de>
5759 PR tree-optimization/85834
5760 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5761 non-constant and non-zero memset arguments.
5763 2018-05-22 Martin Liska <mliska@suse.cz>
5766 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5768 2018-05-22 Richard Biener <rguenther@suse.de>
5770 PR tree-optimization/85863
5771 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5772 comparisons when vectype is specified.
5773 (vectorizable_condition): Do not specify vectype for
5774 vect_is_simple_cond when SLP vectorizing.
5776 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5779 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5780 define __ibm128 as long double.
5781 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5782 as a distinct type when IEEE 128-bit support is enabled.
5783 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5784 128-bit types to use the correct functions.
5785 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5786 convert between 128-bit floating point types that have different
5787 modes but the same representation, instead of using gen_lowpart to
5789 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5791 (IFKF_reg): New attributes to give the register constraints for
5793 (extend<mode>tf2_internal): New insns to mark an explicit
5794 conversion between 128-bit floating point types that have a
5795 different mode but share the same representation.
5797 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5799 PR tree-optimization/85814
5800 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5801 a null return from get_strinfo when unsharing the next
5802 strinfo in the chain.
5804 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5807 * varasm.c (weak_finish): Clean up weak_decls.
5809 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5811 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5812 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5813 UNSPEC_UADALP values.
5814 * config/aarch64/iterators.md (ABAL): New int iterator.
5817 (sur): Add mappings for the above.
5818 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5820 (aarch64_<sur>abal<mode>_4): Likewise.
5821 (aarch64_<sur>adalp<mode>_3): Likewise.
5822 (<sur>sadv16qi): New define_expand.
5824 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5826 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5827 (*movdf_internal): Ditto.
5828 (*rcpsf2_sse): Ditto.
5829 (*rsqrtsf2_sse): Ditto.
5830 (*sqrt<mode>2_sse): Ditto.
5832 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5834 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5836 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5837 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5838 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5839 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5841 * config/aarch64/arm_neon.h: Likewise.
5842 * config/aarch64/iterators.md (VQ_I): New.
5844 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5846 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5848 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5851 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5853 * config/nds32/constraints.md (S): New constraint.
5854 * config/nds32/nds32.md (call_internal): Use constraint S.
5855 (call_value_internal): Likewise.
5856 (sibcall_internal): Likewise.
5857 (sibcall_value_internal): Likewise.
5859 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5860 Chung-Ju Wu <jasonwucj@gmail.com>
5862 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5865 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5866 Chung-Ju Wu <jasonwucj@gmail.com>
5868 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5869 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5870 (nds32_rtx_costs_impl): Simplify.
5871 (nds32_address_cost_impl): Simplify.
5872 (nds32_init_rtx_costs): New function.
5873 (nds32_rtx_costs_speed_prefer): Likewise.
5874 (nds32_rtx_costs_size_prefer): Likewise.
5875 (nds32_address_cost_speed_prefer): Likewise.
5876 (nds32_address_cost_speed_fwprop): Likewise.
5877 (nds32_address_cost_size_prefer): Likewise.
5878 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5879 * config/nds32/nds32.c (nds32_option_override): Use
5880 nds32_init_rtx_costs function.
5882 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5884 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5885 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5886 (TARGET_PIPELINE_N8): Likewise.
5887 (TARGET_PIPELINE_N10): Likewise.
5888 (TARGET_PIPELINE_N13): Likewise.
5889 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5891 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5893 * config/nds32/nds32-fpu.md: Update copyright year.
5895 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5897 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5899 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5901 * config/nds32/nds32.c
5902 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5903 * config/nds32/nds32.opt (minline-asm-r15): New option.
5905 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5907 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5909 * config/nds32/nds32.md (abssi2): New pattern.
5911 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5913 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5914 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5915 Merge insn pattern from sse_cvtsi2ss<round_name> and
5916 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5917 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5918 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5919 using SWI48 mode iterator.
5920 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5921 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5922 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5923 pattern from sse_cvttss2si<round_saeonly_name>
5924 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5925 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5926 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5927 using SWI48 mode iterator.
5928 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5929 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5930 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5931 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5932 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5933 using SWI48 mode iterator.
5934 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5935 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5936 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5937 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5938 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5939 SWI48 mode iterator.
5940 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5941 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5942 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5943 pattern from sse_cvttsd2si<round_saeonly_name>
5944 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5946 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5948 * config/nds32/nds32-md-auxiliary.c
5949 (nds32_valid_smw_lwm_base_p): Refine.
5950 (nds32_output_smw_single_word): Refine.
5951 (nds32_output_smw_double_word): New.
5952 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5954 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5956 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5957 (nds32_output_stack_pop): Refine.
5958 (nds32_expand_unaligned_load): Refine.
5959 (nds32_expand_unaligned_store): Refine.
5961 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5962 Chung-Ju Wu <jasonwucj@gmail.com>
5964 * config/nds32/constants.md: Add TP_REGNUM constant.
5965 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5966 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5968 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5969 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5970 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5971 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5972 and PIC code generation.
5973 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5975 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5977 * config/nds32/nds32.md: Support TLS and PIC.
5978 * config/nds32/nds32.c: Support TLS and PIC.
5979 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5980 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5983 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5985 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5986 mode with E_ prefix.
5988 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5989 Chung-Ju Wu <jasonwucj@gmail.com>
5991 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5992 * config/nds32/nds32-md-auxiliary.c
5993 (symbolic_reference_mentioned_p): New.
5994 (nds32_legitimize_ict_address): New.
5995 (nds32_expand_ict_move): New.
5996 (nds32_indirect_call_referenced_p): New.
5997 (nds32_symbol_binds_local_p): Delete.
5998 (nds32_long_call_p): Modify.
5999 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6000 * config/nds32/nds32-protos.h
6001 (symbolic_reference_mentioned_p): Declare.
6002 (nds32_legitimize_ict_address): Declare.
6003 (nds32_expand_ict_move): Declare.
6004 (nds32_indirect_call_referenced_p): Declare.
6005 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6006 (nds32_relax_group): Use nds32_ict_const_p as condition.
6007 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6008 (nds32_asm_file_start): Output ict_model directive in asm code.
6009 (nds32_legitimate_address_p): Consider indirect call.
6010 (nds32_print_operand): Consider indirect call.
6011 (nds32_print_operand_address): Consider indirect call.
6012 (nds32_insert_attributes): Handle "indirect_call" attribute.
6013 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6014 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6015 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6016 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6017 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6018 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6019 (TARGET_ICT_MODEL_SMALL): Define.
6020 (TARGET_ICT_MODEL_LARGE): Define.
6021 * config/nds32/nds32.md (movsi): Consider ict model.
6022 (call, call_value): Consider ict model.
6023 (sibcall, sibcall_value): Consider ict model.
6024 * config/nds32/nds32.opt (mict-model): New option.
6025 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6028 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6029 Monk Chiang <sh.chiang04@gmail.com>
6030 Jim Wilson <jimw@sifive.com>
6032 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6033 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6034 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6035 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6036 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6037 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6038 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6039 compute save_libcall_adjustment properly.
6040 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6041 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6042 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6043 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6044 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6045 (ABI_SPEC): Handle mabi=ilp32e.
6046 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6047 (RVE): Add RVE mask.
6048 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6049 <-march>: Add rv32e as an example.
6051 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6054 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6055 (intra_create_variable_infos): Handle C++ constructors.
6057 2018-05-18 Martin Liska <mliska@suse.cz>
6059 * passes.def: Remove a redundant pass.
6061 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6064 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6066 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6068 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6070 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6072 (armv2, armv2a, armv3, armv3m): Delete architectures.
6073 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6074 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6075 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6077 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6078 (*mulsidi3adddi): Likewise.
6079 (mulsidi3): Likewise.
6080 (*mulsidi3_nov6): Likewise.
6081 (umulsidi3): Likewise.
6082 (umulsidi3_nov6): Likewise.
6083 (umaddsidi4): Likewise.
6084 (*umulsidi3adddi): Likewise.
6085 (smulsi3_highpart): Likewise.
6086 (*smulsi3_highpart_nov6): Likewise.
6087 (umulsi3_highpart): Likewise.
6088 (*umulsi3_highpart_nov6): Likewise.
6089 * config/arm/arm.h (arm_arch3m): Delete.
6090 * config/arm/arm.c (arm_arch3m): Delete.
6091 (arm_option_override_internal): Update armv3-related comment.
6092 (arm_configure_build_target): Delete use of isa_bit_mode32.
6093 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6094 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6095 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6097 (mulusa3): Likewise.
6098 * config/arm/arm-protos.h (arm_arch3m): Delete.
6099 * config/arm/arm-tables.opt: Regenerate.
6100 * config/arm/arm-tune.md: Likewise.
6101 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6102 deleted architectures.
6104 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6106 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6107 (armv5t, armv5te): New features.
6108 (ARMv5, ARMv5e): Delete fgroups.
6109 (ARMv5t, ARMv5te): Adjust for above changes.
6111 (armv5, armv5e): Delete arches.
6112 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6114 (*call_reg_arm): Likewise.
6115 (*call_value_reg_armv5): Likewise.
6116 (*call_value_reg_arm): Likewise.
6117 (*call_symbol): Likewise.
6118 (*call_value_symbol): Likewise.
6119 (*sibcall_insn): Likewise.
6120 (*sibcall_value_insn): Likewise.
6122 (prefetch): Likewise.
6123 (define_split and define_peephole2 dependent on arm_arch5):
6125 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6127 (TARGET_ARM_QBIT): Likewise.
6128 (TARGET_DSP_MULTIPLY): Likewise.
6129 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6130 (arm_arch5, arm_arch5e): Delete.
6131 (arm_arch5t, arm_arch5te): Declare.
6132 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6133 (arm_arch5t): Declare.
6134 (arm_option_reconfigure_globals): Update for the above.
6135 (arm_options_perform_arch_sanity_checks): Update comment, replace
6136 use of arm_arch5 with arm_arch5t.
6137 (use_return_insn): Likewise.
6138 (arm_emit_call_insn): Likewise.
6139 (output_return_instruction): Likewise.
6140 (arm_final_prescan_insn): Likewise.
6141 (arm_coproc_builtin_available): Likewise.
6142 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6143 arm_arch5e with arm_arch5t and arm_arch5te.
6144 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6145 (arm_arch5t, arm_arch5te): Declare.
6146 * config/arm/arm-tables.opt: Regenerate.
6147 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6148 * config/arm/t-multilib: Likewise.
6149 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6150 instead of arm_arch5.
6151 (*call_reg_thumb1): Likewise.
6152 (*call_value_reg_thumb1_v5): Likewise.
6153 (*call_value_reg_thumb1): Likewise.
6154 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6156 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6158 2018-05-18 Martin Liska <mliska@suse.cz>
6160 PR gcov-profile/84846
6161 * doc/gcov.texi: Document -t option of gcov tool.
6163 2018-05-18 Martin Liska <mliska@suse.cz>
6165 PR gcov-profile/84846
6166 * gcov.c (print_usage): Add new -t option.
6167 (process_args): Handle the option.
6168 (generate_results): Use stdout as output when requested by
6171 2018-05-18 Martin Liska <mliska@suse.cz>
6173 PR gcov-profile/84846
6174 * coverage.c (coverage_init): Write PWD to .gcno file.
6175 * doc/gcov.texi: Document how working directory is printed.
6176 * gcov-dump.c (dump_gcov_file): Print PWD.
6177 * gcov.c (output_intermediate_file): Likewise.
6178 (read_graph_file): Read PWD string.
6179 (output_lines): Print PWD.
6181 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6184 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6185 for retval and return false if all args to phi are zero.
6187 2018-05-18 Richard Biener <rguenther@suse.de>
6189 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6191 (evrp_dom_walker::before_dom_children): Call it.
6193 2018-05-18 Richard Biener <rguenther@suse.de>
6195 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6196 results when processing array refs with variable index.
6198 2018-05-18 Toon Moene <toon@moene.org>
6200 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6201 directly after that of -floop-interchange. Indicate that both
6202 options are enabled by default when specifying -O3.
6204 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6206 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6207 iterator. Delete separate integer-mode vec_set<mode> expander.
6208 (aarch64_simd_vec_setv2di): Delete.
6209 (vec_setv2di): Delete.
6210 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6211 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6212 the "w, r" alternative.
6214 2018-05-18 Martin Liska <mliska@suse.cz>
6216 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6217 * tree-pass.h (make_pass_lower_switch_O0): New function.
6218 * tree-switch-conversion.c (node_has_low_bound): Remove.
6219 (node_has_high_bound): Likewise.
6220 (node_is_bounded): Likewise.
6221 (class pass_lower_switch): Make it a template type and create
6223 (pass_lower_switch::execute): Add template argument.
6224 (make_pass_lower_switch): New function.
6225 (make_pass_lower_switch_O0): New function.
6226 (do_jump_if_equal): Remove.
6227 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6228 all the hard work to tree optimization passes.
6230 2018-05-18 Martin Liska <mliska@suse.cz>
6232 * dbgcnt.c (limit_low): Renamed from limit.
6233 (limit_high): New variable.
6234 (dbg_cnt_is_enabled): Check for upper limit.
6235 (dbg_cnt): Adjust dumping.
6236 (dbg_cnt_set_limit_by_index): Add new argument for high
6238 (dbg_cnt_set_limit_by_name): Likewise.
6239 (dbg_cnt_process_single_pair): Parse new format.
6240 (dbg_cnt_process_opt): Use strtok.
6241 (dbg_cnt_list_all_counters): Remove 'value' and add
6243 * doc/invoke.texi: Document changes.
6245 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6247 * doc/sourcebuild.texi (scalar_all_fma): Document.
6248 * tree.def (FMA_EXPR): Delete.
6249 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6250 * internal-fn.c (ternary_direct): New macro.
6251 (expand_ternary_optab_fn): Likewise.
6252 (direct_ternary_optab_supported_p): Likewise.
6253 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6254 * builtins.c (fold_builtin_fma): Delete.
6255 (fold_builtin_3): Don't call it.
6256 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6257 * expr.c (expand_expr_real_2): Likewise.
6258 * fold-const.c (operand_equal_p): Likewise.
6259 (fold_ternary_loc): Likewise.
6260 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6261 * gimple.c (DEFTREECODE): Likewise.
6262 * gimplify.c (gimplify_expr): Likewise.
6263 * optabs-tree.c (optab_for_tree_code): Likewise.
6264 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6265 * tree-eh.c (operation_could_trap_p): Likewise.
6266 (stmt_could_throw_1_p): Likewise.
6267 * tree-inline.c (estimate_operator_cost): Likewise.
6268 * tree-pretty-print.c (dump_generic_node): Likewise.
6269 (op_code_prio): Likewise.
6270 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6271 * tree-ssa-operands.c (get_expr_operands): Likewise.
6272 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6273 * fold-const-call.h (fold_fma): Delete.
6274 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6275 CFN_FNMA and CFN_FNMS.
6277 * genmatch.c (combined_fn): New enum.
6278 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6279 (commutative_op): New function.
6280 (commutate): Use it. Handle more than 2 operands.
6281 (dt_operand::gen_gimple_expr): Use commutative_op.
6282 (parser::parse_expr): Allow :c to be used with non-binary
6283 operators if the commutative operand is known.
6284 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6285 CFN_FMS, CFN_FNMA and CFN_FNMS.
6286 (backprop::process_assign_use): Remove FMA_EXPR handling.
6287 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6288 (gen_hsa_fma): New function.
6289 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6290 IFN_FNMA and IFN_FNMS.
6291 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6292 * gimple-fold.h (follow_all_ssa_edges): Declare.
6293 * gimple-fold.c (follow_all_ssa_edges): New function.
6294 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6295 gimple_build interface and use follow_all_ssa_edges to fold the result.
6296 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6297 instead of checking for optabs directly.
6298 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6299 rather than FMA_EXPRs.
6300 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6301 call to IFN_FMA instead of an FMA_EXPR.
6303 2018-05-17 Jim Wilson <jimw@sifive.com>
6305 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6306 sign extended promoted subreg, and the range does not have the sign bit
6307 set, then do a sign extend.
6309 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6310 test, check for sign extended subreg and/or constant operands, and
6311 do a sign extend in that case.
6313 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6315 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6316 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6318 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6319 Change logics_shift_reg to logics_shift_imm.
6320 (thunderx2t99_fp_loadpair_basic): Delete.
6321 (thunderx2t99_fp_storepair_basic): Delete.
6322 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6323 (thunderx2t99_asimd_polynomial): Delete.
6324 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6325 and neon_fp_mul_d_scalar_q.
6326 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6327 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6328 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6329 (thunderx2t99_asimd_lut): Add missing tbl types.
6330 (thunderx2t99_asimd_ext): Delete.
6331 (thunderx2t99_asimd_load1_1_mult): Delete.
6332 (thunderx2t99_asimd_load1_2_mult): Delete.
6333 (thunderx2t99_asimd_load1_ldp): New.
6334 (thunderx2t99_asimd_load1): New.
6335 (thunderx2t99_asimd_load2): Add missing *load2* types.
6336 (thunderx2t99_asimd_load3): New.
6337 (thunderx2t99_asimd_load4): New.
6338 (thunderx2t99_asimd_store1_1_mult): Delete.
6339 (thunderx2t99_asimd_store1_2_mult): Delete.
6340 (thunderx2t99_asimd_store2_mult): Delete.
6341 (thunderx2t99_asimd_store2_onelane): Delete.
6342 (thunderx2t99_asimd_store_stp): New.
6343 (thunderx2t99_asimd_store1): New.
6344 (thunderx2t99_asimd_store2): New.
6345 (thunderx2t99_asimd_store3): New.
6346 (thunderx2t99_asimd_store4): New.
6348 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6350 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6351 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6353 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6354 Segher Boessenkool <segher@kernel.crashing.org>
6357 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6360 2018-05-17 Richard Biener <rguenther@suse.de>
6362 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6363 for pruning loop and prune defs feeding only already visited PHIs.
6365 2018-05-17 Richard Biener <rguenther@suse.de>
6367 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6369 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6370 Richard Biener <rguenther@suse.de>
6372 PR tree-optimization/85793
6373 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6374 for VMAT_ELEMENTWISE.
6376 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6378 * internal-fn.h (lookup_internal_fn): Declare
6379 * internal-fn.c (lookup_internal_fn): New function.
6380 * gimple.c (gimple_build_call_from_tree): Handle calls to
6382 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6383 internal function names.
6384 * tree-pretty-print.c (dump_generic_node): Likewise.
6385 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6387 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6389 * gimple-fold.h (gimple_build): Make the function forms take
6390 combined_fn rather than built_in_function.
6391 (gimple_simplify): Likewise.
6392 * gimple-match-head.c (gimple_simplify): Likewise.
6393 * gimple-fold.c (gimple_build): Likewise.
6394 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6395 rather than gimple_build_call_internal.
6396 (get_initial_defs_for_reduction): Likewise.
6397 (vect_create_epilog_for_reduction): Likewise.
6398 (vectorizable_live_operation): Likewise.
6400 2018-05-17 Martin Liska <mliska@suse.cz>
6402 * gimple-ssa-sprintf.c (format_directive): Do not use
6403 space in between 'G_' and '('.
6405 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6408 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6409 even if the mask is not all ones.
6412 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6414 (ix86_gimple_fold_builtin): Likewise.
6417 * config/i386/i386.c: Include tree-vector-builder.h.
6418 (ix86_vector_shift_count): New function.
6419 (ix86_fold_builtin): Fold shift builtins by scalar count.
6420 (ix86_gimple_fold_builtin): Likewise.
6422 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6423 _mm512_setzero): New intrinsics.
6425 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6426 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6428 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6429 code generation for cases where splatting a value is not useful.
6430 * simplify-rtx.c (simplify_ternary_operation): Simplify
6431 vec_merge across a vec_duplicate and a paradoxical subreg forming
6432 a vector mode to a vec_concat.
6434 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6436 * config.gcc: Support "goldmont-plus".
6437 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6439 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6440 PROCESSOR_GOLDMONT_PLUS.
6441 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6442 (processor_target_table): Add "goldmont-plus".
6443 (PTA_GOLDMONT_PLUS): Define.
6444 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6445 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6446 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6447 (fold_builtin_cpu): Add "goldmont-plus".
6448 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6449 (ix86_option_override_internal): Add "goldmont-plus".
6450 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6451 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6452 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6453 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6455 2018-05-17 Richard Biener <rguenther@suse.de>
6457 PR tree-optimization/85757
6458 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6459 remove defs that only feed that PHI from further processing.
6461 2018-05-16 Jim Wilson <jimw@sifive.com>
6463 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6465 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6467 2018-05-16 Mark Wielaard <mark@klomp.org>
6469 * dwarf2out.c (count_index_strings): New function.
6470 (output_indirect_strings): Call count_index_strings and generate
6471 header for dwarf_version >= 5.
6473 2018-05-16 Mark Wielaard <mark@klomp.org>
6475 * dwarf2out.c (dwarf_FORM): New function.
6476 (set_indirect_string): Use dwarf_FORM.
6477 (reset_indirect_string): Likewise.
6478 (size_of_die): Likewise.
6479 (value_format): Likewise.
6480 (output_die): Likewise.
6481 (add_skeleton_AT_string): Likewise.
6482 (output_macinfo_op): Likewise.
6483 (index_string): Likewise.
6484 (output_index_string_offset): Likewise.
6485 (output_index_string): Likewise.
6486 (count_index_strings): Likewise.
6488 2018-05-16 Carl Love <cel@us.ibm.com>
6490 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6491 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6493 2018-05-16 Martin Jambor <mjambor@suse.cz>
6495 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6496 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6498 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6500 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6501 (fnma<mode>4): Likewise.
6502 (fms<mode>4): Likewise.
6503 (fnms<mode>4): Likewise.
6504 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6505 (aarch64_fnma<mode>4): Likewise.
6506 (aarch64_fms<mode>4): Likewise.
6507 (aarch64_fnms<mode>4): Likewise.
6508 (aarch64_fnmadd<mode>4): Likewise.
6510 2018-05-16 Jason Merrill <jason@redhat.com>
6512 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6514 2018-05-16 Richard Biener <rguenther@suse.de>
6516 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6517 (dump_stmt_cost): Declare.
6518 (add_stmt_cost): Dump cost we add.
6519 (add_stmt_costs): New function.
6520 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6522 (vect_analyze_stmt): Adjust prototype.
6523 (vectorizable_condition): Likewise.
6524 (vectorizable_live_operation): Likewise.
6525 (vectorizable_reduction): Likewise.
6526 (vectorizable_induction): Likewise.
6527 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6528 cost vector to pass to vectorizable_ and record afterwards.
6529 (vect_model_reduction_cost): Take cost vector argument and adjust.
6530 (vect_model_induction_cost): Likewise.
6531 (vectorizable_reduction): Likewise.
6532 (vectorizable_induction): Likewise.
6533 (vectorizable_live_operation): Likewise.
6534 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6535 SLP_TREE_NUMBER_OF_VEC_STMTS.
6536 (vect_analyze_slp_cost_1): Remove.
6537 (vect_analyze_slp_cost): Likewise.
6538 (vect_slp_analyze_node_operations): Take visited args and
6539 a target cost vector. Avoid processing already visited stmt sets.
6540 (vect_slp_analyze_operations): Use a local cost vector to gather
6541 costs and register those of non-discarded instances.
6542 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6543 (vect_schedule_slp_instance): Remove copying of
6544 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6546 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6547 adding cost. Record cost entry location.
6548 (vect_prologue_cost_for_slp_op): Function to compute cost of
6549 a constant or invariant generated for SLP vect in the prologue,
6550 split out from vect_analyze_slp_cost_1.
6551 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6552 (vect_model_promotion_demotion_cost): Likewise.
6553 (vect_model_store_cost): Likewise, make static.
6554 (vect_model_load_cost): Likewise.
6555 (vectorizable_bswap): Add cost vector arg and adjust.
6556 (vectorizable_call): Likewise.
6557 (vectorizable_simd_clone_call): Likewise.
6558 (vectorizable_conversion): Likewise.
6559 (vectorizable_assignment): Likewise.
6560 (vectorizable_shift): Likewise.
6561 (vectorizable_operation): Likewise.
6562 (vectorizable_store): Likewise.
6563 (vectorizable_load): Likewise.
6564 (vectorizable_condition): Likewise.
6565 (vectorizable_comparison): Likewise.
6566 (can_vectorize_live_stmts): Likewise.
6567 (vect_analyze_stmt): Likewise.
6568 (vect_transform_stmt): Adjust calls to vectorizable_*.
6569 * tree-vectorizer.c: Include gimple-pretty-print.h.
6570 (dump_stmt_cost): New function.
6572 2018-05-16 Richard Biener <rguenther@suse.de>
6574 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6575 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6576 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6577 (check_name): New callback.
6578 (dse_classify_store): Track cycles via a visited bitmap of PHI
6579 defs and simplify handling of in-loop and across loop dead stores
6580 and properly fail for loop-variant refs. Handle byte-tracking with
6581 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6584 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6586 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6587 (vect_get_mask_type_for_stmt): Likewise.
6588 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6590 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6591 to determine the statement's vector type and the vector type that
6592 should be used for calculating nunits. Deal with cases in which
6593 the type has to be deferred.
6594 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6595 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6596 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6597 (vect_determine_vf_for_stmt): New functions, split out from...
6598 (vect_determine_vectorization_factor): ...here.
6599 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6600 (vect_get_mask_type_for_stmt): New functions, split out from
6601 vect_determine_vectorization_factor.
6603 2018-05-16 Richard Biener <rguenther@suse.de>
6605 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6606 verify the [VEC_]COND_EXPR embedded comparison.
6608 2018-05-15 Martin Sebor <msebor@redhat.com>
6610 PR tree-optimization/85753
6611 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6612 RECORD_TYPE in addition to ARRAY_TYPE.
6614 2018-05-15 Martin Sebor <msebor@redhat.com>
6617 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6619 2018-05-15 Richard Biener <rguenther@suse.de>
6621 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6622 add by_clobber_p one. Change algorithm to collect all defs
6623 representing uses we need to walk and try reducing them to
6624 a single one before failing.
6625 (dse_dom_walker::dse_optimize_stmt): Adjust.
6627 2018-05-13 Mark Wielaard <mark@klomp.org>
6629 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6630 (size_of_loc_descr): Likewise.
6631 (output_loc_operands): Likewise.
6632 (output_loc_operands_raw): Likewise.
6633 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6634 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6635 (hash_loc_operands): Likewise.
6636 (compare_loc_operands): Likewise.
6638 2018-05-14 Mark Wielaard <mark@klomp.org>
6640 * dwarf2out.c (count_index_addrs): New function.
6641 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6643 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6645 PR tree-optimization/83648
6646 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6647 return value as malloc candidate.
6649 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6652 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6653 param as true in call to suggest_attribute.
6655 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6657 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6658 -mreadonly-in-sdata.
6660 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6662 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6664 (aarch64_crypto_aesd_fused): Likewise.
6666 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6668 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6669 (movsi_aarch64): Likewise.
6670 (load_pairsi): Likewise.
6671 (load_pairdi): Likewise.
6672 (store_pairsi): Likewise.
6673 (store_pairdi): Likewise.
6674 (load_pairsf): Likewise.
6675 (load_pairdf): Likewise.
6676 (store_pairsf): Likewise.
6677 (store_pairdf): Likewise.
6678 (zero_extend): Likewise.
6679 (trunc): Swap alternatives.
6680 (fcvt_target): Add '?' to prefer w over r.
6682 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6685 * config/i386/i386.md: Disallow non-commutative arithmetics in
6686 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6687 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6688 in the peephole2 before it.
6690 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6692 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6693 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6694 (ix86_handle_option): Handle -mcldemote.
6695 * config.gcc: New header.
6696 * config/i386/cldemoteintrin.h: New file.
6697 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6698 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6700 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6701 OPTION_MASK_ISA_CLDEMOTE.
6702 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6703 (ix86_valid_target_attribute_inner_p): Ditto.
6704 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6705 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6706 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6707 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6708 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6710 * config/i386/i386.opt: Add -mcldemote.
6711 * config/i386/x86intrin.h: New header.
6712 * doc/invoke.texi: Add -mcldemote.
6714 2018-05-14 Richard Biener <rguenther@suse.de>
6716 * doc/match-and-simplify.texi: Adjust :s documentation.
6718 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6720 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6721 intended memcpy size.
6722 (REORDER_45): Likewise.
6724 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6726 * sort.cc: New file.
6727 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6728 * vec.c (qsort_chk): Use gcc_qsort.
6729 * Makefile.in (OBJS-libcommon): Add sort.o.
6730 (build/sort.o): New target. Use it...
6731 (BUILD_RTL): ... here, and...
6732 (build/gencfn-macros): ... here, and...
6733 (build/genmatch): ... here.
6735 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6736 Chung-Ju Wu <jasonwucj@gmail.com>
6738 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6739 * config/nds32/nds32-graywolf.md: New file.
6740 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6741 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6743 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6744 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6745 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6746 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6747 * config/nds32/pipelines.md: Include n15 settings.
6749 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6750 Chung-Ju Wu <jasonwucj@gmail.com>
6752 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6753 * config/nds32/nds32-n13.md: New file.
6754 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6755 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6757 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6758 * config/nds32/nds32.md (pipeline_model): Add n13.
6759 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6760 * config/nds32/pipelines.md: Include n13 settings.
6762 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6763 Chung-Ju Wu <jasonwucj@gmail.com>
6765 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6766 * config/nds32/nds32-n10.md: New file.
6767 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6768 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6770 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6771 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6772 * config/nds32/nds32.md (pipeline_model): Add n10.
6773 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6774 * config/nds32/pipelines.md: Include n10 settings.
6776 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6777 Kito Cheng <kito.cheng@gmail.com>
6778 Chung-Ju Wu <jasonwucj@gmail.com>
6780 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6781 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6782 Add enum values for DSP extension instructions.
6783 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6785 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6786 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6788 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6789 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6790 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6791 * config/nds32/nds32-intrinsic.md: Likewise.
6792 * config/nds32/nds32_intrinsic.h: Likewise.
6793 * config/nds32/nds32-md-auxiliary.c: Likewise.
6794 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6795 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6796 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6797 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6798 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6799 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6800 TYPE_DMAC in switch statement.
6801 * config/nds32/nds32.c: New checking and implementation for DSP
6802 extension instructions.
6803 * config/nds32/nds32.h: Likewise.
6804 * config/nds32/nds32.md: Likewise.
6805 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6806 * config/nds32/predicates.md: Implement new predicates for DSP
6809 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6811 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6812 Reformat alternatives and attributes so it is easier to identify
6813 which constraints/attributes go with which instruction.
6814 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6815 (mov<mode>_softfloat32, FMOVE64): Likewise.
6816 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6817 (mov<mode>_softfloat64, FMOVE64): Likewise.
6819 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6821 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6823 (Basic PowerPC Built-in Functions): The new name of the
6824 subsection previously known as "PowerPC Built-in Functions".
6825 (Basic PowerPC Built-in Functions Available on all Configurations):
6827 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6828 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6829 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6830 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6832 2018-05-11 Martin Jambor <mjambor@suse.cz>
6835 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6838 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6841 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6843 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6845 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6846 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6847 (ix86_handle_option): Handle -mwaitpkg.
6848 * config.gcc: New header.
6849 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6850 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6851 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6853 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6854 OPTION_MASK_ISA_WAITPKG.
6855 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6856 (ix86_option_override_internal): Add PTA_WAITPKG.
6857 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6858 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6859 IX86_BUILTIN_TPAUSE.
6860 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6861 __builtin_ia32_umwait and __builtin_ia32_tpause.
6862 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6863 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6864 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6865 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6866 UNSPECV_TPAUSE): New.
6867 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6868 * config/i386/i386.opt: Add -mwaitpkg.
6869 * config/i386/waitpkgintrin.h: New file.
6870 * config/i386/x86intrin.h: New header.
6871 * doc/invoke.texi: Add -mwaitpkg.
6873 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6876 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6878 (cortex-m0): Use armv6s-m isa.
6879 (cortex-m0plus): Likewise.
6880 (cortex-m1): Likewise.
6881 (cortex-m0.small-multiply): Likewise.
6882 (cortex-m0plus.small-multiply): Likewise.
6883 (cortex-m1.small-multiply): Likewise.
6885 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6886 Jakub Jelinek <jakub@redhat.com>
6888 PR tree-optimization/85692
6889 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6890 source permute as well.
6892 2018-05-11 Martin Liska <mliska@suse.cz>
6895 * doc/extend.texi: Document LLVM style format for no_sanitize
6898 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6900 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6901 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6902 (mode_supports_vsx_dform_quad): Likewise.
6903 (mode_supports_vmx_dform): Move these functions to be next to the
6904 other mode_supports functions.
6905 (mode_supports_dq_form): Likewise.
6906 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6907 mode_supports_dq_form.
6908 (reg_offset_addressing_ok_p): Likewise.
6909 (offsettable_ok_by_alignment): Likewise.
6910 (rs6000_legitimate_offset_address_p): Likewise.
6911 (legitimate_lo_sum_address_p): Likewise.
6912 (rs6000_legitimize_address): Likewise.
6913 (rs6000_legitimize_reload_address): Likewise.
6914 (rs6000_secondary_reload_inner): Likewise.
6915 (rs6000_preferred_reload_class): Likewise.
6916 (rs6000_output_move_128bit): Likewise.
6918 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6920 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6921 Generate SImode target register for null target.
6922 <case IX86_BUILTIN_XGETBV>: Ditto.
6923 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6924 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6926 2018-05-10 Carl Love <cel@us.ibm.com>
6928 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6929 dcbtt and dcbtstt if operands[2] is 0.
6931 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6934 * config/i386/sse.md (usadv64qi): New expander.
6936 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6938 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6939 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6940 -maltivec=be support.
6941 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6942 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6943 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6944 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6945 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6946 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6947 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6948 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6949 altivec_vsumsws): Adjust.
6950 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6951 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6952 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6954 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6955 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6956 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6957 (altivec_lve<VI_char>x): Delete expand.
6958 (*altivec_lve<VI_char>x_internal): Rename to...
6959 (altivec_lve<VI_char>x): ... this.
6960 (altivec_lvxl_<mode>): Delete expand.
6961 (*altivec_lvxl_<mode>_internal): Rename to ...
6962 (altivec_lvxl_<mode>): ... this.
6963 (altivec_stvxl_<mode>): Delete expand.
6964 (*altivec_stvxl_<mode>_internal): Rename to ...
6965 (altivec_stvxl_<mode>): ... this.
6966 (altivec_stve<VI_char>x): Delete expand.
6967 (*altivec_stve<VI_char>x_internal): Rename to ...
6968 (altivec_stve<VI_char>x): ... this.
6969 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6970 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6971 reduc_plus_scal_<mode>): Adjust.
6972 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6974 (rs6000_cpu_cpp_builtins): Adjust.
6975 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6976 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6977 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6978 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6979 -maltivec=be support.
6980 (rs6000_split_vec_extract_var): Adjust.
6981 (rs6000_split_v4si_init): Adjust.
6982 (swap_selector_for_mode): Delete.
6983 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6984 altivec_expand_stvex_be): Delete.
6985 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6986 -maltivec=be support.
6987 (rs6000_gimple_fold_builtin): Ditto.
6988 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6990 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6991 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6992 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6993 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6994 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6995 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6996 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6997 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6998 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6999 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7000 anonymous split): Adjust.
7001 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7002 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7004 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7006 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7007 when --with-gxx-include-dir is also specified.
7008 * configure: Regenerate.
7010 2018-05-09 Jim Wilson <jimw@sifive.com>
7013 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7014 * config/riscv/t-withmultilib: New.
7015 * config/riscv/withmultilib.h: New.
7016 * doc/install.texi: Document RISC-V --with-multilib-list support.
7018 2018-05-09 Richard Biener <rguenther@suse.de>
7020 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7022 (vect_bb_vectorization_profitable_p): Adjust. Compute
7023 actual scalar cost using the cost vector and the add_stmt_cost
7026 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7028 PR rtl-optimization/85645
7029 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7030 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7032 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7034 PR rtl-optimization/85645
7035 * shrink-wrap.c (spread_components): Return a boolean saying if
7036 anything was changed.
7037 (try_shrink_wrapping_separate): Iterate spread_components until
7038 nothing changes anymore.
7040 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7042 PR rtl-optimization/85645
7043 * regrename.c (build_def_use): Also kill the chains that include the
7044 destination of a REG_CFA_REGISTER note.
7046 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7048 PR rtl-optimization/85645
7049 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7050 insn that has a REG_CFA_REGISTER note.
7052 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7054 * cfgexpand.c (expand_clobber): New function.
7055 (expand_gimple_stmt_1): Use it.
7056 * tree-vect-stmts.c (vect_clobber_variable): New function,
7058 (vectorizable_simd_clone_call): ...here.
7059 (vectorizable_store): Emit a clobber either side of an
7060 IFN_STORE_LANES sequence.
7061 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7063 2018-05-09 Tom de Vries <tom@codesourcery.com>
7066 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7067 (define_insn "trap_if_false"): Add exit after trap.
7069 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7071 PR rtl-optimization/85638
7072 * bb-reorder.c: Include common/common-target.h.
7073 (create_forwarder_block): New function extracted from...
7074 (fix_up_crossing_landing_pad): ...here. Rename into...
7075 (dw2_fix_up_crossing_landing_pad): ...this.
7076 (sjlj_fix_up_crossing_landing_pad): New function.
7077 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7078 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7079 from both partitions and exit the loop after one iteration.
7081 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7084 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7086 (Basic PowerPC Built-in Functions): The new name of the
7087 subsection previously known as "PowerPC Built-in Functions".
7088 (Basic PowerPC Built-in Functions Available on all Configurations):
7090 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7092 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7094 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7096 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7099 2018-05-08 Jim Wilson <jimw@sifive.com>
7101 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7102 (LD_EMUL_SUFFIX): New.
7103 (LINK_SPEC): Use it.
7105 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7107 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7109 (Basic PowerPC Built-in Functions): The new name of the
7110 subsection previously known as "PowerPC Built-in Functions".
7111 (Basic PowerPC Built-in Functions Available on all Configurations):
7113 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7115 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7117 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7119 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7122 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7125 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7126 after cmpelim optimization.
7128 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7130 * config.gcc: Support "goldmont".
7131 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7132 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7134 * config/i386/i386.c (m_GOLDMONT): Define.
7135 (processor_target_table): Add "goldmont".
7136 (PTA_GOLDMONT): Define.
7137 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7138 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7139 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7140 (fold_builtin_cpu): Add "goldmont".
7141 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7142 (ix86_option_override_internal): Add "goldmont".
7143 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7144 (processor_type): Add PROCESSOR_GOLDMONT.
7145 * config/i386/i386.md: Add CPU "glm".
7146 * config/i386/glm.md: New file.
7147 * config/i386/x86-tune.def: Add m_GOLDMONT.
7148 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7150 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7153 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7155 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7156 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7157 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7160 * config/i386/i386.c (ix86_fold_builtin): Handle
7161 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7164 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7165 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7167 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7170 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7171 (check_arch): Likewise.
7172 (check_fpu): Return the result rather than printing it.
7173 (end arch): Fix operator precedence.
7174 (end cpu): Likewise.
7175 (END): Print the result from check_fpu.
7177 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7178 Alan Hayward <alan.hayward@arm.com>
7179 David Sherwood <david.sherwood@arm.com>
7181 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7182 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7183 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7184 (*fcmuo<mode>_and): New patterns.
7186 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7188 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7189 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7190 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7191 (cmp_op, sve_imm_con): New code attributes.
7192 (SVE_COND_INT_CMP, imm_con): Delete.
7193 (cmp_op): Remove above unspecs from int attribute.
7194 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7196 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7197 comparison-specific unspecs.
7198 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7199 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7200 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7201 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7202 (*vec_fcm<cmp_op><mode>): Rename to...
7203 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7204 (*vec_fcmuo<mode>): Rename to...
7205 (*fcmuo<mode>): ...this and adjust likewise.
7206 (*pred_fcm<cmp_op><mode>): New pattern.
7207 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7208 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7210 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7212 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7213 (aarch64_emit_sve_predicated_cond): New function.
7214 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7215 (aarch64_emit_unspec_cond_or): Replace with...
7216 (aarch64_emit_sve_or_conds): ...this new function. Use
7217 aarch64_emit_sve_ptrue_op for the individual comparisons and
7218 aarch64_emit_binop to OR them together.
7219 (aarch64_emit_inverted_unspec_cond): Replace with...
7220 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7221 aarch64_emit_sve_ptrue_op for the comparison and
7222 aarch64_emit_unop to invert the result.
7223 (aarch64_expand_sve_vec_cmp_float): Update after the above
7224 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7226 2018-05-07 Nathan Sidwell <nathan@acm.org>
7228 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7229 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7230 (Backwards Compatibility): Likewise.
7232 2018-05-07 Luis Machado <luis.machado@linaro.org>
7236 2018-05-07 Luis Machado <luis.machado@linaro.org>
7238 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7239 <prefetch_dynamic_strides>: New const bool field.
7240 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7241 prefetch_dynamic_strides.
7242 (exynosm1_prefetch_tune): Likewise.
7243 (thunderxt88_prefetch_tune): Likewise.
7244 (thunderx_prefetch_tune): Likewise.
7245 (thunderx2t99_prefetch_tune): Likewise.
7246 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7248 (aarch64_override_options_internal): Update to set
7249 PARAM_PREFETCH_DYNAMIC_STRIDES.
7250 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7251 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7252 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7253 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7254 prefetch-dynamic-strides setting.
7256 2018-05-07 Luis Machado <luis.machado@linaro.org>
7258 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7259 <minimum_stride>: New const int field.
7260 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7261 minimum_stride field.
7262 (exynosm1_prefetch_tune): Likewise.
7263 (thunderxt88_prefetch_tune): Likewise.
7264 (thunderx_prefetch_tune): Likewise.
7265 (thunderx2t99_prefetch_tune): Likewise.
7266 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7267 (aarch64_override_options_internal): Update to set
7268 PARAM_PREFETCH_MINIMUM_STRIDE.
7269 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7270 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7271 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7272 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7273 stride is constant and is below the minimum stride threshold.
7275 2018-05-07 Luis Machado <luis.machado@linaro.org>
7277 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7280 2018-05-07 Luis Machado <luis.machado@linaro.org>
7282 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7283 <prefetch_dynamic_strides>: New const bool field.
7284 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7285 prefetch_dynamic_strides.
7286 (exynosm1_prefetch_tune): Likewise.
7287 (thunderxt88_prefetch_tune): Likewise.
7288 (thunderx_prefetch_tune): Likewise.
7289 (thunderx2t99_prefetch_tune): Likewise.
7290 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7292 (aarch64_override_options_internal): Update to set
7293 PARAM_PREFETCH_DYNAMIC_STRIDES.
7294 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7295 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7296 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7297 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7298 prefetch-dynamic-strides setting.
7300 2018-05-07 Luis Machado <luis.machado@linaro.org>
7302 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7303 <minimum_stride>: New const int field.
7304 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7305 minimum_stride field.
7306 (exynosm1_prefetch_tune): Likewise.
7307 (thunderxt88_prefetch_tune): Likewise.
7308 (thunderx_prefetch_tune): Likewise.
7309 (thunderx2t99_prefetch_tune): Likewise.
7310 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7311 (aarch64_override_options_internal): Update to set
7312 PARAM_PREFETCH_MINIMUM_STRIDE.
7313 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7314 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7315 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7316 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7317 stride is constant and is below the minimum stride threshold.
7319 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7322 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7323 the type is addressable. Don't force op into register if it has
7326 2018-05-05 Roland McGrath <mcgrathr@google.com>
7329 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7330 any section for which we don't know a specific type it should have,
7331 regardless of name. Previously this was done only for the exact
7332 names ".init_array", ".fini_array", and ".preinit_array".
7333 (default_elf_asm_named_section): Add comment about
7334 relationship with default_section_type_flags and SECTION_NOTYPE.
7335 (get_section): Don't consider it a type conflict if one side has
7336 SECTION_NOTYPE and the other doesn't, as long as neither has the
7337 SECTION_BSS et al used in the default_section_type_flags logic.
7339 2018-05-05 Tom de Vries <tom@codesourcery.com>
7342 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7343 (workaround_barsyncs): New function.
7344 (nvptx_reorg): Use workaround_barsyncs.
7345 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7346 (define_expand "nvptx_membar_cta"): New define_expand.
7347 (define_insn "*nvptx_membar_cta"): New insn.
7349 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7351 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7352 To improve optimization opportunities.
7353 * builtin-types.def: The new needed builtin types for the above.
7355 2018-05-04 Richard Biener <rguenther@suse.de>
7357 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7358 * gimple-ssa-store-merging.c
7359 (imm_store_chain_info::output_merged_store): Remove redundant create,
7360 release split_store vector contents on failure.
7361 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7362 scalar stmt vector on cache hit.
7364 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7366 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7368 * config.gcc (powerpc-xilinx-eabi*): Remove.
7369 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7371 (fusion_addis_mem_combo_load): Ditto.
7372 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7374 (rs6000_cpu_cpp_builtins): Ditto.
7375 * config/rs6000/rs6000-linux.c
7376 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7377 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7378 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7380 (rs6000_setup_reg_addr_masks): Ditto.
7381 (rs6000_init_hard_regno_mode_ok): Ditto.
7382 (rs6000_option_override_internal): Ditto.
7383 (legitimate_lo_sum_address_p): Ditto.
7384 (rs6000_legitimize_address): Ditto.
7385 (rs6000_legitimize_reload_address): Ditto.
7386 (rs6000_legitimate_address_p): Ditto.
7387 (abi_v4_pass_in_fpr): Ditto.
7388 (setup_incoming_varargs): Ditto.
7389 (rs6000_gimplify_va_arg): Ditto.
7390 (rs6000_split_multireg_move): Ditto.
7391 (rs6000_savres_strategy): Ditto.
7392 (rs6000_emit_prologue_components): Ditto.
7393 (rs6000_emit_epilogue_components): Ditto.
7394 (rs6000_emit_prologue): Ditto.
7395 (rs6000_emit_epilogue): Ditto.
7396 (rs6000_elf_file_end): Ditto.
7397 (rs6000_function_value): Ditto.
7398 (rs6000_libcall_value): Ditto.
7399 * config/rs6000/rs6000.h: Ditto.
7400 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7401 (TARGET_MINMAX): ... this. New.
7402 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7403 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7404 (*movsi_internal1_single): Delete.
7405 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7406 mfpu=, mxilinx-fpu): Delete.
7407 * config/rs6000/singlefp.h: Delete.
7408 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7409 * config/rs6000/t-rs6000: Ditto.
7410 * config/rs6000/t-xilinx: Delete.
7411 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7412 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7413 (VStype_simple): Delete.
7414 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7415 * config/rs6000/xfpu.h: Delete.
7416 * config/rs6000/xfpu.md: Delete.
7417 * config/rs6000/xilinx.h: Delete.
7418 * config/rs6000/xilinx.opt: Delete.
7419 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7420 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7422 2018-05-04 Tom de Vries <tom@codesourcery.com>
7425 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7428 2018-05-04 Richard Biener <rguenther@suse.de>
7431 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7432 (expand_complex_div_wide): Likewise.
7433 (expand_complex_operations_1): Likewise.
7434 (expand_complex_libcall): Preserve EH info of the original stmt.
7435 (tree_lower_complex): Handle removed blocks.
7436 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7437 on complex multiplication and division libcall builtins.
7439 2018-05-04 Richard Biener <rguenther@suse.de>
7442 * fold-const.c (negate_expr_p): Restrict negation of operand
7443 zero of a division to when we know that can happen without
7445 (fold_negate_expr_1): Likewise.
7447 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7450 * real.h (real_nextafter): Declare.
7451 * real.c (real_nextafter): New function.
7452 * fold-const-call.c (fold_const_nextafter): New function.
7453 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7454 CASE_CFN_NEXTTOWARD.
7455 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7456 even when arg1_mode is different from arg0_mode.
7458 2018-05-03 Nathan Sidwell <nathan@acm.org>
7460 * doc/extend.texi (Deprecated Features): Remove
7462 (Backwards Compatibility): Likewise.
7463 * doc/invoke.texi (C++ Language Options): Likewise.
7464 (C++ Dialect Options): Likewise.
7466 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7469 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7470 _mm512_mask_mullox_epi64): New intrinsics.
7472 2018-05-03 Tom de Vries <tom@codesourcery.com>
7475 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7476 dump files): Add offload-tree.
7478 2018-05-03 Richard Biener <rguenther@suse.de>
7480 PR tree-optimization/85615
7481 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7482 to loops not nested in BBs loop father to avoid creating multi-entry
7485 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7487 PR tree-optimization/70291
7488 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7489 arguments. Change return type to tree. Emit libcall as a new
7490 statement rather than replacing existing one when inplace_p is true.
7491 (expand_complex_multiplication_components): New function.
7492 (expand_complex_multiplication): Expand floating-point complex
7493 multiplication using the above.
7494 (expand_complex_division): Rename inner_type parameter to type.
7495 Update expand_complex_libcall call-site.
7496 (expand_complex_operations_1): Update expand_complex_multiplication
7497 and expand_complex_division call-sites.
7499 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7502 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7503 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7504 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7505 the highest significant bit of the shift count mask is clear. In
7506 check whether and[sq]i3 is needed verify that all significant bits
7507 of the shift count other than the highest are set.
7509 2018-05-02 Tom de Vries <tom@codesourcery.com>
7512 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7513 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7514 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7515 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7516 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7517 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7518 __builtin_goacc_parlevel_size.
7520 2018-05-02 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/85597
7523 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7524 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7527 2018-05-02 Tom de Vries <tom@codesourcery.com>
7530 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7531 dump files): Add ltrans-tree.
7533 2018-05-02 Tom de Vries <tom@codesourcery.com>
7536 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7537 dump files): Add wpa-ipa.
7539 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7541 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7542 powerpc*-*-linux*paired* target.
7543 * config/rs6000/750cl.h: Delete.
7544 * config/rs6000/paired.h: Delete.
7545 * config/rs6000/paired.md: Delete.
7546 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7548 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7549 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7550 comment. Remove paired float support.
7551 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7552 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7554 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7555 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7557 * config/rs6000/rs6000.c: Remove paired float support.
7558 (paired_expand_vector_init, paired_expand_vector_move,
7559 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7560 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7561 paired_expand_builtin, paired_expand_predicate_builtin,
7562 paired_init_builtins): Delete.
7563 * config/rs6000/rs6000.h: Remove paired float support.
7564 * config/rs6000/rs6000.md: Remove paired float support.
7565 (move_from_CR_ov_bit): Delete.
7566 * config/rs6000/rs6000.opt (mpaired): Delete.
7567 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7568 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7570 2018-05-02 Richard Biener <rguenther@suse.de>
7573 * gimplify.c (gimplify_save_expr): When in SSA form allow
7574 SAVE_EXPRs to compute to SSA vars.
7576 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7579 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7580 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7581 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7582 clobber operands[2], instead use a new pseudo. Formatting fixes.
7584 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7586 PR tree-optimization/85586
7587 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7588 exit early for statements in the same group if the accesses are
7591 2018-05-02 Tom de Vries <tom@codesourcery.com>
7594 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7597 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7599 PR tree-optimization/85143
7600 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7602 2018-05-01 Tom de Vries <tom@codesourcery.com>
7605 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7606 not found" error message.
7608 2018-05-01 Tom de Vries <tom@codesourcery.com>
7611 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7612 * vec.c (test_ordered_remove_if): New function.
7613 (vec_c_tests): Call test_ordered_remove_if.
7614 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7615 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7616 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7617 VEC_ORDERED_REMOVE_IF.
7619 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7621 PR tree-optimization/82665
7622 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7623 pointer subtraction where arguments come from a memchr call.
7625 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7627 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7628 --push-state --as-needed and --pop-state instead of --as-needed and
7629 --no-as-needed if ld supports it.
7630 * configure: Regenerated.
7633 * doc/install.texi2html: Replace _002d with - and _002a with * in
7634 generated html files using sed.
7636 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7639 * gcc-rich-location.c (blank_line_before_p): New function.
7640 (use_new_line): New function.
7641 (gcc_rich_location::add_fixit_insert_formatted): New function.
7642 * gcc-rich-location.h
7643 (gcc_rich_location::add_fixit_insert_formatted): New function.
7645 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7647 * selftest.c (assert_streq): Rename "expected" and "actual" to
7648 "val1" and "val2". Extend NULL-handling to cover both inputs
7649 symmetrically, while still requiring both to be non-NULL for a pass.
7650 * selftest.h (assert_streq): Rename "expected" and "actual" to
7652 (ASSERT_EQ): Likewise.
7653 (ASSERT_EQ_AT): Likewise.
7654 (ASSERT_KNOWN_EQ): Likewise.
7655 (ASSERT_KNOWN_EQ_AT): Likewise.
7656 (ASSERT_NE): Likewise.
7657 (ASSERT_MAYBE_NE): Likewise.
7658 (ASSERT_MAYBE_NE_AT): Likewise.
7659 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7660 the assertion to pass.
7661 (ASSERT_STREQ_AT): Likewise.
7663 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7665 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7666 interaction with -pie.
7668 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7670 * selftest.h: Fix alphabetization of per-source-file selftest
7673 2018-04-30 Jason Merrill <jason@redhat.com>
7675 PR c++/61982 - dead stores to destroyed objects.
7676 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7679 2018-04-30 Jason Merrill <jason@redhat.com>
7681 * tree.c (build_clobber): New.
7682 * tree.h: Declare it.
7683 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7685 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7687 * diagnostic-show-locus.c (layout::layout): Update for
7688 location_get_source_line returning a char_span.
7689 (struct char_span): Move to input.h.
7690 (struct correction): Update for fields in char_span becoming
7692 (struct source_line): Update for location_get_source_line
7693 returning a char_span.
7694 (layout::print_line): Likewise.
7695 * edit-context.c (edited_file::print_content): Likewise.
7696 (edited_file::print_diff_hunk): Likewise.
7697 (edited_file::print_run_of_changed_lines): Likewise.
7698 (edited_file::get_num_lines): Likewise.
7699 (edited_line::edited_line): Likewise.
7700 * final.c (asm_show_source): Likewise.
7701 * input.c (location_get_source_line): Convert return type
7702 from const char * to char_span, losing the final "line_len"
7704 (dump_location_info): Update for the above.
7705 (get_substring_ranges_for_loc): Likewise. Use a char_span
7706 when handling the literal within the line.
7707 (test_reading_source_line): Update for location_get_source_line
7708 returning a char_span.
7709 * input.h (class char_span): Move here from
7710 diagnostic-show-locus.c, converting from a struct to a class.
7711 Make data members private.
7712 (char_span::operator bool): New.
7713 (char_span::length): New.
7714 (char_span::get_buffer): New.
7715 (char_span::operator[]): New.
7716 (char_span::subspan): Make const.
7717 (char_span::xstrdup): New.
7718 (location_get_source_line): Convert return type from const char *
7719 to char_span, losing the final "line_size" param.
7721 2018-04-30 Jan Hubicka <jh@suse.cz>
7723 * lto-wrapper.c (ltrans_priorities): New static var.
7724 (cmp_priority): New.
7725 (run_gcc): Read priorities and if doing parallel build order
7726 the Makefile by them.
7728 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7730 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7732 2018-04-30 Richard Biener <rguenther@suse.de>
7734 * tree-cfg.c (verify_address): Remove base argument, add
7735 flag whether to check TREE_ADDRESSABLE and do that.
7736 (verify_expr): Remove.
7737 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7738 (verify_gimple_assign_single): Likewise.
7739 (verify_gimple_switch): Likewise.
7740 (verify_expr_location_1): Dereference tp once. Add (disabled)
7741 piece from verify_expr.
7742 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7744 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7746 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7748 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7750 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7751 (small_data_pattern): Likewise.
7752 (arc_rewrite_small_data): Likewise.
7753 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7754 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7755 (get_symbol_alignment): New function.
7756 (legitimate_small_data_address_p): Likewise.
7757 (legitimate_scaled_address): Update, call
7758 legitimate_small_data_address_p.
7759 (output_sdata): New static variable.
7760 (arc_print_operand): Update how we handle small data operands.
7761 (arc_print_operand_address): Likewise.
7762 (arc_legitimate_address_p): Update, use
7763 legitimate_small_data_address_p.
7764 (arc_rewrite_small_data_p): Remove.
7765 (arc_rewrite_small_data_1): Likewise.
7766 (arc_rewrite_small_data): Likewise.
7767 (small_data_pattern): Likewise.
7768 (compact_sda_memory_operand): Update to use
7769 legitimate_small_data_address_p and get_symbol_alignment.
7770 (prepare_move_operands): Don't rewite sdata pattern.
7771 (prepare_extend_operands): Remove.
7772 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7774 (zero_extendqisi2): Likewise.
7775 (zero_extendhisi2): Likewise.
7776 (extendqihi2): Likewise.
7777 (extendqisi2): Likewise.
7778 (extendhisi2): Likewise.
7782 * config/arc/constraints.md (Usd): Change it to memory constraint.
7784 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7786 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7787 as source of std instructions.
7788 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7789 allow 6-bit constants as source for store instructions.
7790 (movdi_insn): Update instruction pattern to allow 6-bit constants
7791 as source for store instructions.
7793 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7795 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7797 2018-04-30 Nathan Sidwell <nathan@acm.org>
7798 Sandra Loosemore <sandra@codesourcery.com>
7800 * dumpfile.c (dump_open): Allow '-' for stdout.
7801 * doc/invoke.texi (Developer Options): Document dump filename
7802 determination early. Document stdin/stdout selection.
7804 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7806 Microblaze Target: PIC data text relative
7808 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7809 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7811 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7812 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7813 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7814 New addressing mode for data-text relative position indepenedent code.
7815 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7816 'ADDRESS_SYMBOLIC_TXT_REL'.
7817 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7818 (microblaze_legitimate_pic_operand): Exclude function calls from
7819 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7820 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7822 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7823 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7824 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7825 for 'address + offset'.
7826 (microblaze_expand_prologue): Add new function prologue call for
7828 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7829 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7830 table in case of TARGET_PIC_DATA_TEXT_REL.
7831 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7832 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7833 Add new macros 'UNSPEC_TEXT',
7834 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7835 + exclude function calls from 'UNSPEC_PLT' in case of data text
7837 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7838 new target hook for generating address diff vector tables in case of
7840 * doc/tm.texi : Regenerate.
7841 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7842 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7843 of addr diff vector generation.
7844 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7845 target hook definition.
7846 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7847 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7848 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7849 option for data text relative.
7851 2018-04-30 Richard Biener <rguenther@suse.de>
7853 * tree-chrec.h (evolution_function_is_constant_p): Remove
7855 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7857 2018-04-30 Richard Biener <rguenther@suse.de>
7860 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7862 2018-04-30 Richard Biener <rguenther@suse.de>
7864 PR tree-optimization/28364
7865 PR tree-optimization/85275
7866 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7867 copying first exit test.
7869 2018-04-28 Mark Wielaard <mark@klomp.org>
7871 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7873 (dwarf_AT): Handle DW_AT_addr_base.
7874 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7876 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7879 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7880 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7881 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7882 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7884 2018-04-28 Richard Biener <rguenther@suse.de>
7886 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7887 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7888 to reflect use. Only add interesting stmts.
7890 2018-04-27 Martin Jambor <mjambor@suse.cz>
7893 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7894 the jump function allows for passing through aggregate values.
7896 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7898 * input.h (in_system_header_at): Convert from macro to inline
7900 (from_macro_expansion_at): Likewise.
7901 (from_macro_definition_at): Likewise.
7903 2018-04-27 Jeff Law <law@redhat.com>
7905 * config.gcc: Mark tile* targets as deprecated/obsolete.
7907 2018-04-27 Richard Biener <rguenther@suse.de>
7909 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7912 2018-04-27 Richard Biener <rguenther@suse.de>
7914 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7916 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7918 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7919 with Yd constraint. Set "preferred_for_speed" attribute from
7920 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7922 (*movdi_internal): Ditto.
7923 (movti_interunit splitters): Remove
7924 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7925 (movdi_interunit splitters): Ditto.
7926 * config/i386/constraints.md (Ye): Remove.
7927 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7929 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7932 * config/aarch64/constraints.md (Usg): Limit to 31.
7935 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7937 PR tree-optimization/85529
7938 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7939 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7940 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7941 zero extension or masking of the MSB bit.
7942 (optimize_range_tests): Add FIRST_BB argument, pass it through
7943 to optimize_range_tests_var_bound.
7944 (maybe_optimize_range_tests, reassociate_bb): Adjust
7945 optimize_range_tests callers.
7947 2018-04-26 Richard Biener <rguenther@suse.de>
7948 Jakub Jelinek <jakub@redhat.com>
7950 * cgraph.h (symbol_table): Just declare debug method here.
7951 * symtab.c (symbol_table::debug): Define.
7953 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7955 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7957 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7959 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7960 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7961 (*movdi_internal): Substitute Yi and Yj constraint with x
7962 and Ym and Yn constraint with y constraint. Update "isa"
7963 attribute and set "preferred_for_speed" attribute from
7964 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7965 (*movsi_internal): Ditto.
7966 (*movdf_internal): Ditto.
7967 (*movsf_internal): Ditto.
7968 (*zero_extendsidi2): Ditto.
7969 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7970 (sse2_loadld): Ditto.
7971 (*vec_extract<ssevecmodelower>_0): Ditto.
7972 (*vec_extractv4si_0_zext_sse4): Ditto.
7973 (vec_concatv2di): Ditto.
7974 (*vec_dup<mode>): Ditto.
7975 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7976 * config/i386/constraints.md (Yi): Remove.
7981 2018-04-26 Nathan Sidwell <nathan@acm.org>
7983 * dumpfile.c (dump_open): New.
7984 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7985 (dump_finish): Detect stdio/stderr by value not name.
7987 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7989 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7991 2018-04-26 Tom de Vries <tom@codesourcery.com>
7994 * config/nvptx/nvptx.c (verify_neutering_jumps)
7995 (verify_neutering_labels): New function
7996 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7998 2018-04-26 Tom de Vries <tom@codesourcery.com>
8001 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8002 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8005 2018-04-26 Richard Biener <rguenther@suse.de>
8006 Tom de Vries <tom@codesourcery.com>
8009 * lto-streamer-out.c (output_function): Fixup loops if required to match
8010 discovery done in the reader.
8012 2018-04-26 Richard Biener <rguenther@suse.de>
8014 PR tree-optimization/85116
8015 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8016 have a loop exit from the single latch predecessor. Remove
8017 case of header with just condition.
8018 (ch_base::copy_headers): Exclude infinite loops from any
8020 (pass_ch::execute): Record exits.
8022 2018-04-26 Richard Biener <rguenther@suse.de>
8024 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8025 prologue cost vector and pass it to vect_get_load_cost.
8026 (vect_get_peeling_costs_all_drs): Likewise.
8027 (vect_peeling_hash_get_lowest_cost): Likewise.
8028 (vect_enhance_data_refs_alignment): Likewise.
8030 2018-04-26 Richard Biener <rguenther@suse.de>
8033 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8034 checking of integer<->pointer conversions.
8035 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8036 sign-/zero-extending pointer types.
8037 (expand_omp_for_static_chunk): Likewise.
8039 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8040 Jean Lee <xiaoyur347@gmail.com>
8042 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8043 (TARGET_ASAN_SHADOW_OFFSET): Define.
8044 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8045 true for -fsanitize=address.
8047 2018-04-25 Mark Wielaard <mark@klomp.org>
8049 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8052 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8054 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8055 than "alu", remove explicit "memory" and "imm_disp" attributes.
8056 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8059 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8060 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8061 gen_lowpart_no_emit.
8063 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8066 * config/i386/i386.c (ix86_expand_builtin): Change memory
8067 operand to XI, extend p0 to Pmode.
8068 * config/i386/i386.md: Change unspec volatile and operand
8069 1 mode to XI, change operand 0 mode to P.
8071 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8073 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8074 GET_MODE_MASK before any checking.
8075 (nds32_can_use_bset_p): Likewise.
8076 (nds32_can_use_btgl_p): Likewise.
8078 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8080 * config/nds32/nds32-doubleword.md: New define_split pattern for
8081 illegal register number.
8083 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8085 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8087 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8089 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8091 2018-04-25 Richard Biener <rguenther@suse.de>
8093 * lto-streamer.h (LTO_major_version): Bump to 8.
8095 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8097 * BASE-VER: Set to 9.0.0.
8099 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8101 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8102 in __abskf2 and __powikf2.
8104 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8107 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8108 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8109 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8110 Use the above on operand 2. Reindent.
8111 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8113 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8116 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8118 * config/i386/i386.opt (mcet): Removed.
8119 * doc/install.texi: Remove -mcet documentation.
8120 * doc/invoke.texi: Likewise.
8122 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8125 * doc/install.texi: Remove -mcet from bootstrap-cet.
8127 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8130 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8131 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8135 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8136 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8137 containing a CONST_VECTOR.
8139 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8141 * doc/install.texi: Update newlib dependency for nvptx.
8143 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8146 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8147 instead of INTVAL when shifting x left.
8149 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8151 PR tree-optimization/85478
8152 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8153 vect_grouped_store_supported for single element vectors.
8155 2018-04-24 Richard Biener <rguenther@suse.de>
8158 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8159 load cost increase to the case of non-constant step.
8161 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8164 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8165 destination if any_malformed_asm.
8167 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8170 * expr.c (store_field): In the bitfield case, if the value comes from
8171 a function call and is returned in registers by means of a PARALLEL,
8172 do not change the mode of the temporary unless BLKmode and VOIDmode.
8174 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8176 PR rtl-optimization/85423
8177 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8178 dependencies to debug insns when the previous insn is non-debug.
8180 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8182 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8183 enums into a single definition.
8184 (fls): Fix predicates and printing.
8187 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8189 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8190 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8191 and short u6 immediate.
8192 (check_if_valid_sleep_operand): Remove.
8193 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8195 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8197 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8198 flag_always_save_lp condition.
8199 * config/nds32/nds32.opt (malways-save-lp): New option.
8201 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8203 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8204 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8205 * config/nds32/nds32.h
8206 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8207 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8209 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8211 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8212 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8214 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8215 Chung-Ju Wu <jasonwucj@gmail.com>
8217 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8219 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8220 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8222 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8224 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8226 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8228 * config/nds32/nds32-protos.h (nds32_data_alignment,
8229 nds32_local_alignment): Declare.
8230 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8231 nds32_local_alignment): New functions.
8232 (TARGET_CONSTANT_ALIGNMENT): Define.
8233 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8235 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8237 * config/nds32/nds32.c
8238 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8239 (TARGET_MODES_TIEABLE_P): Likewise.
8241 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8243 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8246 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8247 Chung-Ju Wu <jasonwucj@gmail.com>
8249 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8250 for unaligned access.
8251 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8253 * config/nds32/nds32-intrinsic.md: Likewise.
8254 * config/nds32/nds32_intrinsic.h: Likewise.
8255 * config/nds32/nds32.h (nds32_builtins): Likewise.
8256 * config/nds32/nds32.opt (munaligned-access): New option.
8257 * config/nds32/nds32.c (nds32_asm_file_start): Display
8258 flag_unaligned_access status.
8260 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8262 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8263 -mno-relax is present.
8264 * config/riscv/linux.h (LINK_SPEC): Ditto.
8266 2018-04-20 Martin Sebor <msebor@redhat.com>
8269 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8271 (gimple_fold_builtin_stxcpy_chk): Same.
8272 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8274 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8277 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8278 __powikf2 when long double is IEEE 128-bit.
8280 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8282 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8283 step to make sure stack always aligned.
8285 2018-04-20 Carl Love <cel@us.ibm.com>
8288 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8289 size check for arg0.
8291 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8292 Tom de Vries <tom@codesourcery.com>
8295 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8296 Emit insns for calls too.
8297 (nvptx_find_par): Always look for worker-level predecessor insn.
8298 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8300 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8301 (nvptx_process_pars): Propagate frames for calls.
8303 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8306 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8308 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8309 (ix86_handle_option): Don't handle OPT_mibt.
8310 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8312 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8314 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8315 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8316 (ix86_target_macros): Define __CET__ with flag_cf_protection
8317 for -fcf-protection.
8318 * config/i386/i386.c (isa2_opts): Remove -mibt.
8319 * config/i386/i386.h (TARGET_IBT): Removed.
8320 (TARGET_IBT_P): Likewise.
8321 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8322 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8323 * config/i386/i386.opt (mcet): Update help message.
8326 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8327 -mcet as an alias for -mshstk.
8329 2018-04-20 Richard Biener <rguenther@suse.de>
8332 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8333 complexity by forcing a single use of the multiply operand.
8335 2018-04-20 Martin Jambor <mjambor@suse.cz>
8338 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8339 recursion dependency to only apply to non-clones.
8341 2018-04-20 Martin Jambor <mjambor@suse.cz>
8344 * ipa-cp.c (create_specialized_node): Check that clones of
8345 self-recursive edges exist during IPA-CP.
8347 2018-04-19 Toon Moene <toon@moene.org>
8349 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8352 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8354 PR tree-optimization/85467
8355 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8356 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8357 VECTOR_CST element to type.
8359 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8362 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8363 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8364 (builtin_longjmp): Likewise.
8365 (save_stack_nonlocal): New pattern.
8366 (restore_stack_nonlocal): Likewise.
8368 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8371 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8372 Replace ASM_OUTPUT_LABEL with fprintf.
8374 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8377 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8378 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8379 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8380 define __IBT__ and __SHSTK__ for -fcf-protection.
8381 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8383 (ix86_trampoline_init): Likewise.
8384 (x86_output_mi_thunk): Likewise.
8385 (ix86_notrack_prefixed_insn_p): Likewise.
8386 (ix86_option_override_internal): Don't disallow -fcf-protection.
8387 * config/i386/i386.md (rdssp<mode>): Also enable for
8389 (incssp<mode>): Likewise.
8390 (nop_endbr): Likewise.
8391 * config/i386/i386.opt (mcet): Change help message to built-in
8395 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8396 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8397 enable CET built-in functions.
8399 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8401 * common/config/i386/i386-common.c
8402 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8403 OPTION_MASK_ISA_MOVDIRI_UNSET,
8404 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8405 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8406 * config.gcc (movdirintrin.h): New header.
8407 * config/i386/cpuid.h (bit_MOVDIRI,
8408 bit_MOVDIR64B): New bits.
8409 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8411 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8412 (VOID, PVOID, PCVOID)): New function types.
8413 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8414 __builtin_ia32_directstoreu_u64,
8415 __builtin_ia32_movdir64b): New builtins.
8416 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8417 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8419 (ix86_valid_target_attribute_inner_p): Ditto.
8420 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8421 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8422 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8423 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8424 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8425 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8426 (movdiri<mode>, movdir64b_<mode>): New.
8427 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8428 * config/i386/immintrin.h: Include movdirintrin.h.
8429 * config/i386/movdirintrin.h: New file.
8430 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8432 2018-04-19 Richard Biener <rguenther@suse.de>
8435 * cfg.c (clear_bb_flags): When loop state says we have
8436 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8438 2018-04-19 Richard Biener <rguenther@suse.de>
8440 PR tree-optimization/84737
8441 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8442 copying restrict info.
8443 (vect_setup_realignment): Use it.
8444 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8445 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8446 the first DR to all generated stores.
8447 (vectorizable_load): Likewise for loads.
8449 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8451 PR tree-optimization/85446
8452 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8453 the integral and pointer types to have the same precision.
8455 * doc/install.texi: Document --disable-cet being the default and
8458 2018-04-18 Martin Liska <mliska@suse.cz>
8460 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8463 2018-04-18 Martin Liska <mliska@suse.cz>
8466 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8469 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8470 arguments if they are comparable.
8472 2018-04-18 Martin Liska <mliska@suse.cz>
8475 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8478 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8481 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8484 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8485 ENDBR after calling __morestack.
8487 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8490 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8491 by using gcc_base_ver to generate a gcc_driver_version, and use
8492 it when generating GCC_DRIVER_NAME.
8493 * configure: Regenerate.
8495 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8498 * config.gcc: Obsolete powerpc*-*-*spe*.
8500 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8503 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8504 (stabstr_D): Change type of unum from unsigned int to
8505 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8508 2018-04-17 Jim Wilson <jimw@sifive.com>
8511 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8512 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8513 Set arg_pointer_offset after using pretend_args_size.
8515 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8517 PR rtl-optimization/85431
8518 * dse.c (record_store): Ignore zero width stores.
8521 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8522 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8523 __builtin_stack_restore rather than after it.
8524 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8525 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8526 argument instead of virtual_dynamic_stack_rtx.
8528 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8530 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8532 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8533 Add note to error message to explain internal mapping of overloaded
8534 built-in function name to non-overloaded built-in function name.
8535 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8538 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8541 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8542 where the inputs overlap with the output.
8544 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8546 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8547 (=v, v) alternative and explicit "memory" attribute.
8548 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8549 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8551 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8552 "sselog1" type instead of "sselog".
8553 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8554 "sselog". Remove explicit "memory" attribute.
8555 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8556 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8558 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8559 "sselog1" type instead of "sselog". Remove explicit "memory"
8561 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8562 use "sselog1" type instead of "sselog". Remove explicit "memory"
8564 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8565 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8567 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8568 "sselog1" type instead of "sselog". Remove explicit "memory"
8570 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8571 use "sselog1" type instead of "sselog". Remove explicit "memory"
8575 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8578 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8581 2018-04-17 Martin Jambor <mjambor@suse.cz>
8584 * ipa-cp.c (create_specialized_node): Call
8585 expand_all_artificial_thunks if necessary.
8587 2018-04-17 Martin Liska <mliska@suse.cz>
8590 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8591 in message, remote space in between '_G' and '('.
8593 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8596 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8597 avx512f_vmcmp<mode>3<round_saeonly_name>,
8598 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8599 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8600 avx512f_rndscale<mode><round_saeonly_name>,
8601 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8602 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8603 Use %<iptr>2 instead of %2 for -masm=intel.
8604 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8605 avx512f_vcvttss2usi<round_saeonly_name>,
8606 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8608 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8609 avx512f_vcvttsd2usi<round_saeonly_name>,
8610 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8611 Use %q1 instead of %1 for -masm=intel.
8612 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8613 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8614 of %3 for -masm=intel.
8615 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8617 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8619 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8621 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8622 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8624 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8626 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8627 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8628 %g1 and one with %0 and %1.
8629 (avx512er_vmrcp28<mode><round_saeonly_name>,
8630 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8632 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8633 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8634 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8635 of %0 and %{%4%} for -masm=intel.
8636 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8637 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8638 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8639 order of %0 and %{%5%}%{z%} for -masm=intel.
8641 2018-04-17 Jan Hubicka <jh@suse.cz>
8644 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8646 2018-04-17 Martin Liska <mliska@suse.cz>
8649 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8650 for target_clone error message. Make default implementation
8651 clone to be a local declaration.
8652 (separate_attrs): Add new argument and check for an empty
8654 (expand_target_clones): Handle it.
8655 (ipa_target_clone): Make redirection just for target_clones
8658 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8659 Tom de Vries <tom@codesourcery.com>
8662 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8663 tiled basic blocks without omp continue statements.
8665 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8668 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8669 vec_extract expression as having side effects to make sure it gets
8672 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8675 * config/i386/i386.c (get_builtin_code_for_version): Check
8678 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8681 * gcc/config.gcc: Support "skylake".
8682 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8684 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8685 (processor_target_table): Add "skylake".
8686 (ix86_option_override_internal): Add "skylake".
8687 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8688 PROCESSOR_CANNONLAKE.
8689 (get_builtin_code_for_version): Fix priority for
8690 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8691 PROCESSOR_SKYLAKE-AVX512.
8692 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8693 (processor_type): Add PROCESSOR_SKYLAKE.
8695 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8696 Jason Merrill <jason@redhat.com>
8699 * convert.c (convert_to_integer_1): Use direct recursion for
8700 enumeral types and types with a precision less than the number
8701 of bits in their mode.
8703 2018-04-16 Julia Koval <julia.koval@intel.com>
8706 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8707 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8709 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8712 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8713 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8714 and -mno-direct-move.
8716 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8719 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8720 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8721 Ensure negative shifts result in {0}.
8723 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8725 PR rtl-optimization/79916
8726 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8727 regs (if any) to define how to gnerate SD moves when LRA is in
8730 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8732 PR rtl-optimization/85393
8733 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8734 * except.c (expand_dw2_landing_pad_for_region): Make static.
8735 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8736 a label and unconditional jump to old_bb, rather than
8737 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8740 PR rtl-optimization/85376
8741 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8742 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8743 instead of a specific value.
8745 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8746 Bin Cheng <bin.cheng@arm.com>
8748 PR tree-optimization/82965
8749 PR tree-optimization/83991
8750 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8751 by_profile_only parameter.
8752 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8753 information if the loop was predicted to iterate too many times.
8754 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8756 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8759 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8762 2018-04-13 Martin Liska <mliska@suse.cz>
8763 Jakub Jelinek <jakub@redhat.com>
8766 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8767 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8768 * builtins.c (expand_builtin_memory_copy_args): Use
8769 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8770 handle dest_addr == pc_rtx.
8772 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8775 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8776 asked to not generate direct moves.
8777 (fix_trunc<mode>si2_stfiwx): Similar.
8778 (fix_trunc<mode>si2_internal): Similar.
8780 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8783 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8784 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8785 lookup if dest in some wider mode is known to be const0_rtx and
8786 if so, record permanent equivalence for it to be ZERO_EXTEND of
8787 the narrower mode destination.
8789 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8791 * lto-streamer-out.c (output_function): Revert 259346.
8792 * omp-expand.c (expand_oacc_for): Likewise.
8794 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8796 PR rtl-optimization/85354
8797 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8798 * sel-sched.c (sel_global_init): ... here.
8800 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8803 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8804 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8805 mode for PE-COFF targets.
8806 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8807 (i386_pe_asm_lto_end): Likewise.
8808 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8809 (TARGET_ASM_LTO_END): Likewise.
8810 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8811 (i386_pe_asm_lto_start): New function.
8812 (i386_pe_asm_lto_end): Likewise.
8814 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8815 Richard Biener <rguenther@suse.de>
8818 * lto-streamer-out.c (output_function): Fix CFG loop state before
8820 * omp-expand.c (expand_oacc_for): Handle calls to internal
8821 functions like regular functions.
8823 2018-04-12 Richard Biener <rguenther@suse.de>
8826 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8827 for the early LTO debug to properly generate references to it
8828 during DIE emission. Do not re-use that for the skeleton for
8830 (dwarf2out_early_finish): Likewise.
8832 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8835 * config/i386/sse.md
8836 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8837 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8838 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8839 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8840 and output is a reg, avoid creating invalid lowpart subreg, but
8841 instead split into a 512-bit move. Don't split if not AVX512VL,
8842 input is xmm16+ reg and output is a mem.
8843 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8844 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8845 xmm16+ reg and output is a mem.
8847 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8849 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8850 also for flag_dwarf2_cfi_asm.
8852 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8854 PR rtl-optimization/85342
8855 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8856 a bool scalar var inside of the loop instead. Don't try to update
8857 recog_data.operand after failed apply_change_group.
8859 2018-04-12 Tom de Vries <tom@codesourcery.com>
8862 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8863 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8864 array with flexible array member as array without given dimension.
8865 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8866 argument for undefined param to true.
8868 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8871 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8872 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8873 from PowerPC section.
8874 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8875 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8876 help text that is too long.
8877 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8878 help text that is too long.
8879 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8880 help text that is too long.
8882 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8884 * config/alpha/alpha.md (stack_probe_internal): Rename
8885 from "probe_stack". Update all callers.
8887 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8889 PR rtl-optimization/84566
8890 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8891 sched_macro_fuse_insns.
8893 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8896 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8897 (compute_block_dependences): ... from here.
8899 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8901 PR tree-optimization/85331
8902 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8903 from int to HOST_WIDE_INT.
8905 2018-04-11 Martin Jambor <mjambor@suse.cz>
8908 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8909 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8910 not the same as the source val.
8911 (cgraph_edge_brings_value_p): New parameter.
8912 (gather_edges_for_value): Pass destination value to
8913 cgraph_edge_brings_value_p.
8914 (perhaps_add_new_callers): Likewise.
8915 (get_info_about_necessary_edges): Likewise and exclude values brought
8916 only by self-recursive edges.
8917 (create_specialized_node): Redirect only clones of self-calling edges.
8918 (+self_recursive_pass_through_p): New function.
8919 (find_more_scalar_values_for_callers_subset): Use it.
8920 (find_aggregate_values_for_callers_subset): Likewise.
8921 (known_aggs_to_agg_replacement_list): Removed.
8922 (decide_whether_version_node): Re-calculate known constants for all
8923 remaining context clones.
8925 2018-04-11 Richard Biener <rguenther@suse.de>
8928 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8929 from early DWARF output.
8930 (dwarf2out_early_finish): Output line info unconditionally into
8931 early DWARF and add reference to it.
8933 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8936 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8937 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8938 other than V2DFmode using iptr mode attribute.
8939 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8941 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8943 PR rtl-optimization/84659
8944 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8946 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8949 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8951 (output_loc_list): Pass address of a dummy size variable even in the
8952 locview handling loop.
8953 (index_location_lists): Add comment on why skip_loc_list_entry can't
8956 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8959 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8962 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8965 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8966 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8967 and -mstring-compare-inline-limit.
8969 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8972 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8973 for stack clash protection in a register whenever we need it to be in
8976 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8978 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8979 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8981 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8984 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8986 (mlong-double-): Ditto.
8987 * config/rs6000/sysv4.opt (msdata=): Ditto.
8988 (mtls-size=): Ditto.
8990 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8992 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8993 erroneous entries for
8994 "vector int vec_ldl (int, long int *)", and
8995 "vector unsigned int vec_ldl (int, unsigned long int *)".
8996 Add comments and entries for
8997 "vector bool char vec_ldl (int, bool char *)",
8998 "vector bool short vec_ldl (int, bool short *)",
8999 "vector bool int vec_ldl (int, bool int *)",
9000 "vector bool long long vec_ldl (int, bool long long *)",
9001 "vector pixel vec_ldl (int, pixel *)",
9002 "vector long long vec_ldl (int, long long *)",
9003 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9004 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9005 type tree bool_long_long_type_node and correct definition of
9006 bool_V2DI_type_node to make reference to this new type tree.
9007 (rs6000_mangle_type): Replace erroneous reference to
9008 bool_long_type_node with bool_long_long_type_node.
9009 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9010 comments to emphasize sign distinctions for char and int types and
9011 replace RS6000_BTI_bool_long constant with
9012 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9013 use of RS6000_BTI_pixel.
9014 (bool_long_type_node): Remove this macro definition.
9015 (bool_long_long_type_node): New macro definition
9017 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9019 PR rtl-optimization/85300
9020 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9021 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9022 simplify_unary_operation fails.
9024 2018-04-10 Martin Liska <mliska@suse.cz>
9026 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9027 cgraph_edge and ipa_ref.
9029 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9033 * config/i386/sse.md
9034 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9035 computation of the VEC_MERGE selector from mask.
9036 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9037 Fix decoding of the VEC_MERGE selector into mask.
9039 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9041 PR tree-optimization/85286
9042 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9044 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9046 * final.c (final_1): Set insn_last_address as well as
9047 insn_current_address.
9049 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9052 * explow.c (emit_stack_probe): Call validize_mem on memory location
9053 before passing it to gen_probe_stack. Create address operand and
9054 legitimize it for the probe_stack_address case.
9056 2018-04-09 Jan Hubicka <jh@suse.cz>
9059 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9060 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9061 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9063 (free_lang_data): Rebuild type inheritance graph.
9065 2018-04-09 Martin Sebor <msebor@redhat.com>
9067 * invoke.texi (-finline-small-functions): Mention other optimization
9069 (-findirect-inlining, -fpartial-inlining): Same.
9070 (-finline-functions-called-once): Same.
9071 (-freorder-blocks-and-partition): Same.
9073 2018-04-09 Jan Hubicka <jh@suse.cz>
9076 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9077 jumps; choose last target that matches the criteria (i.e.
9078 no partition changes for non-crossing jumps).
9079 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9080 support for redirecting crossing jumps to non-crossing.
9082 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9084 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9085 also for naked functions.
9087 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9089 * config/arc/arc.md (add_shift): New pattern.
9090 (add_shift2): Likewise.
9091 (sub_shift): Likewise.
9092 (sub_shift_cmp0_noout): Likewise.
9093 (compare_si_ashiftsi): Likewise.
9094 (xbfu_cmp0_noout): New combine pattern.
9095 (xbfu_cmp0"): Likewise.
9096 (movsi_set_cc_insn): Place the predicable variant first.
9097 (commutative_binary_cmp0_noout): Remove clobber.
9098 (commutative_binary_cmp0): New pattern.
9099 (noncommutative_binary_cmp0): Likewise.
9100 (noncommutative_binary_cmp0_noout): Likewise.
9101 (noncommutative_binary_comparison_result_used): Removed.
9102 (rsub_cmp0): New pattern.
9103 (rsub_cmp0_noout): Likewise.
9104 (extzvsi): Changed, keep only meaningful variants.
9105 (SQH, SEZ): New iterators.
9106 (SQH_postfix): New mode attribute.
9107 (SEZ_prefix): New code attribute.
9108 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9109 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9110 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9112 (noncommutative_operator): Check the availability of barrel
9115 2018-04-09 Richard Biener <rguenther@suse.de>
9117 PR tree-optimization/85284
9118 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9119 Only use the niter constraining form of simple_iv when the exit
9122 2018-04-09 Tom de Vries <tom@codesourcery.com>
9125 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9126 (define_expand "*memory_barrier"): New define_expand.
9127 (define_insn "memory_barrier"): New insn.
9129 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9131 PR rtl-optimization/80463
9132 PR rtl-optimization/83972
9133 PR rtl-optimization/83480
9135 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9136 correct producer for the insn.
9137 (tidy_control_flow): Fixup seqnos in case of debug insns.
9139 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9141 PR rtl-optimization/83913
9143 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9144 different sched-times when merging exprs.
9146 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9148 PR rtl-optimization/83962
9150 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9151 tidy_fallthru_edge and tidy_control_flow.
9153 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9155 PR rtl-optimization/83530
9157 * sel-sched.c (force_next_insn): New global variable.
9158 (remove_insn_for_debug): When force_next_insn is true, also leave only
9159 next insn in the ready list.
9160 (sel_sched_region): When the region wasn't scheduled, make another pass
9161 over it with force_next_insn set to 1.
9163 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9165 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9167 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9168 for interrupt control.
9169 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9170 functions for interrupt control.
9171 * config/nds32/nds32-intrinsic.md: Likewise.
9172 * config/nds32/nds32_intrinsic.h: Likewise.
9173 * config/nds32/nds32.h (nds32_builtins): Likewise.
9175 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9177 * config/nds32/nds32.c (nds32_init_machine_status,
9178 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9179 strict_aligned_p field.
9180 (nds32_expand_to_rtl_hook): New function.
9181 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9182 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9184 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9185 Chung-Ju Wu <jasonwucj@gmail.com>
9187 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9188 * config/nds32/nds32-n7.md: New file.
9189 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9190 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9192 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9193 * config/nds32/nds32.md (pipeline_model): Add n7.
9194 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9195 * config/nds32/pipelines.md: Include n7 settings.
9197 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9198 Chung-Ju Wu <jasonwucj@gmail.com>
9200 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9201 * config/nds32/nds32-e8.md: New file.
9202 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9203 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9205 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9206 * config/nds32/nds32.md (pipeline_model): Add e8.
9207 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9208 * config/nds32/pipelines.md: Include e8 settings.
9210 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9211 Chung-Ju Wu <jasonwucj@gmail.com>
9213 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9214 * config/nds32/nds32-n8.md: New file.
9215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9218 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9219 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9220 * config/nds32/nds32.md (pipeline_model): Add n8.
9221 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9222 * config/nds32/pipelines.md: Include n8 settings.
9224 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9225 Chung-Ju Wu <jasonwucj@gmail.com>
9227 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9228 * config/nds32/nds32-n9-2r1w.md: New file.
9229 * config/nds32/nds32-n9-3r2w.md: New file.
9230 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9231 nds32_register_ports): New or modify for cpu n9.
9232 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9234 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9235 * config/nds32/nds32-utils.c: New file.
9236 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9237 TARGET_MUL_SLOW): Define.
9238 * config/nds32/nds32.md (pipeline_model): New attribute.
9239 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9240 New options that support cpu n9.
9241 * config/nds32/pipelines.md: Include n9 settings.
9242 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9244 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9246 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9247 information if necessary.
9248 (output_cond_branch_compare_zero): Likewise.
9249 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9250 (nds32_target_alignment): Refine for alignment.
9251 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9252 (FUNCTION_BOUNDARY): Modify.
9253 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9255 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9257 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9259 * config/nds32/constants.md (unspec_volatile_element): Add values for
9260 TLB operation and data prefetch.
9261 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9262 functions for TLB operation and data prefetch.
9263 * config/nds32/nds32-intrinsic.md: Likewise.
9264 * config/nds32/nds32_intrinsic.h: Likewise.
9265 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9266 (nds32_print_operand): Likewise.
9267 * config/nds32/nds32.h (nds32_builtins): Likewise.
9269 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9270 Andrew Pinski <pinsika@gcc.gnu.org>
9273 * match.pd: Use constant_boolean_node of correct type instead of
9274 boolean_true_node or boolean_false_node for simplifying
9275 pointer comparisons to zero.
9277 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9279 PR tree-optimization/80021
9280 * tree.c (verify_type_variant): Make error call in verify_variant_match
9281 translatable and remove final full stop.
9283 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9285 * config/nds32/constants.md (unspec_volatile_element): Add
9286 UNSPEC_VOLATILE_EH_RETURN.
9287 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9288 nds32_output_stack_pop): Support dwarf exception handling process.
9289 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9290 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9291 exception handling process.
9292 (nds32_compute_stack_frame): Likewise.
9293 (nds32_return_addr_rtx): Likewise.
9294 (nds32_initial_elimination_offset): Likewise.
9295 (nds32_expand_prologue): Likewise.
9296 (nds32_expand_epilogue): Likewise.
9297 (nds32_dynamic_chain_address): New function.
9298 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9300 (DYNAMIC_CHAIN_ADDRESS): Define.
9301 (EH_RETURN_DATA_REGNO): Define.
9302 (EH_RETURN_STACKADJ_RTX): Define.
9303 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9304 patterns for dwarf exception handling.
9306 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9308 * config/nds32/nds32.h: Clean up obsolete macros.
9310 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9312 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9313 Add enum values for particular instructions.
9314 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9315 particular intrinsic functions.
9316 * config/nds32/nds32-intrinsic.md: Likewise.
9317 * config/nds32/nds32_intrinsic.h: Likewise.
9318 * config/nds32/nds32.h (nds32_builtins): Likewise.
9319 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9320 (btst, ave): New patterns for particular instructions.
9322 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9324 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9325 Add enum values for atomic load/store and memory sync.
9326 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9328 * config/nds32/nds32-intrinsic.md: Likewise.
9329 * config/nds32/nds32_intrinsic.h: Likewise.
9330 * config/nds32/nds32.h (nds32_builtins): Likewise.
9332 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9334 PR tree-optimization/85257
9335 * fold-const.c (native_encode_vector): If not all elts could fit
9336 and off is -1, return 0 rather than offset.
9337 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9338 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9339 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9340 adjust buffer in native_interpret_expr call.
9342 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9344 * config/nds32/constants.md (unspec_volatile_element): Add cache
9345 control enum values.
9346 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9347 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9348 * config/nds32/nds32.c (nds32_cctl_names): New.
9349 (nds32_print_operand): Handle cache control register names.
9350 * config/nds32/nds32.h (nds32_builtins): New enum values.
9351 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9353 * config/nds32/nds32.md (type): Add mmu.
9354 * config/nds32/pipelines.md (simple_insn): Add mmu.
9356 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9358 * config/nds32/nds32.md (type): Remove call.
9359 * config/nds32/pipelines.md (simple_insn): Likewise.
9361 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9363 * config/nds32/constants.md (unspec_volatile_element): Add
9364 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9365 UNSPEC_VOLATILE_FMFCFG.
9366 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9367 description for fmfcfg and fmfcsr.
9368 (bdesc_1arg): Add fmtcsr.
9369 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9370 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9371 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9372 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9373 unspec_fmfcfg): New patterns.
9374 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9375 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9376 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9377 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9378 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9379 __nds32__fmfcfg): Define.
9381 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9383 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9384 intrinsic register names.
9385 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9386 intrinsic register enum values and macros.
9388 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9390 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9391 for load/store addressing form.
9392 (nds32_print_operand_address): Likewise.
9394 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9397 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9398 based on LABEL_REF. Remove useless assertion.
9399 (pic_address_needs_scratch): Fix formatting.
9400 (sparc_legitimize_pic_address): Minor tweaks.
9401 (sparc_delegitimize_address): Adjust assertion accordingly.
9402 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9403 into symbolic_operand.
9404 (movsi_high_pic_label_ref): Likewise.
9405 (movsi_lo_sum_pic_label_ref): Likewise.
9406 (movdi_pic_label_ref): Likewise.
9407 (movdi_high_pic_label_ref): Likewise.
9408 (movdi_lo_sum_pic_label_ref): Likewise.
9410 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9412 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9413 custom LIB_SPEC setup.
9415 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9416 Kito Cheng <kito.cheng@gmail.com>
9418 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9419 * config/riscv/freebsd.h: New.
9421 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9423 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9424 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9427 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9428 Kito Cheng <kito.cheng@gmail.com>
9430 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9431 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9432 * config/nds32/nds32-protos.h (nds32_output_call,
9433 nds32_output_return): Declare.
9434 * config/nds32/nds32.md: Refine all the call and return patterns.
9436 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9439 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9440 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9442 PR rtl-optimization/84872
9443 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9444 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9447 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9449 * expr.c (copy_blkmode_to_reg): Revert 254862.
9450 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9452 2018-04-06 Richard Biener <rguenther@suse.de>
9455 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9456 after seeing a component reference with an adjacent field. Treat
9457 refs to arrays at struct end of external decls similar to
9458 refs to unconstrained commons.
9460 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9463 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9464 look through SAVE_EXPRs with non-side-effects argument. Adjust
9466 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9469 2018-04-06 Richard Biener <rguenther@suse.de>
9472 * alias.c (find_base_term): New wrapper around find_base_term
9473 unwinding CSELIB_VAL_PTR changes.
9474 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9477 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9479 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9481 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9482 constant definitions.
9483 ("nop"): lr 0,0 -> nopr r0
9484 ("nop_lr0", "nop_lr1"): New insn definitions.
9486 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9488 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9489 NDS32_V3PUSH_AVAILABLE_P macro.
9491 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9492 Chung-Ju Wu <jasonwucj@gmail.com>
9494 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9495 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9496 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9497 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9498 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9499 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9500 * config/nds32/constraints.md: New constraints and checking for hard
9501 float configuration.
9502 * config/nds32/iterators.md: New mode iterator and attribute for hard
9503 float configuration.
9504 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9506 * config/nds32/nds32-fpu.md: New file.
9507 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9508 deal with hard float code generation.
9509 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9511 (abi_type, float_reg_number): New enum type.
9512 * config/nds32/nds32-predicates.c: New predicates for hard float.
9513 * config/nds32/nds32-protos.h: Declare functions for hard float.
9514 * config/nds32/nds32.c: Implementation for hard float configuration.
9515 * config/nds32/nds32.h: Definitions for hard float configuration.
9516 * config/nds32/nds32.md: Include hard float machine description and
9517 modify patterns for hard float configuration.
9518 * config/nds32/nds32.opt: New options for hard float configuration.
9519 * config/nds32/predicates.md: New predicates for hard float
9522 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9524 * common/config/nds32/nds32-common.c
9525 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9527 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9530 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9531 CONSTRUCTOR_ELT (ctor, ...)->value.
9533 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9536 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9538 2018-04-05 Tom de Vries <tom@codesourcery.com>
9541 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9544 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9545 Kito Cheng <kito.cheng@gmail.com>
9547 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9548 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9549 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9550 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9552 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9553 Kito Cheng <kito.cheng@gmail.com>
9555 * config/nds32/constraints.md (Ufe): New memory constraint.
9556 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9557 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9558 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9560 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9561 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9563 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9565 * config/nds32/nds32.md: Use optimize_size in the condition for
9566 alu-shift instructions.
9568 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9570 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9572 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9574 * config/nds32/nds32.md (negsi2): Refine pattern.
9576 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9577 Chung-Ju Wu <jasonwucj@gmail.com>
9579 * config/nds32/iterators.md (shift_rotate): New code iterator.
9580 (shift): New code attribute.
9581 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9582 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9583 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9584 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9585 bit-wise operations.
9586 (andsi3, *andsi3): Ditto.
9587 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9588 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9589 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9590 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9591 nds32_ior_operand, nds32_xor_operand): New predicates.
9593 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9595 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9596 (addsi3, subsi3): ... this.
9598 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9600 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9602 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9604 * config/nds32/nds32.md: Adjust indention.
9606 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9608 * config/nds32/nds32.md (feature): New attribute.
9610 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9612 * config/nds32/nds32.md (subtype): New attribute.
9614 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9617 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9618 expansion to perform a bitwise AND of the argument followed by a
9619 boolean negation of the result.
9621 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9623 PR rtl-optimization/84878
9624 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9625 the basic block. Assert the use reference is not artificial and that
9626 it has an associated insn.
9628 2018-04-04 Michael Matz <matz@suse.de>
9630 * builtins.c (compute_objsize): Pass correct operand
9631 to array_at_struct_end_p.
9633 2018-04-04 Richard Biener <rguenther@suse.de>
9636 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9637 from contexts for DINFO_LEVEL_TERSE and below.
9639 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9641 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9642 resiter_operand condition.
9643 * config/nds32/nds32.md (*move<mode>): Ditto.
9645 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9646 Monk Chiang <sh.chiang04@gmail.com>
9648 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9650 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9652 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9654 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9655 Kito Cheng <kito.cheng@gmail.com>
9657 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9658 nds32_cond_code_str, output_cond_branch,
9659 output_cond_branch_compare_zero, nds32_expand_cbranch,
9660 nds32_expand_cstore, nds32_expand_movcc,
9661 nds32_output_cbranchsi4_equality_zero,
9662 nds32_output_cbranchsi4_equality_reg,
9663 nds32_output_cbranchsi4_equality_reg_or_const_int,
9664 nds32_output_cbranchsi4_greater_less_zero: New functions.
9665 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9666 nds32_expand_cstore, nds32_expand_movcc,
9667 nds32_output_cbranchsi4_equality_zero,
9668 nds32_output_cbranchsi4_equality_reg,
9669 nds32_output_cbranchsi4_equality_reg_or_const_int,
9670 nds32_output_cbranchsi4_greater_less_zero): Declare.
9671 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9672 nds32_rimm11s_operand): New predicates.
9673 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9674 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9677 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9679 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9680 * config/nds32/nds32.md: Ditto.
9681 * config/nds32/pipelines.md: Ditto.
9683 2018-04-04 Richard Biener <rguenther@suse.de>
9685 PR tree-optimization/85168
9686 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9687 propagating abnormals.
9689 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9691 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9693 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9694 Kito Cheng <kito.cheng@gmail.com>
9696 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9697 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9698 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9699 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9700 * config/nds32/nds32.md (sibcall_internal): New.
9701 (sibcall_register): Remove.
9702 (sibcall_immediate): Remove.
9703 (sibcall_value_internal): New.
9704 (sibcall_value_register): Remove.
9705 (sibcall_value_immediate): Remove.
9706 * config/nds32/predicates.md (nds32_general_register_operand): New.
9707 (nds32_call_address_operand): New.
9709 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9711 PR rtl-optimization/85167
9712 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9713 bb_defs if *split_p, instead preinitialize it to NULL.
9715 PR tree-optimization/85156
9716 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9717 evaluating the argument multiple times.
9719 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9721 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9723 (_mm_cvtpd_ps): Likewise.
9724 (_mm_cvttpd_epi32): Likewise.
9725 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9726 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9727 vector, pixel, and bool following altivec.h include.
9729 2018-04-03 Martin Sebor <msebor@redhat.com>
9731 * doc/extend.texi (Common Function Attributes): Clarify.
9732 (const attribute): Likewise.
9733 (pure attribute): Likewise.
9735 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9738 * config/i386/i386.c (ix86_expand_vector_set): Use
9739 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9741 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9743 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9744 instructions when changing rounding bits to preserve precision bits
9745 in the x87 control word.
9747 2018-04-03 Martin Liska <mliska@suse.cz>
9749 PR tree-optimization/82491
9750 * rtl.h (strip_offset_and_add): Replace += suboffset with
9751 poly_uint64 () + suboffset.
9753 2018-03-29 Martin Liska <mliska@suse.cz>
9754 Martin Jambor <mjambor@suse.cz>
9757 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9758 param_type is not an integral or pointer type.
9760 2018-04-03 Richard Biener <rguenther@suse.de>
9762 * sese.h (recompute_all_dominators): Remove.
9764 2018-04-02 Martin Sebor <msebor@redhat.com>
9766 * doc/invoke.texi (-Wrestrict): Fix typos.
9768 2018-04-02 Jim Wilson <jimw@sifive.com>
9770 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9771 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9772 (<optab>di3, <optab>si3_extend): Likewise.
9773 (<optab>si3_mask, <optab>si3_mask_1): New.
9774 (<optab>di3_mask, <optab>di3_mask_1): New.
9775 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9776 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9777 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9779 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9781 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9784 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9786 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9787 (nds32_canonicalize_comparison): New function.
9789 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9790 Kito Cheng <kito.cheng@gmail.com>
9791 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9793 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9794 * config/nds32/constants.md (unspec_volatile_element): Add
9795 UNSPEC_VOLATILE_RELAX_GROUP.
9796 * config/nds32/nds32-relax-opt.c: New file.
9797 * config/nds32/nds32-predicates.c
9798 (nds32_symbol_load_store_p): New function.
9799 * config/nds32/nds32-protos.h
9800 (nds32_symbol_load_store_p): Declare function.
9801 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9802 * config/nds32/nds32.c
9803 (nds32_register_pass): New function to register pass.
9804 (nds32_register_passes): New function to register passes.
9805 * config/nds32/nds32.md (relax_group): New pattern.
9806 * config/nds32/nds32.opt (mrelax-hint): New option.
9807 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9809 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9811 * config/nds32/t-nds32: Modify files dependency.
9813 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9815 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9816 (PROFILE_HOOK): Define its implementation.
9818 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9820 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9821 type and 32-bit size.
9823 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9826 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9827 (V_128_256): New mode iterator.
9828 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9829 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9830 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9832 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9833 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9835 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9838 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9839 NaN inputs correctly.
9841 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9844 * config/rs6000/vsx.md (??r): New mode attribute.
9845 (*vsx_mov<mode>_64bit): Use it.
9846 (*vsx_mov<mode>_32bit): Likewise.
9848 2018-03-30 Martin Sebor <msebor@redhat.com>
9850 PR tree-optimization/84818
9851 * builtins.c (check_access): Use warning_n.
9853 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9856 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9858 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9861 2018-03-30 Julia Koval <julia.koval@intel.com>
9864 * x86-tune.def (movx, partial_reg_dependency): Enable for
9867 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9870 * lra-constraints.c (process_alt_operands): Move setting
9871 this_alternative_matches below.
9873 2018-03-29 Martin Liska <mliska@suse.cz>
9876 * doc/invoke.texi: Document how LTO works with debug info.
9877 Describe auto-load support of binutils. Mention 'x86-64'
9878 as valid option value of -march option.
9880 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9882 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9885 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9886 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9887 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9890 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9893 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9894 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9895 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9896 for RS6000_BTM_POWERPC64.
9897 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9898 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9899 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9904 2018-03-28 Carl Love <cel@us.ibm.com>
9907 2017-09-27 Carl Love <cel@us.ibm.com>
9909 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9910 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9911 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9914 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9916 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9917 instead of __vector bool.
9918 (_mm_max_pu8): Likewise.
9919 (_mm_min_pi16): Likewise.
9921 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9924 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9925 (DIVWEUO): Likewise.
9927 (DIVDEUO): Likewise.
9928 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9929 DIVWEUO and DIVDEUO.
9930 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9931 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9932 (div_extend): Likewise.
9933 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9935 (__builtin_divweuo): Likewise.
9936 (__builtin_divdeo): Likewise.
9937 (__builtin_divdeuo): Likewise.
9939 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9942 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9943 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9945 PR tree-optimization/82004
9946 * gimple-match-head.c (optimize_pow_to_exp): New function.
9947 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9948 Don't fold to exp if optimize_pow_to_exp is false.
9950 2018-03-28 Martin Liska <mliska@suse.cz>
9953 * calls.c (initialize_argument_information): Fix trailing space.
9954 * common.opt: Fix typo and provide better explanation for
9955 -fsanitize-coverage option.
9956 * config/i386/i386.opt: Fix typo.
9958 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9959 Martin Liska <mliska@suse.cz>
9962 * gimplify.c (asan_poison_variable): Don't do the check for
9963 gimplify_omp_ctxp here.
9964 (gimplify_decl_expr): Do it here.
9965 (gimplify_target_expr): Likewise.
9967 2018-03-28 Martin Liska <mliska@suse.cz>
9970 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9971 chkp_type_bounds_count if MPX is not enabled.
9973 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9975 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9977 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9980 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9981 function to create the function decl for complex long double
9982 multiply and divide for -mabi=ieeelongdouble.
9983 (init_float128_ieee): Call it.
9985 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9988 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9989 -fcf-protection=branch -mibt.
9990 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9992 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9995 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9996 arm_word_relocations.
9998 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10001 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10002 extern array declarations.
10004 2018-03-27 Richard Biener <rguenther@suse.de>
10006 PR middle-end/84067
10007 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10008 explicit single_use checks.
10010 2018-03-27 Richard Biener <rguenther@suse.de>
10012 PR tree-optimization/85082
10013 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10016 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10018 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10019 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10020 Turn on fasynchronous-unwind-tables and funwind-tables.
10022 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10025 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10026 (*bmi_blsr_<mode>_ccz): Ditto.
10028 2018-03-26 Tom de Vries <tom@codesourcery.com>
10030 PR tree-optimization/85063
10031 * omp-general.c (offloading_function_p): New function. Factor out
10033 * omp-offload.c (pass_omp_target_link::gate): ... here.
10034 * omp-general.h (offloading_function_p): Declare.
10035 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10036 with attribute omp declare target for offloading functions.
10038 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10040 PR tree-optimization/84005
10041 * tree-data-ref.h (get_base_for_alignment): Declare.
10042 * tree-data-ref.c (get_base_for_alignment_1): New function.
10043 (get_base_for_alignment): Likewise.
10044 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10045 get_base_for_alignment to find a suitable base object, instead
10046 of always using drb->base_address.
10048 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10050 PR inline-asm/85022
10051 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10052 known size by default.
10054 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10056 PR inline-asm/85030
10057 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10058 and non BLKmode operands.
10060 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10063 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10064 Clean up attributes.
10066 2018-03-23 Richard Biener <rguenther@suse.de>
10069 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10070 we are going to emit early debug for LTO.
10072 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10074 PR inline-asm/85034
10075 * function.c (match_asm_constraints_1): Don't optimize if input
10076 doesn't satisfy general_operand predicate for output's mode.
10078 PR inline-asm/85022
10079 * alias.c (write_dependence_p): Don't require for x_canonicalized
10080 non-VOIDmode if x has VOIDmode.
10083 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10084 just don't try to optimize it rather than assert it never happens.
10086 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10088 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10089 macro expansions for definition of ST_INTERNAL_<mode> and
10090 LD_INTERNAL_<mode> builtins.
10091 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10093 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10095 (altivec_expand_st_builtin): Likewise.
10096 (altivec_expand_builtin): Remove calls to deleted functions.
10097 (rs6000_address_for_altivec): Delete this function.
10098 * config/rs6000/vector.md: Remove expands for
10099 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10101 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10104 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10105 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10106 re-computing once computed.
10107 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10108 (arm_init_machine_status): Initialize
10109 machine->static_chain_stack_bytes.
10111 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10114 * doc/extend.texi: Add four new prototypes for vec_ld.
10115 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10116 definitions for more logical presentation.
10117 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10118 entries for V1TI variants of __builtin_altivec_ld builtin.
10119 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10120 handling of V1TI variant of LVX icode pattern.
10121 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10122 (rs6000_gimple_fold_builtin): Likewise.
10123 (altivec_init_builtins): Add code to define
10124 __builtin_altivec_lvx_v1ti function.
10126 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10128 PR inline-asm/84941
10129 * function.c (match_asm_constraints_1): Don't do the optimization
10130 if input isn't a REG, SUBREG, MEM or constant.
10132 2018-03-22 Tom de Vries <tom@codesourcery.com>
10134 PR tree-optimization/84956
10135 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10136 bb_has_abnormal_pred.
10138 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10141 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10142 DECL_INITIAL (decl) to decl at the end.
10143 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10144 adjust the comment.
10146 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10148 * doc/extend.texi (__builtin_tgmath): Document when complex
10149 integer types are treated as _Complex _Float64.
10151 2018-03-21 Tom de Vries <tom@codesourcery.com>
10153 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10155 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10157 PR tree-optimization/84960
10158 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10159 if it is ENTRY block, move them into single succ of ENTRY in that case.
10161 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10163 PR tree-optimization/84811
10164 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10165 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10166 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10167 (known_subrange_p): Update accordingly. Cast each value involved
10168 in the size comparison, rather than casting the result of the
10171 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10173 PR tree-optimization/84982
10174 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10175 by flipping the least significant bit rather than all bits from
10176 bitpos to bitpos + bitsize - 1.
10178 2018-03-21 Nathan Sidwell <nathan@acm.org>
10180 * doc/extend.texi (Deprecated Features): Remove mention of
10181 long-deleted deprecations.
10183 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10186 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10187 * configure: Regenerate.
10189 2018-03-21 Tom de Vries <tom@codesourcery.com>
10191 PR tree-optimization/83126
10192 * tree-parloops.c (num_phis): New function.
10193 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10195 2018-03-21 Nathan Sidwell <nathan@acm.org>
10197 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10198 mention anon-struct/union members and trailing attributes.
10200 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10202 PR tree-optimization/84969
10203 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10204 builtin memset partitions if they set different rhs values.
10206 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10208 PR rtl-optimization/84989
10209 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10210 VEC_DUPLICATE with scalar result mode.
10212 2018-03-21 Martin Liska <mliska@suse.cz>
10215 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10216 not intended return statement.
10218 2018-03-21 Martin Liska <mliska@suse.cz>
10221 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10222 (chkp_find_bound_slots_1): Limit number of iterations.
10224 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10227 * Minor grammar fixes for x86 options.
10229 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10232 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10233 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10235 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10238 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10239 (altivec_lvx_<mode>_1op): Likewise.
10240 (altivec_stvx_<mode>_2op): Likewise.
10241 (altivec_stvx_<mode>_1op): Likewise.
10242 (altivec_lvx_<VM2:mode>): New define_expand.
10243 (altivec_stvx_<VM2:mode>): Likewise.
10244 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10245 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10246 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10247 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10248 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10249 (rs6000_gen_lvx): Likewise.
10250 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10251 (altivec_expand_stv_builtin): Likewise.
10252 (altivec_expand_builtin): Likewise.
10253 * config/rs6000/vector.md: Likewise.
10255 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10258 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10261 2018-03-20 Richard Biener <rguenther@suse.de>
10264 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10265 sign-conversions as zero, fall back to standard scalar_stmt
10268 2018-03-20 Martin Liska <mliska@suse.cz>
10271 * predict.c (rebuild_frequencies): Handle case when we have
10272 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10274 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10277 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10278 flag_section_anchors.
10279 * varasm.c (use_blocks_for_decl_p): Remove hack for
10280 dw2_force_const_mem.
10283 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10285 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10286 be created, use lowpart_subreg of operands[0] rather than operands[0]
10288 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10289 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10290 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10291 and n constraint instead of aarch64_shift_imm_di and Usd.
10292 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10293 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10295 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10298 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10299 to favor GPR over NEON registers.
10300 (<shift>di3_neon): Likewise.
10302 2018-03-20 Tom de Vries <tom@codesourcery.com>
10305 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10306 (nvptx_process_pars): Emit bar.sync asap and alap.
10308 2018-03-20 Tom de Vries <tom@codesourcery.com>
10311 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10312 seen_label if seen_label is already set.
10314 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10317 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10318 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10319 Use 1U instead of 1. Formatting fixes.
10322 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10323 instead of TREE_TYPE (s1) for the return value.
10325 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10327 PR tree-optimization/84946
10328 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10329 bitsize + bitsize in poly_uint64 rather than poly_int64.
10332 * dwarf2asm.c: Include fold-const.c.
10333 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10334 of decl rather than decl itself.
10336 PR rtl-optimization/84643
10337 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10339 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10342 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10343 calling assemble_variable.
10345 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10348 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10349 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10351 2018-03-19 Jim Wilson <jimw@sifive.com>
10354 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10355 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10356 (riscv_first_stack_step): Likewise.
10357 (riscv_option_override): Use STACK_BOUNDARY instead of
10358 MIN_STACK_BOUNDARY.
10359 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10360 MIN_STACK_BOUNDARY.
10361 (BIGGEST_ALIGNMENT): Set to 128.
10362 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10363 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10366 2018-03-19 Richard Biener <rguenther@suse.de>
10368 PR tree-optimization/84933
10369 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10370 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10372 2018-03-19 Richard Biener <rguenther@suse.de>
10374 PR tree-optimization/84859
10375 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10376 (cond_if_else_store_replacement): Perform sinking operation on
10377 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10378 Generalize what a BB with a single eligible store is.
10380 2018-03-19 Richard Biener <rguenther@suse.de>
10382 PR tree-optimization/84929
10383 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10384 chrec_is_positive against non-chrec arg.
10386 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10389 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10391 2018-03-18 Martin Liska <mliska@suse.cz>
10393 PR rtl-optimization/84635
10394 * regrename.c (build_def_use): Use matches_mode only when
10397 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10399 PR tree-optimization/84913
10400 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10401 vectorize chains of COND_EXPRs.
10403 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10405 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10407 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10409 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10411 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10413 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10415 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10416 Kito Cheng <kito.cheng@gmail.com>
10418 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10419 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10420 (nds32_adjust_reg_alloc_order): New function.
10421 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10423 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10425 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10426 nds32_print_operand, nds32_print_operand_address): Use
10427 HOST_WIDE_INT_PRINT_DEC instead.
10429 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10431 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10433 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10436 * config/i386/i386.c (initial_ix86_tune_features,
10437 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10438 unsigned long long.
10439 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10440 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10441 rather than 1u << ix86_tune. Formatting fix.
10442 (ix86_option_override_internal): Change ix86_arch_mask from
10443 unsigned int to unsigned HOST_WIDE_INT, initialize to
10444 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10445 (ix86_function_specific_restore): Likewise.
10447 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10450 * postreload.c (reload_combine_recognize_pattern): Perform
10451 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10452 truncate_int_for_mode the result for the destination's mode.
10455 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10456 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10458 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10461 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10464 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10465 regno_allocno_class_array and sorted_pseudos.
10466 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10467 insns where regno is used.
10469 2018-03-16 Martin Liska <mliska@suse.cz>
10472 * multiple_target.c (create_dispatcher_calls): Redirect
10473 reference in the symbol table.
10475 2018-03-16 Martin Liska <mliska@suse.cz>
10478 * multiple_target.c (create_dispatcher_calls): Redirect also
10481 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10485 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10486 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10487 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10489 2018-03-16 Julia Koval <julia.koval@intel.com>
10491 * doc/invoke.texi (Skylake Server): Add CLWB.
10492 Cannonlake): Remove CLWB.
10494 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10496 PR tree-optimization/84841
10497 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10499 (FLOAT_ONE_CONST_TYPE): Define.
10500 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10501 (sort_by_operand_rank): Put entries with higher constant_type last
10502 rather than first to match comments.
10504 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10506 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10509 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10512 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10513 check and use address of target always.
10515 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10518 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10519 (indirect_thunk_bnd_needed): Likewise.
10520 (indirect_thunks_used): Likewise.
10521 (indirect_thunks_bnd_used): Likewise.
10522 (indirect_return_needed): New.
10523 (indirect_return_bnd_needed): Likewise.
10524 (output_indirect_thunk_function): Add a bool argument for
10526 (output_indirect_thunk_function): Don't generate alias for
10527 function return thunk.
10528 (ix86_code_end): Call output_indirect_thunk_function to generate
10529 function return thunks.
10530 (ix86_output_function_return): Set indirect_return_bnd_needed
10531 and indirect_return_needed instead of indirect_thunk_bnd_needed
10532 and indirect_thunk_needed.
10534 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10536 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10537 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10538 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10540 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10541 Paul Hua <paul.hua.gm@gmail.com>
10544 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10546 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10548 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10549 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10550 resp. SFmode cases.
10552 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10555 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10556 instead of GET_MODE_SIZE when comparing Units.
10558 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10561 * varasm.c (hash_section): Return an unchangeble hash value
10562 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10563 Return !aarch64_can_use_per_function_literal_pools_p ().
10565 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10568 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10569 rather than address of cmode as last argument to prepare_cmp_insn.
10571 2018-03-15 Julia Koval <julia.koval@intel.com>
10573 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10574 F_AVX512VNNI, F_AVX512BITALG): New.
10576 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10579 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10580 insn for floating-point loads and stores.
10582 2018-03-14 Carl Love <cel@us.ibm.com>
10584 * config/rs6000/rs6000-c.c: Add macro definitions for
10585 ALTIVEC_BUILTIN_VEC_PERMXOR.
10586 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10587 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10588 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10589 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10591 * config/doc/extend.texi: Add prototypes for vec_permxor.
10593 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10596 * diagnostic-show-locus.c (class layout_point): Convert m_line
10597 from int to linenum_type.
10598 (line_span::comparator): Use linenum "compare" function when
10599 comparing line numbers.
10600 (test_line_span): New function.
10601 (layout_range::contains_point): Convert param "row" from int to
10603 (layout_range::intersects_line_p): Likewise.
10604 (layout::will_show_line_p): Likewise.
10605 (layout::print_source_line): Likewise.
10606 (layout::should_print_annotation_line_p): Likewise.
10607 (layout::print_annotation_line): Likewise.
10608 (layout::print_leading_fixits): Likewise.
10609 (layout::annotation_line_showed_range_p): Likewise.
10610 (struct line_corrections): Likewise for field m_row.
10611 (line_corrections::line_corrections): Likewise for param "row".
10612 (layout::print_trailing_fixits): Likewise.
10613 (layout::get_state_at_point): Likewise.
10614 (layout::get_x_bound_for_row): Likewise.
10615 (layout::print_line): Likewise.
10616 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10617 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10618 * input.c (selftest::test_linenum_comparisons): New function.
10619 (selftest::input_c_tests): Call it.
10620 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10621 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10622 * selftest.h (ASSERT_GT): New macro.
10623 (ASSERT_GT_AT): New macro.
10624 (ASSERT_LT): New macro.
10625 (ASSERT_LT_AT): New macro.
10627 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10629 PR rtl-optimization/84780
10630 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10632 2018-03-14 Martin Liska <mliska@suse.cz>
10634 * tree.c (record_node_allocation_statistics): Use
10635 get_stats_node_kind.
10636 (get_stats_node_kind): New function extracted from
10637 record_node_allocation_statistics.
10638 (free_node): Use get_stats_node_kind.
10640 2018-03-14 Richard Biener <rguenther@suse.de>
10642 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10643 that the value-set of ANTIC_IN doesn't grow.
10646 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10648 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10649 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10651 2018-03-14 Julia Koval <julia.koval@intel.com>
10653 * config.gcc (icelake-client, icelake-server): New.
10655 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10656 (initial_ix86_arch_features): Ditto.
10657 (PTA_SKYLAKE): Add SGX.
10658 (PTA_ICELAKE): Remove.
10659 (PTA_ICELAKE_CLIENT): New.
10660 (PTA_ICELAKE_SERVER): New.
10661 (ix86_option_override_internal): Split up icelake on icelake client and
10663 (get_builtin_code_for_version): Ditto.
10664 (fold_builtin_cpu): Ditto.
10665 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10666 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10667 * config/i386/i386.h (processor_type): Ditto.
10668 * doc/invoke.texi: Ditto.
10670 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10673 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10674 INTEGER_CST offset, add it together with bitpos / 8 and
10675 sign extend based on POINTER_SIZE.
10679 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10682 * config/i386/constraints.md (Yc): New register constraint.
10683 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10684 Use Yc constraint for alternative 2 of operand 0. Remove
10685 preferred_for_speed attribute.
10687 2018-03-14 Richard Biener <rguenther@suse.de>
10689 PR tree-optimization/84830
10690 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10691 with the old one to avoid oscillations.
10693 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10696 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10698 (assign_by_spills): Return a flag of reload assignment failure.
10699 Do not process the reload assignment failures. Do not spill other
10700 reload pseudos if they has the same reg class. Update n if
10702 (lra_assign): Add a return arg. Set up from the result of
10703 assign_by_spills call.
10704 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10705 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10706 usage_insns if it is not NULL.
10707 (spill_hard_reg_in_range): New function.
10708 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10709 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10710 function prototypes.
10711 (lra_assign): Change prototype.
10712 * lra.c (lra): Add code to deal with fails by splitting hard reg
10715 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10717 * config/riscv/riscv.opt (mrelax): New option.
10718 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10719 "norelax" when riscv_mrelax is disabled.
10720 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10722 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10725 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10726 reassociation for int modes.
10728 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10730 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10731 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10733 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10734 * config/aarch64/aarch64-sve.md
10735 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10736 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10737 (*extend<mode><Vwide>2): Rename to...
10738 (aarch64_sve_extend<mode><Vwide>2): ...this.
10739 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10740 renaming the old pattern to...
10741 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10743 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10744 define_expand, renaming the old pattern to...
10745 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10746 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10747 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10748 account when deciding which SVE instruction the optab should use.
10749 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10751 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10753 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10754 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10755 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10756 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10757 (tlsdesc_small_<mode>): Turn a define_expand and use
10758 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10759 (tlsdesc_small_advsimd_<mode>): ...this.
10760 (tlsdesc_small_sve_<mode>): New pattern.
10762 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10764 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10765 (UNSPEC_UMUL_HIGHPART): New constants.
10766 (MUL_HIGHPART): New int iteraor.
10767 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10768 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10770 (*<su>mul<mode>3_highpart): New define_insn.
10772 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10775 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10778 2018-03-13 Martin Liska <mliska@suse.cz>
10781 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10783 (sem_item_optimizer::~sem_item_optimizer): Release it.
10784 (sem_item_optimizer::merge_classes): Register variable aliases.
10785 (sem_item_optimizer::fixup_pt_set): New function.
10786 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10787 * ipa-icf.h: Declare new variables and functions.
10789 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10791 PR middle-end/84834
10792 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10793 integer_pow2p@2 and test integer_pow2p in condition.
10794 (A < 0 ? C : 0): Similarly for @1.
10796 PR middle-end/84831
10797 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10798 characters starting at p contain '\0' character, don't look beyond
10802 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10803 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10806 * reg-stack.c (change_stack): Change update_end var from int to
10807 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10808 also call set_block_for_insn on the newly added insns and rescan.
10811 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10812 on the last operand.
10815 * tree.c (stabilize_reference_1): Return save_expr (e) for
10816 STATEMENT_LIST even if it doesn't have side-effects.
10818 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10820 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10822 2018-03-12 Renlin Li <renlin.li@arm.com>
10824 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10825 aarch64_output_scalar_simd_mov_immediate.
10827 2018-03-12 Martin Sebor <msebor@redhat.com>
10829 PR tree-optimization/83456
10830 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10831 for perfectly overlapping calls to memcpy.
10832 (gimple_fold_builtin_memory_chk): Same.
10833 (gimple_fold_builtin_strcpy): Handle no-warning.
10834 (gimple_fold_builtin_stxcpy_chk): Same.
10835 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10837 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10839 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10840 parameter. Use it for SFmode.
10841 (rs6000_function_arg_advance_1): Adjust.
10842 (rs6000_function_arg): Adjust.
10843 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10845 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10847 PR rtl-optimization/84169
10848 PR rtl-optimization/84780
10849 * combine.c (can_combine_p): Check for a 2-insn combination whether
10850 the destination register is used between the two insns, too.
10852 2018-03-12 Richard Biener <rguenther@suse.de>
10854 PR tree-optimization/84803
10855 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10856 for refs DR analysis didn't process.
10858 2018-03-12 Richard Biener <rguenther@suse.de>
10860 PR tree-optimization/84777
10861 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10862 force-vectorize loops ignore whether we are optimizing for size.
10864 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10866 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10867 (TARGET_MD_ASM_ADJUST): Define.
10869 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10870 Kito Cheng <kito.cheng@gmail.com>
10871 Chung-Ju Wu <jasonwucj@gmail.com>
10873 * config/nds32/nds32.c (nds32_compute_stack_frame,
10874 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10875 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10876 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10877 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10878 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10879 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10880 * config/nds32/nds32.md (prologue, epilogue): Use macro
10881 NDS32_V3PUSH_AVAILABLE_P to do checking.
10883 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10886 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10887 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10888 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10889 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10890 addition of most attributes on !orig_type_die or the attribute not
10891 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10893 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10894 Chung-Ju Wu <jasonwucj@gmail.com>
10896 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10897 __NDS32_VH__ macro.
10898 * config/nds32/nds32.opt (mvh): New option.
10900 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10901 Chung-Ju Wu <jasonwucj@gmail.com>
10903 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10905 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10906 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10909 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10910 Chung-Ju Wu <jasonwucj@gmail.com>
10912 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10914 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10915 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10917 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10918 Kito Cheng <kito.cheng@gmail.com>
10919 Chung-Ju Wu <jasonwucj@gmail.com>
10921 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10922 UNSPEC_FFMISM and UNSPEC_FLMISM.
10923 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10924 for ffb, ffmism and flmism.
10925 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10926 (unspec_ffmism): Ditto.
10927 (unspec_flmism): Ditto.
10928 (nds32_expand_builtin_impl): Check if string extension is available.
10929 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10930 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10932 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10935 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10938 * lra-assigns.c (assign_by_spills): Return a flag of reload
10939 assignment failure. Do not process the reload assignment
10940 failures. Do not spill other reload pseudos if they has the same
10942 (lra_assign): Add a return arg. Set up from the result of
10943 assign_by_spills call.
10944 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10945 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10946 usage_insns if it is not NULL.
10947 (spill_hard_reg_in_range): New function.
10948 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10949 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10950 function prototypes.
10951 (lra_assign): Change prototype.
10952 * lra.c (lra): Add code to deal with fails by splitting hard reg
10955 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10958 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10960 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10963 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10964 (dw_val_node): Add val_symbolic_view.
10965 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10966 (symview_upper_bound): New.
10967 (new_line_info_table): Initialize symviews_since_reset.
10968 (dwarf2out_source_line): Count symviews_since_reset and set
10969 symview_upper_bound.
10970 (dw_val_equal_p): Handle symview.
10971 (add_AT_symview): New.
10972 (print_dw_val): Handle symview.
10973 (attr_checksum, attr_checksum_ordered): Likewise.
10974 (same_dw_val_p, size_of_die): Likewise.
10975 (value_format, output_die): Likewise.
10976 (add_high_low_attributes): Use add_AT_symview for entry_view.
10977 (dwarf2out_finish): Reset symview_upper_bound, clear
10980 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10983 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10984 Add strict argument and use it.
10985 (rs6000_split_multireg_move): Update for new strict argument.
10986 (mem_operand_gpr): Disallow all non-offsettable addresses.
10987 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10989 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10992 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10993 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10994 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10997 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10998 decl, use remap_type if we want to use the type.
11000 2018-03-09 Martin Sebor <msebor@redhat.com>
11002 PR tree-optimization/84526
11003 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11005 (builtin_access::generic_overlap): Be prepared to handle non-array
11008 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11010 PR rtl-optimization/84682
11011 * lra-constraints.c (process_address_1): Check is_address flag
11012 for address constraints.
11013 (process_alt_operands): Likewise.
11014 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11015 preprocess_constraints.
11016 * recog.h (preprocess_constraints): Add oploc parameter.
11018 * recog.c (preprocess_constraints): Test address_operand for
11019 CT_ADDRESS constraints.
11021 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11024 * lra-assigns.c (assign_by_spills): Return a flag of reload
11025 assignment failure. Do not process the reload assignment
11026 failures. Do not spill other reload pseudos if they has the same
11028 (lra_assign): Add a return arg. Set up from the result of
11029 assign_by_spills call.
11030 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11031 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11032 usage_insns if it is not NULL.
11033 (spill_hard_reg_in_range): New function.
11034 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11035 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11036 function prototypes.
11037 (lra_assign): Change prototype.
11038 * lra.c (lra): Add code to deal with fails by splitting hard reg
11041 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11044 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11045 Accept complain bool parameter. Only emit errors if it is true.
11046 (arm_parse_cpu_option_name): Likewise.
11047 (arm_target_thumb_only): Adjust callers of the above.
11048 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11049 prototype to take a default true bool parameter.
11050 (arm_parse_arch_option_name): Likewise.
11052 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11053 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11057 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11058 * configure: Regenerate.
11059 * configure.ac ("linker --version-script option"): New.
11060 ("linker soname option"): New.
11062 2018-03-09 Richard Biener <rguenther@suse.de>
11064 PR tree-optimization/84775
11065 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11066 immediate uses of predicate stmts and mark them modified.
11069 PR tree-optimization/84178
11070 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11072 (version_loop_for_if_conversion): Delay update_ssa call.
11073 (tree_if_conversion): Delay update_ssa until after predicate
11076 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11079 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11080 when the function accesses prior frames.
11082 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11085 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11086 gen_llsym, otherwise call maybe_gen_llsym.
11088 PR inline-asm/84742
11089 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11090 has ',' character inside of it.
11092 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11095 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11096 as clobbering CC_REGNUM.
11098 2018-03-08 Richard Biener <rguenther@suse.de>
11100 PR middle-end/84552
11101 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11102 (follow_copies_to_constant): Do not follow SSA names registered
11105 2018-03-08 Richard Biener <rguenther@suse.de>
11107 PR tree-optimization/84178
11108 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11110 (version_loop_for_if_conversion): Delay update_ssa call.
11111 (tree_if_conversion): Delay update_ssa until after predicate
11114 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11116 PR tree-optimization/84178
11117 * tree-if-conv.c (release_bb_predicate): Remove the
11118 the assertion that the stmts have NULL use_ops.
11119 Discard the statements, asserting that they haven't
11120 yet been added to a BB.
11122 2018-03-08 Richard Biener <rguenther@suse.de>
11124 PR tree-optimization/84746
11125 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11126 (phi_translate): Pass in destination ANTIC_OUT set.
11127 (phi_translate_1): Likewise. For a simplified result lookup
11128 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11129 (phi_translate_set): Adjust.
11130 (do_pre_regular_insertion): Likewise.
11131 (do_pre_partial_partial_insertion): Likewise.
11133 2018-03-08 Martin Liska <mliska@suse.cz>
11135 PR gcov-profile/84735
11136 * doc/gcov.texi: Document usage of profile files.
11137 * gcov-io.h: Document changes in the format.
11139 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11143 * dwarf2out.c (struct dw_line_info_table): Update comments for
11145 (FORCE_RESET_NEXT_VIEW): New.
11146 (FORCE_RESETTING_VIEW_P): New.
11147 (RESETTING_VIEW_P): Check for -1 too.
11148 (ZERO_VIEW_P): Likewise.
11149 (new_line_info_table): Force-reset next view.
11150 (dwarf2out_begin_function): Likewise.
11151 (dwarf2out_source_line): Simplify zero_view_p initialization.
11152 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11153 view directly. Omit view when omitting .loc at line 0.
11155 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11157 PR tree-optimization/84740
11158 * tree-switch-conversion.c (process_switch): Call build_constructors
11159 only if info.phi_count is non-zero.
11161 PR tree-optimization/84739
11162 * tree-tailcall.c (find_tail_calls): Check call arguments against
11163 DECL_ARGUMENTS (current_function_decl) rather than
11164 DECL_ARGUMENTS (func) when checking for tail recursion.
11166 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11168 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11169 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11170 Volker Reichelt's entry and add entries for people that perform
11171 GCC fuzzy testing and report numerous bugs.
11173 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11176 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11177 readonly data in sdata, if that is disabled.
11178 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11179 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11180 -mreadonly-in-sdata option.
11182 2018-03-07 Martin Sebor <msebor@redhat.com>
11184 PR tree-optimization/84468
11185 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11186 basic block when looking for nul assignment.
11188 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11191 * except.h (output_function_exception_table): Adjust prototype.
11192 * except.c (output_function_exception_table): Remove FNNAME parameter
11193 and add SECTION parameter. Ouput one part of the table at a time.
11194 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11195 the first part of the exception table and emit unwind directives.
11196 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11197 (i386_pe_seh_cold_init): Likewise.
11198 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11199 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11200 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11201 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11202 * config/i386/winnt.c: Include except.h.
11203 (struct seh_frame_state): Add reg_offset, after_prologue and
11204 in_cold_section fields.
11205 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11206 (i386_pe_seh_cold_init): New function.
11207 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11208 to seh->in_cold_section.
11209 (seh_emit_push): Record the offset of the push.
11210 (seh_emit_save): Record the offet of the save.
11211 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11212 Test seh->after_prologue to disregard the epilogue.
11213 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11214 (i386_pe_end_cold_function): New function.
11216 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11219 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11220 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11223 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11225 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11226 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11228 PR middle-end/84723
11229 * multiple_target.c: Include tree-inline.h and intl.h.
11230 (expand_target_clones): Diagnose and fail if node->definition and
11231 !tree_versionable_function_p (node->decl).
11233 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11235 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11237 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11238 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11239 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11241 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11244 * combine.c (try_combine): Use reg_or_subregno instead of handling
11245 just paradoxical SUBREGs and REGs.
11247 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11249 * config/arc/arc.c (arc_finalize_pic): Remove function.
11250 (arc_must_save_register): We use single base PIC register, remove
11251 checks to save/restore the PIC register.
11252 (arc_expand_prologue): Likewise.
11253 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11255 (arc_verify_short): Likewise.
11256 (arc_attr_type): Likewise.
11257 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11258 (walk_stores): Likewise.
11259 (arc_address_cost): Make it static.
11260 (arc_verify_short): Likewise.
11261 (branch_dest): Likewise.
11262 (arc_attr_type): Likewise.
11263 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11264 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11265 (arc_final_prescan_insn): Remove inserting the nops due to
11266 hardware hazards. It is done in reorg step.
11267 (insn_length_variant_t): Remove.
11268 (insn_length_parameters_t): Likewise.
11269 (arc_insn_length_parameters): Likewise.
11270 (arc_get_insn_variants): Likewise.
11271 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11273 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11275 PR inline-asm/84683
11276 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11279 PR tree-optimization/84687
11280 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11282 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11284 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11286 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11287 Rename to ppc_speculation_barrier.
11288 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11289 __builtin_ppc_speculation_barrier.
11291 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11294 * combine.c (combine_simplify_rtx): Don't try to simplify if
11295 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11298 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11300 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11301 to 32 bytes when compiling for POWER9.
11303 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11306 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11307 regparm >= 3 with no arg reg available also for calls with
11308 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11311 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11313 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11315 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11318 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11320 2018-03-05 Richard Biener <rguenther@suse.de>
11322 PR tree-optimization/84486
11323 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11324 When inserting a __builtin_assume_aligned call set the LHS
11325 SSA name alignment info accordingly.
11327 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11329 PR tree-optimization/84114
11330 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11331 Avoid reassociation of FLOAT_MODE addition.
11333 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11335 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11336 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11337 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11338 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11339 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11340 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11341 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11343 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11344 __builtin_ia32_wbinvd): New builtins.
11345 (SPECIAL_ARGS2): New.
11346 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11347 (SPECIAL_ARGS2): New.
11348 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11349 (ix86_valid_target_attribute_inner_p): Ditto.
11350 (ix86_init_mmx_sse_builtins): Add special_args2.
11351 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11352 TARGET_WBNOINVD_P): New.
11353 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11354 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11355 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11356 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11357 * config/i386/pconfigintrin.h: New file.
11358 * config/i386/wbnoinvdintrin.h: Ditto.
11359 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11361 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11363 2018-03-05 Richard Biener <rguenther@suse.de>
11365 PR tree-optimization/84670
11366 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11368 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11369 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11370 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11371 doesn't grow if all successors (recursively) were visited at least
11374 2018-03-05 Richard Biener <rguenther@suse.de>
11376 PR tree-optimization/84650
11377 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11378 if executed in the loop pipeline.
11380 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11382 * doc/configfiles.texi (Configuration Files): Move info about
11383 conditionalizing $target-protos.h to...
11384 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11385 differs from $target-protos.h.
11387 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11388 Chung-Ju Wu <jasonwucj@gmail.com>
11390 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11391 * config/nds32/nds32-multiple.md (setmemsi): Define.
11392 * config/nds32/nds32-memory-manipulation.c
11393 (nds32_gen_dup_4_byte_to_word_value): New.
11394 (emit_setmem_word_loop): New.
11395 (emit_setmem_byte_loop): New.
11396 (nds32_expand_setmem_loop): New.
11397 (nds32_expand_setmem_loop_v3m): New.
11398 (nds32_expand_setmem_unroll): New.
11399 (nds32_expand_setmem): New.
11401 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11402 Chung-Ju Wu <jasonwucj@gmail.com>
11404 * config/nds32/nds32-memory-manipulation.c
11405 (nds32_emit_load_store): New.
11406 (nds32_emit_post_inc_load_store): New.
11407 (nds32_emit_mem_move): New.
11408 (nds32_emit_mem_move_block): New.
11409 (nds32_expand_movmemsi_loop_unknown_size): New.
11410 (nds32_expand_movmemsi_loop_known_size): New.
11411 (nds32_expand_movmemsi_loop): New.
11412 (nds32_expand_movmemsi_unroll): New.
11413 (nds32_expand_movmemqi): Rename ...
11414 (nds32_expand_movmemsi): ... to this.
11415 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11416 (movmemsi): ... to this.
11417 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11418 (nds32_expand_movmemsi): ... to this.
11420 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11421 Monk Chiang <sh.chiang04@gmail.com>
11422 Chung-Ju Wu <jasonwucj@gmail.com>
11424 * config/nds32/nds32-protos.h
11425 (nds32_expand_load_multiple): New arguments.
11426 (nds32_expand_store_multiple): Ditto.
11427 (nds32_valid_multiple_load_store): Rename ...
11428 (nds32_valid_multiple_load_store_p): ... to this.
11429 * config/nds32/nds32-memory-manipulation.c
11430 (nds32_expand_load_multiple): Refine implementation.
11431 (nds32_expand_store_multiple): Ditto.
11432 * config/nds32/nds32-multiple.md
11433 (load_multiple): Update nds32_expand_load_multiple interface.
11434 (store_multiple): Update nds32_expand_store_multiple interface.
11435 * config/nds32/nds32-predicates.c
11436 (nds32_valid_multiple_load_store): Rename ...
11437 (nds32_valid_multiple_load_store_p): ... to this and refine
11439 * config/nds32/predicates.md
11440 (nds32_load_multiple_and_update_address_operation): New predicate.
11441 (nds32_store_multiple_and_update_address_operation): New predicate.
11443 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11444 Chung-Ju Wu <jasonwucj@gmail.com>
11446 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11447 (combo): New attribute.
11448 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11450 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11452 * config/nds32/nds32.opt: Change -mcmodel= default value.
11454 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11455 Monk Chiang <sh.chiang04@gmail.com>
11456 Chung-Ju Wu <jasonwucj@gmail.com>
11458 * config/nds32/constants.md (unspec_element): New enum.
11459 * config/nds32/constraints.md (Umw): New constraint.
11460 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11461 * config/nds32/nds32-intrinsic.md: Likewise.
11462 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11463 (nds32_valid_smw_lwm_base_p): New.
11464 (nds32_output_smw_single_word): New.
11465 (nds32_output_lmw_single_word): New.
11466 (nds32_expand_unaligned_load): New.
11467 (nds32_expand_unaligned_store): New.
11468 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11469 (nds32_output_smw_single_word): Declare.
11470 (nds32_output_lmw_single_word): Declare.
11471 (nds32_expand_unaligned_load): Declare.
11472 (nds32_expand_unaligned_store): Declare.
11473 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11474 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11475 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11476 NDS32_BUILTIN_UASTORE_DW.
11477 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11480 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11481 Kito Cheng <kito.cheng@gmail.com>
11482 Chung-Ju Wu <jasonwucj@gmail.com>
11484 * config/nds32/nds32-intrinsic.c
11485 (nds32_expand_builtin_null_ftype_reg): Delete.
11486 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11487 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11488 (nds32_read_argument): New.
11489 (nds32_legitimize_target): Ditto.
11490 (nds32_legitimize_argument): Ditto.
11491 (nds32_check_constant_argument): Ditto.
11492 (nds32_expand_unop_builtin): Ditto.
11493 (nds32_expand_unopimm_builtin): Ditto.
11494 (nds32_expand_binop_builtin): Ditto.
11495 (nds32_builtin_decl_impl): Ditto.
11496 (builtin_description): Ditto.
11497 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11498 (nds32_init_builtins_impl): Ditto.
11499 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11500 (nds32_builtin_decl): New.
11501 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11502 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11504 2018-03-02 Jeff Law <law@redhat.com>
11506 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11507 (redundant_insn, fill_simple_delay_slots): Likewise.
11508 (fill_slots_from_thread): Likewise.
11509 * resource.c (mark_referenced_resources): Likewise.
11510 (mark_set_resources, find_dead_or_set_registers): Likewise.
11512 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11514 * substring-locations.h (format_warning_va): Formatting fix for
11515 ATTRIBUTE_GCC_DIAG.
11516 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11518 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11519 * substring-locations.c: Include intl.h.
11520 (format_warning_va): Turned into small wrapper around
11521 format_warning_n_va, renamed to ...
11522 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11523 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11525 (format_warning_at_substring_n): New function.
11526 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11527 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11528 format_warning_at_substring with just a shorter name instead of
11529 const function pointer.
11530 (fmtwarn_n): New function.
11531 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11532 appropriate, get rid of all the fmtstr temporaries, move conditionals
11533 with G_() wrapped string literals directly into fmtwarn arguments,
11534 cast dir.len to (int), formatting fixes.
11536 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11538 * doc/invoke.texi: Remove "Cilk Plus" references.
11540 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11541 Richard Biener <rguenther@suse.de>
11544 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11545 for error or warning attributes if CALL_FROM_THUNK_P is set.
11548 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11551 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11552 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11555 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11556 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11558 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11559 instead of -1U in last predictors element's probability member.
11561 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11564 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11565 arguments if they are comparable.
11567 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11569 PR tree-optimization/84634
11570 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11571 masks and masked_loop_p with a single loop_masks, making sure it's
11572 null for bb vectorization.
11574 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11576 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11577 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11578 loop->force_vectorize to check whether there is no alias.
11580 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11583 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11585 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11587 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11588 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11589 instead of a loop around prev_real_insn.
11590 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11593 PR inline-asm/84625
11594 * config/i386/i386.c (ix86_print_operand): Use conditional
11595 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11598 2018-03-02 Richard Biener <rguenther@suse.de>
11600 PR tree-optimization/84427
11601 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11602 (bitmap_set_subtract_values): Rewrite to handle multiple
11605 (prune_clobbered_mems): Likewise.
11606 (phi_translate): Take edge instead of pred/phiblock.
11607 (phi_translate_1): Likewise.
11608 (phi_translate_set): Likewise. Insert all translated
11609 exprs for a value into the set, keeping possibly multiple
11610 expressions per value.
11611 (compute_antic_aux): Adjust for phi_translate changes.
11612 When intersecting union the expressions and prune those
11613 not in the final value set, keeping possibly multiple
11614 expressions per value. Do not use value-insertion
11615 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11616 all expressions. Add verification that the value-sets
11617 only shrink during iteration.
11618 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11619 (do_pre_regular_insertion): Likewise.
11620 (do_pre_partial_partial_insertion): Likewise.
11622 2018-03-02 Richard Biener <rguenther@suse.de>
11625 * config/darwin.c (saved_debug_info_level): New static global.
11626 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11627 (darwin_asm_lto_end): Restore debug info generation settings.
11629 2018-03-01 Martin Liska <mliska@suse.cz>
11632 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11633 volatile arguments.
11635 2018-03-01 Richard Biener <rguenther@suse.de>
11638 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11639 type annotation with LTO when debug was disabled at compile-time.
11641 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11643 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11645 (mips_final_postscan_insn): Likewise.
11647 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11649 PR rtl-optimization/84528
11650 * alias.c (init_alias_target): Add commentary.
11651 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11652 a unique base value if the frame pointer is not eliminated
11653 to the stack pointer.
11655 2018-03-01 Tom de Vries <tom@codesourcery.com>
11657 PR rtl-optimization/83327
11658 * lra-int.h (hard_regs_spilled_into): Declare.
11659 * lra.c (hard_regs_spilled_into): Define.
11660 (init_reg_info): Init hard_regs_spilled_into.
11661 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11662 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11663 (process_bb_lives): Handle hard_regs_spilled_into.
11664 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11665 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11667 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11669 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11670 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11671 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11672 * config/rs6000/aix72.h: New file.
11674 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11676 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11677 instead of warning_at with conditional singular and plural messages
11681 * stor-layout.c (update_alignment_for_field): For
11682 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11683 && !DECL_PACKED (field), do the alignment update, just use
11684 only desired_align instead of MAX (type_align, desired_align)
11686 (place_field): Don't do known_align < desired_align handling
11687 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11688 is non-NULL, instead do it after rli->prev_field handling and
11689 only if not within a bitfield word. For DECL_PACKED (field)
11690 use type_align of BITS_PER_UNIT.
11692 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11694 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11695 superfluous parentheses and trailing spaces.
11697 2018-02-28 Richard Biener <rguenther@suse.de>
11699 PR tree-optimization/84584
11700 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11701 SCoPs with fake exit edge.
11703 2018-02-28 Martin Liska <mliska@suse.cz>
11706 * timevar.c (timer::print): Fix format to properly print 100%
11709 2018-02-28 Richard Biener <rguenther@suse.de>
11711 PR middle-end/84607
11712 * genmatch.c (capture_info::walk_match): Do not mark
11713 captured expressions without operands as expr_p given
11714 they act more like predicates and should be subject to
11715 "lost tail" side-effect preserving.
11717 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11719 PR rtl-optimization/81611
11720 * auto-inc-dec.c (attempt_change): Move dead note from
11721 mem_insn if it's the next use of regno
11722 (find_address): Take address use of reg holding
11723 non-incremented value. Add parm to limit search to the named
11725 (merge_in_block): Attempt to use a mem insn that is the next
11726 use of the original regno.
11728 2018-02-27 Martin Sebor <msebor@redhat.com>
11731 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11732 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11734 2018-02-27 Martin Sebor <msebor@redhat.com>
11736 PR translation/84207
11737 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11738 n argument to unsigned HOST_WIDE_INT.
11739 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11740 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11741 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11742 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11744 2018-02-27 Richard Biener <rguenther@suse.de>
11746 PR tree-optimization/84512
11747 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11748 Do not use the estimate returned from record_stmt_cost for
11749 the scalar iteration cost but sum properly using add_stmt_cost.
11751 2018-02-27 Richard Biener <rguenther@suse.de>
11753 PR tree-optimization/84466
11754 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11755 Adjust last change to less strictly validate use operands.
11757 2018-02-27 Martin Liska <mliska@suse.cz>
11759 PR gcov-profile/84548
11760 * gcov.c (process_file): Allow partial overlap and consider it
11761 also as group functions.
11762 (output_lines): Properly calculate range of lines for a group.
11764 2018-02-27 Martin Liska <mliska@suse.cz>
11766 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11767 'ggc' suffixes. Change first column width.
11768 (timer::print): Fix formatting of the column.
11770 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11772 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11773 preserve inline entry blocks for the sake of debug inline
11774 entry point markers alone.
11775 (remove_unused_locals): Suggest in comments a better place to
11776 force the preservation of inline entry blocks that are
11777 otherwise unused, but do not preserve them.
11779 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11781 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11783 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11786 * config/i386/constraints.md (Bs): Replace
11787 ix86_indirect_branch_register with
11788 TARGET_INDIRECT_BRANCH_REGISTER.
11790 * config/i386/i386.md (indirect_jump): Likewise.
11791 (tablejump): Likewise.
11792 (*sibcall_memory): Likewise.
11793 (*sibcall_value_memory): Likewise.
11794 Peepholes of indirect call and jump via memory: Likewise.
11795 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11796 (*sibcall_value_GOT_32): Likewise.
11797 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11798 (GOT_memory_operand): Likewise.
11799 (call_insn_operand): Likewise.
11800 (sibcall_insn_operand): Likewise.
11801 (GOT32_symbol_operand): Likewise.
11802 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11804 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11806 PR rtl-optimization/83496
11807 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11808 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11809 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11810 redundant insn, if any.
11811 (relax_delay_slots): Likewise.
11812 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11814 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11816 PR tree-optimization/83965
11817 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11818 that grouped statements are part of a reduction chain. Return
11819 true if the statement is not marked as a reduction itself but
11820 is part of a group.
11821 (vect_recog_dot_prod_pattern): Don't check whether the statement
11822 is part of a group here.
11823 (vect_recog_sad_pattern): Likewise.
11824 (vect_recog_widen_sum_pattern): Likewise.
11826 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11829 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11831 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11834 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11836 (ix86_output_indirect_function_return): New prototype.
11837 (ix86_split_simple_return_pop_internal): Likewise.
11838 * config/i386/i386.c (indirect_return_via_cx): New.
11839 (indirect_return_via_cx_bnd): Likewise.
11840 (indirect_thunk_name): Handle return va CX_REG.
11841 (output_indirect_thunk_function): Create alias for
11842 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11843 (ix86_output_indirect_jmp): Remove the bool argument.
11844 (ix86_output_indirect_function_return): New function.
11845 (ix86_split_simple_return_pop_internal): Likewise.
11846 * config/i386/i386.md (*indirect_jump): Don't pass false
11847 to ix86_output_indirect_jmp.
11848 (*tablejump_1): Likewise.
11849 (simple_return_pop_internal): Change it to define_insn_and_split.
11850 Call ix86_split_simple_return_pop_internal to split it for
11851 -mfunction-return=.
11852 (simple_return_indirect_internal): Call
11853 ix86_output_indirect_function_return instead of
11854 ix86_output_indirect_jmp.
11856 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11859 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11860 memset and value initialization afterwards.
11862 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11864 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11866 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11869 * common/config/aarch64/aarch64-common.c
11870 (aarch_option_optimization_table[]): Switch
11871 off fomit-frame-pointer
11873 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11874 Chung-Ju Wu <jasonwucj@gmail.com>
11876 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11878 (store_multiple): Ditto.
11880 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11882 * config.gcc: Add --with-cpu support for nds32 target.
11883 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11884 * config/nds32/nds32.opt: Add -mcpu= option.
11886 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11888 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11889 isel=yes): Warn for these deprecated options.
11891 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11893 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11894 ISA_2_5_MASKS_EMBEDDED.
11896 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11898 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11899 p->max as pointers rather than using iterative_hash_expr.
11901 2018-02-23 Carl Love <cel@us.ibm.com>
11903 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11904 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11906 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11907 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11908 P8V_BUILTIN_VEC_VUNSIGNED2.
11910 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11913 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11914 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11916 * lra-constraints.c (curr_insn_transform): Set up
11917 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11919 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11920 * lra-eliminations.c (spill_pseudos): Ditto.
11921 (process_insn_for_elimination): Ditto.
11922 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11923 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11924 LRA_NON_CLOBBERED_ALT.
11926 2018-02-22 Martin Sebor <msebor@redhat.com>
11928 PR tree-optimization/84480
11929 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11930 to maybe_diag_stxncpy_trunc. Call it.
11931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11932 from gimple_fold_builtin_strcpy. Print inlining stack.
11933 (handle_builtin_stxncpy): Print inlining stack.
11934 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11936 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11939 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11940 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11941 and -fcheck-pointer-bounds are used together.
11942 (indirect_thunk_prefix): New enum.
11943 (indirect_thunk_need_prefix): New function.
11944 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11945 "_nt" instead of "_bnd" for NOTRACK prefix.
11946 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11947 (output_indirect_thunk_function): Likewise.
11949 (ix86_code_end): Update output_indirect_thunk_function calls.
11950 (ix86_output_indirect_branch_via_reg): Replace
11951 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11952 (ix86_output_indirect_branch_via_push): Likewise.
11953 (ix86_output_function_return): Likewise.
11954 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11955 incompatible with -fcf-protection=branch and
11956 -fcheck-pointer-bounds.
11958 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11961 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11962 Change gcc_assert call to output_operand_lossage.
11964 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11966 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11968 2018-02-22 DJ Delorie <dj@redhat.com>
11969 Sebastian Perta <sebastian.perta@renesas.com>
11970 Oleg Endo <olegendo@gcc.gnu.org>
11972 * config/rx/rx.c (rx_rtx_costs): New function.
11973 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11975 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11977 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11979 2018-02-22 Martin Liska <mliska@suse.cz>
11982 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11983 Add "native" as a possible value.
11985 2018-02-22 Martin Liska <mliska@suse.cz>
11988 * config/i386/i386.c (ix86_option_override_internal):
11989 Add "native" as a possible value for -march and -mtune.
11991 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11994 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11995 to all type variants.
11997 PR tree-optimization/84503
11998 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11999 width as info->bitpos + info->bitsize - start.
12000 (merged_store_group::merge_overlapping): Simplify width computation.
12001 (check_no_overlap): New function.
12002 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12003 start + width and last_order of the group, fail if check_no_overlap
12005 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12006 to group if check_no_overlap fails.
12008 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12010 * config/rs6000/altivec.md: Delete contraint arguments to
12011 define_expand, define_split, and define_peephole2, and in
12012 define_insn_and_split if always unused.
12013 * config/rs6000/darwin.md: Ditto.
12014 * config/rs6000/dfp.md: Ditto.
12015 * config/rs6000/rs6000.md: Ditto.
12016 * config/rs6000/sync.md: Ditto.
12017 * config/rs6000/vector.md: Ditto.
12018 * config/rs6000/vsx.md: Ditto.
12020 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12022 * config/rs6000/altivec.md: Write output control strings as braced
12023 blocks instead of double-quoted strings.
12024 * config/rs6000/darwin.md: Ditto.
12025 * config/rs6000/rs6000.md: Ditto.
12026 * config/rs6000/vector.md: Ditto.
12027 * config/rs6000/vsx.md: Ditto.
12029 2018-02-21 Jason Merrill <jason@redhat.com>
12031 PR c++/84314 - ICE with templates and fastcall attribute.
12032 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12034 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12036 * ipa-cp.c (determine_versionability): Fix comment typos.
12038 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12041 * ipa-cp.c (determine_versionability): Do not version functions caling
12044 2018-02-21 Martin Liska <mliska@suse.cz>
12047 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12048 Add "native" as a possible value.
12049 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12050 the macro when native cpu detection is available.
12052 2018-02-21 Martin Liska <mliska@suse.cz>
12055 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12056 Add "native" as a possible value.
12057 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12058 when native cpu detection is available.
12060 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12061 Martin Sebor <msebor@redhat.com>
12063 PR tree-optimization/84478
12064 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12066 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12067 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12068 support which is conservatively correct, for 2 only stay conservative
12069 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12070 argument to the 2 argument get_range_strlen, adjust 6 arg
12071 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12073 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12074 (gimple_fold_builtin_strlen): Pass true as last argument to
12077 2018-02-20 Martin Sebor <msebor@redhat.com>
12079 PR middle-end/84095
12080 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12081 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12082 (builtin_memref::builtin_memref): Factor out parts into
12083 set_base_and_offset and call it.
12085 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12087 PR middle-end/84406
12088 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12089 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12090 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12091 search at the associated MODE_INT.
12093 2018-02-20 Jeff Law <law@redhat.com>
12095 PR middle-end/82123
12096 PR tree-optimization/81592
12097 PR middle-end/79257
12098 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12099 for range data rather than using global data.
12100 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12101 range data rather than using global data.
12102 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12103 pass it to children as needed.
12104 (struct directive::fmtresult): Similarly.
12105 (struct directive::set_width): Similarly.
12106 (struct directive::set_precision): Similarly.
12107 (format_integer, format_directive, parse_directive): Similarly.
12108 (format_none): Accept unnamed vr_values parameter.
12109 (format_percent, format_floating, format_character): Similarly.
12110 (format_string, format_plain): Similarly.
12111 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12112 the EVRP range analyzer for range data rather than using global data.
12113 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12114 gimple-ssa-evrp-analyze.h
12115 (class sprintf_dom_walker): Add after_dom_children member function.
12116 Add evrp_range_analyzer member.
12117 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12118 range analyzer as needed.
12119 (sprintf_dom_walker::after_dom_children): New member function.
12120 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12122 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12123 (evrp_range_analyzer::pop_to_marker): Likewise.
12125 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12127 PR tree-optimization/84419
12128 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12129 with the required type if its current type is compatible but
12132 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12134 PR middle-end/82004
12135 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12136 after vectorization.
12138 2018-02-20 Martin Liska <mliska@suse.cz>
12141 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12142 possible values if we don't have a hint.
12144 2018-02-20 Martin Liska <mliska@suse.cz>
12148 * final.c (shorten_branches): Build align_tab array with one
12150 * opts.c (finish_options): Add alignment option limit check.
12151 (MAX_CODE_ALIGN): Likewise.
12152 (MAX_CODE_ALIGN_VALUE): Likewise.
12153 * doc/invoke.texi: Document maximum allowed option value for
12154 all -falign-* options.
12156 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12159 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12160 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12161 * var-tracking.c (emit_note_insn_var_location): Remove all references
12162 to NOTE_INSN_CALL_ARG_LOCATION.
12163 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12164 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12165 Use copy_rtx_if_shared.
12166 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12167 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12168 (dwarf2out_var_location): Remove handling of
12169 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12171 * final.c (final_scan_insn): Remove all references to
12172 NOTE_INSN_CALL_ARG_LOCATION.
12173 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12174 before dumping final insns.
12175 * except.c (emit_note_eh_region_end): Remove all references to
12176 NOTE_INSN_CALL_ARG_LOCATION.
12177 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12178 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12179 * config/arc/arc.c (hwloop_optimize): Likewise.
12180 * config/arm/arm.c (create_fix_barrier): Likewise.
12181 * config/s390/s390.c (s390_chunkify_start): Likewise.
12182 * config/sh/sh.c (find_barrier): Likewise.
12183 * config/i386/i386.c (rest_of_insert_endbranch,
12184 ix86_seh_fixup_eh_fallthru): Likewise.
12185 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12186 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12187 * config/frv/frv.c (frv_function_prologue): Likewise.
12188 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12190 (note_outside_basic_block_p): Remove all references to
12191 NOTE_INSN_CALL_ARG_LOCATION.
12192 * gengtype.c (adjust_field_rtx_def): Likewise.
12193 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12195 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12196 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12199 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12202 PR tree-optimization/84452
12203 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12204 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12207 2018-02-19 Martin Liska <mliska@suse.cz>
12210 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12212 2018-02-19 Martin Liska <mliska@suse.cz>
12213 Richard Sandiford <richard.sandiford@linaro.org>
12215 PR tree-optimization/82491
12216 * gimple-fold.c (get_base_constructor): Make earlier bail out
12219 2018-02-19 Carl Love <cel@us.ibm.com>
12221 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12222 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12224 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12225 P8V_BUILTIN_VEC_NEG.
12227 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12229 * config/rl78/rl78.md (movdf): New define expand.
12231 2018-02-19 Martin Liska <mliska@suse.cz>
12234 * doc/invoke.texi: Fix typo.
12235 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12237 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12239 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12240 handle rs6000_single_float and rs6000_double_float specially for
12243 2018-02-16 Jeff Law <law@redhat.com>
12245 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12246 (pop_regs): Use it.
12248 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12251 * ipa-inline.c (inline_small_functions): Fix a typo.
12253 2018-02-16 Nathan Sidwell <nathan@acm.org>
12255 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12257 2018-02-16 Carl Love <cel@us.ibm.com>
12259 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12260 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12261 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12262 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12263 expansion to P8V_BUILTIN_VEC_FLOAT2.
12265 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12267 PR rtl-optimization/70023
12268 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12269 src_regno into account.
12271 2018-02-16 Carl Love <cel@us.ibm.com>
12273 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12274 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12275 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12276 * config/rs6000/rs6000.c: Remove case statements for
12277 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12278 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12279 and P9V_BUILTIN_VEC_VINSERT4B.
12280 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12281 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12282 * config/rs6000/vsx.md:
12283 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12286 2018-02-16 Carl Love <cel@us.ibm.com>
12288 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12290 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12292 * config/rs6000/rs6000-c.c: Add the definitions for
12293 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12294 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12295 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12296 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12297 definition for insert4b and define insn *insert3b_internal.
12298 * doc/extend.texi: Add documentation for vec_extract4b.
12300 2018-02-16 Nathan Sidwell <nathan@acm.org>
12302 * doc/extend.texi (Backwards Compatibility): Mention friend
12303 injection. Note for-scope is deprecated.
12304 * doc/invoke.texi (-ffriend-injection): Deprecate.
12306 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12308 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12309 that moved to I2, also allow destinations that are a paradoxical
12310 subreg (instead of a normal reg).
12312 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12315 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12318 2018-02-16 Richard Biener <rguenther@suse.de>
12320 PR tree-optimization/84037
12321 PR tree-optimization/84016
12323 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12324 Adjust vec_construct for the fact we need additional higher latency
12325 128bit inserts for AVX256 and AVX512 vector builds.
12326 (ix86_add_stmt_cost): Scale vector construction cost for
12329 2018-02-16 Richard Biener <rguenther@suse.de>
12331 PR tree-optimization/84417
12332 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12333 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12334 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12336 2018-02-16 Martin Liska <mliska@suse.cz>
12339 * internal-fn.def (ASAN_CHECK): Set proper flags.
12340 (ASAN_MARK): Likewise.
12342 2018-02-16 Julia Koval <julia.koval@intel.com>
12344 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12345 from PTA_CANNONLAKE.
12347 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12350 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12351 Use ++iter rather than iter++ for std::list iterators.
12352 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12353 defer deleting them until all nodes in the forest are processed. Do
12354 free even leaf nodes. Change to_process into auto_vec.
12357 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12358 * vec.h (vec_default_construct): Use memset instead of placement new
12359 if BROKEN_VALUE_INITIALIZATION is defined.
12360 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12361 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12364 PR rtl-optimization/83723
12365 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12366 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12367 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12369 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12371 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12373 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12375 PR rtl-optimization/81443
12376 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12377 from inner REGs to paradoxical SUBREGs.
12379 2018-02-16 Richard Biener <rguenther@suse.de>
12381 PR tree-optimization/84399
12382 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12383 For operands we can analyze at their definition make sure we can
12384 analyze them at each use as well.
12386 2018-02-16 Richard Biener <rguenther@suse.de>
12388 PR tree-optimization/84190
12389 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12390 volatile accesses if the decl isn't volatile.
12392 2018-02-15 Jason Merrill <jason@redhat.com>
12394 PR c++/84314 - ICE with templates and fastcall attribute.
12395 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12396 TYPE_CANONICAL on an existing type.
12398 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12400 PR tree-optimization/84383
12401 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12402 dstoff nor call operand_equal_p if dstbase is NULL.
12404 PR tree-optimization/84334
12405 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12406 also a CONSTANT_CLASS_P, punt.
12408 2018-02-14 Jim Wilson <jimw@sifive.com>
12410 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12411 first SMALL_OPERAND check. New local min_second_step. Move assert
12412 to where locals are set. Add TARGET_RVC support.
12413 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12415 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12417 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12419 2018-02-14 Martin Sebor <msebor@redhat.com>
12421 PR tree-optimization/83698
12422 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12423 arrays constrain the offset range to their bounds.
12424 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12425 (builtin_access::overlap): Avoid setting the size of overlap if it's
12427 (maybe_diag_overlap): Also consider arrays when deciding what values
12428 of offsets to include in diagnostics.
12430 2018-02-14 Martin Sebor <msebor@redhat.com>
12433 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12434 that correspond to the kind of a declaration.
12436 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12439 * config/pa/pa.md: Load address of PIC label using the linkage table
12440 if the label is nonlocal.
12442 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12444 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12445 warning message if user requests -maltivec=be.
12446 * doc/invoke.texi: Document deprecation of -maltivec=be.
12448 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12451 * config/rs6000/rs6000-c.c: Update definitions for
12452 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12453 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12455 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12458 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12459 add _get_ssp intrinsics. Remove argument from
12460 __builtin_ia32_rdssp[d|q].
12461 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12462 * config/i386/i386-builtin.def: Remove argument from
12463 __builtin_ia32_rdssp[d|q].
12464 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12465 ix86_expand_special_args_builtin for _rdssp[d|q].
12466 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12467 Clear register before usage.
12468 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12469 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12471 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12473 PR tree-optimization/84357
12474 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12475 operand 1 of an ARRAY_REF too.
12477 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12480 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12481 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12483 (set_of_reg): New struct.
12484 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12485 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12486 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12488 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12489 Split into bitclr, bitset, bitinvert patterns if appropriate.
12490 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12491 use rx_fuse_in_memory_bitop.
12492 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12493 to named insn, correct maximum insn length.
12495 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12498 * machmode.def: Define a complex mode for PARTIAL_INT.
12499 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12501 * doc/rtl.texi: Document CSPImode.
12502 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12504 (msp430_hard_regno_nregs_with_padding): Likewise.
12506 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12509 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12511 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12513 PR rtl-optimization/84169
12514 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12515 we generated a parallel as new i3 and we split that to new i2 and i3
12516 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12517 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12518 those to i2, not i1. Partially rewrite this scan code.
12520 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12523 * stor-layout.c (place_field): For variable length fields, adjust
12524 offset_align afterwards not just based on the field's alignment,
12525 but also on the size.
12527 PR middle-end/84309
12528 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12529 of exps and logs in the use_exp2 case.
12531 2018-02-13 Jeff Law <law@redhat.com>
12533 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12534 entry for "vector".
12536 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12539 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12543 * common.opt (gas-loc-support, gas-locview-support): New.
12544 (ginline-points, ginternal-reset-location-views): New.
12545 * doc/invoke.texi: Document them. Use @itemx where intended.
12546 (gvariable-location-views): Adjust.
12547 * target.def (reset_location_view): New.
12548 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12549 (TARGET_RESET_LOCATION_VIEW): New.
12550 * doc/tm.texi: Rebuilt.
12551 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12552 (dwarf2out_default_as_locview_support): New.
12553 (output_asm_line_debug_info): Use option variables.
12554 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12555 (output_loc_list): Likewise.
12556 (add_high_low_attributes): Check option variables.
12557 Don't output entry view attribute in strict mode.
12558 (gen_inlined_subroutine_die): Check option variables.
12559 (dwarf2out_inline_entry): Likewise.
12560 (init_sections_and_labels): Likewise.
12561 (dwarf2out_early_finish): Likewise.
12562 (maybe_reset_location_view): New, from...
12563 (dwarf2out_var_location): ... here. Call it.
12564 * debug.h (dwarf2out_default_as_loc_support): Declare.
12565 (dwarf2out_default_as_locview_support): Declare.
12566 * hooks.c (hook_int_rtx_insn_0): New.
12567 * hooks.h (hook_int_rtx_insn_0): Declare.
12568 * toplev.c (process_options): Take -gas-loc-support and
12569 -gas-locview-support from dwarf2out. Enable
12570 -gvariable-location-views by default only with locview
12571 assembler support. Enable -ginternal-reset-location-views by
12572 default only if the target defines the corresponding hook.
12573 Enable -ginline-points by default if location views are
12574 enabled; force it disabled if statement frontiers are
12576 * tree-inline.c (expand_call_inline): Check option variables.
12577 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12579 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12581 PR tree-optimization/84321
12582 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12583 handling. Also check whether the anti-range contains any values
12584 that satisfy the mask; switch to a VR_RANGE if not.
12586 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12589 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12591 2018-02-13 Martin Jambor <mjambor@suse.cz>
12594 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12595 of call statements, also set location of a load to a temporary.
12597 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12599 * config/rl78/rl78.c (add_vector_labels): New function.
12600 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12601 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12602 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12603 which checks that no arguments are passed.
12604 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12605 * doc/extend.texi: Documentation for the new attribute.
12607 2018-02-13 Andreas Schwab <schwab@suse.de>
12609 * config/riscv/linux.h (CPP_SPEC): Define.
12611 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12614 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12615 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12616 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12617 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12618 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12619 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12620 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12621 temporarily for AES and PCLMUL builtins.
12623 PR tree-optimization/84339
12624 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12625 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12628 PR middle-end/84309
12629 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12630 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12631 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12633 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12635 * omp-simd-clone.h: New file.
12636 * omp-simd-clone.c: Include omp-simd-clone.h.
12637 (expand_simd_clones): No longer static.
12638 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12639 cgraph.h and omp-simd-clone.h.
12640 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12641 (vect_recog_widen_shift_pattern): Formatting fix.
12642 (vect_pattern_recog_1): Don't check optab for calls.
12645 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12646 operands[2] into a REG before using gen_lowpart on it.
12648 2018-02-12 Jeff Law <law@redhat.com>
12651 * config/sh/sh.c (find_barrier): Consider a sibling call
12654 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12655 successfully back substituting a reg.
12657 2018-02-12 Richard Biener <rguenther@suse.de>
12659 PR tree-optimization/84037
12660 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12661 parameter, move visited init to caller.
12662 (vect_slp_analyze_operations): Separate cost from validity
12663 check, initialize visited once for all instances.
12664 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12666 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12668 (vect_model_promotion_demotion_cost): Likewise.
12669 (vectorizable_bswap): Guard cost modeling with !slp_node
12670 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12672 (vectorizable_call): Likewise.
12673 (vectorizable_conversion): Likewise.
12674 (vectorizable_assignment): Likewise.
12675 (vectorizable_shift): Likewise.
12676 (vectorizable_operation): Likewise.
12677 (vectorizable_store): Likewise.
12678 (vectorizable_load): Likewise.
12679 (vectorizable_condition): Likewise.
12680 (vectorizable_comparison): Likewise.
12682 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12685 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12686 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12687 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12689 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12691 PR middle-end/83665
12692 * params.def (inline-min-speedup): Increase from 8 to 15.
12693 (max-inline-insns-auto): Decrease from 40 to 30.
12694 * ipa-split.c (consider_split): Add some buffer for function to
12695 be considered inlining candidate.
12696 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12699 2018-02-12 Richard Biener <rguenther@suse.de>
12701 PR tree-optimization/84037
12702 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12703 matched stmts if we cannot swap the non-matched ones.
12705 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12707 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12708 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12709 _mm_maskz_scalef_round_ss): New intrinsics.
12710 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12711 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12712 __builtin_ia32_scalefss_round): Remove.
12713 (__builtin_ia32_scalefsd_mask_round,
12714 __builtin_ia32_scalefss_mask_round): New intrinsics.
12715 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12716 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12717 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12718 "<round_constraint>")): Changed to ...
12719 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12720 "<round_scalar_constraint>")): ... this.
12721 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12722 %0, %1, %2<round_op3>}"): Changed to ...
12723 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12724 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12725 %2<round_scalar_mask_op3>}"): ... this.
12726 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12728 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12730 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12731 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12732 (_mm_maskz_sqrt_round_ss): New intrinsics.
12733 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12734 (__builtin_ia32_sqrtsd_mask_round)
12735 (__builtin_ia32_sqrtss_mask_round): New builtins.
12736 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12737 (__builtin_ia32_sqrtss_round): Remove.
12738 (__builtin_ia32_sqrtsd_mask_round)
12739 (__builtin_ia32_sqrtss_mask_round): New builtins.
12740 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12741 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12742 ((match_operand:VF_128 1 "vector_operand"
12743 "xBm,<round_constraint>")): Changed to ...
12744 ((match_operand:VF_128 1 "vector_operand"
12745 "xBm,<round_scalar_constraint>")): ... this.
12746 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12747 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12748 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12749 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12750 %<iptr>1<round_scalar_mask_op3>}): ... this.
12751 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12752 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12754 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12757 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12758 Cast vec_cmpeq result to correct type.
12759 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12760 Cast vec_cmpgt result to correct type.
12762 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12764 * final.c (final_scan_insn_1): Renamed from...
12765 (final_scan_insn): ... this. New wrapper, to recover
12766 seen from the outermost call in recursive ones.
12767 * config/sparc/sparc.c (output_return): Drop seen from call.
12768 (output_sibcall): Likewise.
12769 * config/visium/visium.c (output_branch): Likewise.
12771 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12773 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12776 2018-02-10 Alan Modra <amodra@gmail.com>
12779 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12780 Specify LR as an input.
12782 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12785 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12786 remove_member_access_dummy_vars): New functions.
12787 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12788 lower_omp_1, execute_lower_omp): Use them.
12790 PR rtl-optimization/84308
12791 * shrink-wrap.c (spread_components): Release todo vector.
12793 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12795 PR rtl-optimization/57193
12796 * ira-color.c (struct allocno_color_data): Add member
12797 conflict_allocno_hard_prefs.
12798 (update_conflict_allocno_hard_prefs): New.
12799 (bucket_allocno_compare_func): Add a preference based on
12800 conflict_allocno_hard_prefs.
12801 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12802 (color_allocnos): Remove a dead code. Initiate
12803 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12805 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12808 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12809 constraint from =wa to wa. Avoid a subreg on the output operand,
12810 instead use a pseudo and subreg it in a move.
12811 (p9_xxbrd_<mode>): Changed to ...
12812 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12813 (p9_xxbrd_v2df): New expander.
12814 (p9_xxbrw_<mode>): Changed to ...
12815 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12816 (p9_xxbrw_v4sf): New expander.
12818 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12820 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12822 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12825 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12826 multiply in 32-bit mode.
12827 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12828 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12831 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12833 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12834 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12835 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12836 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12838 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12841 * dwarf2out.c (is_trivial_indirect_ref): New function.
12842 (dwarf2out_late_global_decl): Do not generate a location
12843 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12844 and that are not defined in the current unit.
12846 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12848 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12849 instead of a libcall for UNORDERED.
12851 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12854 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12855 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12857 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12860 * config/s390/s390.c (s390_set_current_function): Invoke
12861 s390_indirect_branch_settings also if fndecl didn't change.
12863 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12865 * config/rs6000/rs6000.md (blockage): Set length to zero.
12867 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12869 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12871 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12874 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12875 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12879 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12880 PARALLEL incoming that failed vt_get_decl_and_offset check.
12882 PR middle-end/84237
12883 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12884 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12886 (get_variable_section): For decls in named .bss* sections pass true as
12887 second argument to bss_initializer_p.
12889 2018-02-09 Marek Polacek <polacek@redhat.com>
12890 Jakub Jelinek <jakub@redhat.com>
12893 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12894 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12895 Sync some changes from cxx_fold_indirect_ref.
12897 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12899 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12901 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12902 (BLOCK_INLINE_ENTRY_LABEL): New.
12903 (dwarf2out_var_location): Disregard inline entry markers.
12904 (inline_entry_data): New struct.
12905 (inline_entry_data_hasher): New hashtable type.
12906 (inline_entry_data_hasher::hash): New.
12907 (inline_entry_data_hasher::equal): New.
12908 (inline_entry_data_table): New variable.
12909 (add_high_low_attributes): Add DW_AT_entry_pc and
12910 DW_AT_GNU_entry_view attributes if a pending entry is found
12911 in inline_entry_data_table. Add old entry_pc attribute only
12912 if debug nonbinding markers are disabled.
12913 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12914 markers are enabled.
12915 (block_within_block_p, dwarf2out_inline_entry): New.
12916 (dwarf2out_finish): Check that no entries remained in
12917 inline_entry_data_table.
12918 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12919 (final_scan_insn, notice_source_line): Likewise.
12920 (rest_of_clean_state): Skip inline entry markers.
12921 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12923 * gimple.c (gimple_build_debug_inline_entry): New.
12924 * gimple.h (enum gimple_debug_subcode): Add
12925 GIMPLE_DEBUG_INLINE_ENTRY.
12926 (gimple_build_debug_inline_entry): Declare.
12927 (gimple_debug_inline_entry_p): New.
12928 (gimple_debug_nonbind_marker_p): Adjust.
12929 * insn-notes.def (INLINE_ENTRY): New.
12930 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12931 inline entry marker notes.
12932 (print_insn): Likewise.
12933 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12934 (INSN_DEBUG_MARKER_KIND): Likewise.
12935 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12936 * tree-inline.c (expand_call_inline): Build and insert
12937 debug_inline_entry stmt.
12938 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12939 inline entry blocks early, if nonbind markers are enabled.
12940 (dump_scope_block): Dump fragment info.
12941 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12942 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12943 (gimple_build_debug_inline_entry): New.
12944 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12945 Enable/disable inline entry points too.
12946 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12947 (DEBUG_INSN): Describe inline entry markers.
12949 * common.opt (gvariable-location-views): New.
12950 (gvariable-location-views=incompat5): New.
12951 * config.in: Rebuilt.
12952 * configure: Rebuilt.
12953 * configure.ac: Test assembler for view support.
12954 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12955 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12956 * dwarf2out.c (var_loc_view): New typedef.
12957 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12958 (dwarf2out_locviews_in_attribute): New.
12959 (dwarf2out_locviews_in_loclist): New.
12960 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12961 (enum dw_line_info_opcode): Add LI_adv_address.
12962 (struct dw_line_info_table): Add view.
12963 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12964 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12965 (zero_view_p): New variable.
12966 (ZERO_VIEW_P): New macro.
12967 (output_asm_line_debug_info): New.
12968 (struct var_loc_node): Add view.
12969 (add_AT_view_list, AT_loc_list): New.
12970 (add_var_loc_to_decl): Add view param. Test it against last.
12971 (new_loc_list): Add view params. Record them.
12972 (AT_loc_list_ptr): Handle loc and view lists.
12973 (view_list_to_loc_list_val_node): New.
12974 (print_dw_val): Handle dw_val_class_view_list.
12975 (size_of_die): Likewise.
12976 (value_format): Likewise.
12977 (loc_list_has_views): New.
12978 (gen_llsym): Set vl_symbol too.
12979 (maybe_gen_llsym, skip_loc_list_entry): New.
12980 (dwarf2out_maybe_output_loclist_view_pair): New.
12981 (output_loc_list): Output view list or entries too.
12982 (output_view_list_offset): New.
12983 (output_die): Handle dw_val_class_view_list.
12984 (output_dwarf_version): New.
12985 (output_compilation_unit_header): Use it.
12986 (output_skeleton_debug_sections): Likewise.
12987 (output_rnglists, output_line_info): Likewise.
12988 (output_pubnames, output_aranges): Update version comments.
12989 (output_one_line_info_table): Output view numbers in asm comments.
12990 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12991 Call maybe_gen_llsym.
12992 (loc_list_from_tree_1): Adjust.
12993 (add_AT_location_description): Create view list attribute if
12994 needed, check it's absent otherwise.
12995 (convert_cfa_to_fb_loc_list): Adjust.
12996 (maybe_emit_file): Call output_asm_line_debug_info for test.
12997 (dwarf2out_var_location): Reset views as needed. Precompute
12998 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12999 attribute. Set view.
13000 (new_line_info_table): Reset next view.
13001 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13002 (dwarf2out_source_line): Likewise. Output view resets and labels to
13003 the assembler, or select appropriate line info opcodes.
13004 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13005 (optimize_string_length): Catch it. Adjust.
13006 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13007 dw_val_class_view_list, and remove it if no longer needed.
13008 (hash_loc_list): Hash view numbers.
13009 (loc_list_hasher::equal): Compare them.
13010 (optimize_location_lists): Check whether a view list symbol is
13011 needed, and whether the locview attribute is present, and
13012 whether they match. Remove the locview attribute if no longer
13014 (index_location_lists): Call skip_loc_list_entry for test.
13015 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13016 Use output_dwarf_version.
13017 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13018 (struct dw_val_node): Add val_view_list.
13019 * final.c (SEEN_NEXT_VIEW): New.
13020 (set_next_view_needed): New.
13021 (clear_next_view_needed): New.
13022 (maybe_output_next_view): New.
13023 (final_start_function): Rename to...
13024 (final_start_function_1): ... this. Take pointer to FIRST,
13025 add SEEN parameter. Emit param bindings in the initial view.
13026 (final_start_function): Reintroduce SEEN-less interface.
13027 (final): Rename to...
13028 (final_1): ... this. Take SEEN parameter. Output final pending
13029 next view at the end.
13030 (final): Reintroduce seen-less interface.
13031 (final_scan_insn): Output pending next view before switching
13032 sections or ending a block. Mark the next view as needed when
13033 outputting variable locations. Notify debug backend of section
13034 changes, and of location view changes.
13035 (rest_of_handle_final): Adjust.
13036 * toplev.c (process_options): Autodetect value for debug variable
13037 location views option. Warn on incompat5 without -gdwarf-5.
13038 * doc/invoke.texi (gvariable-location-views): New.
13039 (gvariable-location-views=incompat5): New.
13040 (gno-variable-location-views): New.
13042 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13044 PR tree-optimization/84136
13045 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13046 that the result of find_edge is non-NULL.
13048 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13051 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13052 storing integer register in SImode. Fix cost of 256 and 512
13053 byte aligned SSE register store.
13055 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13057 * config/i386/i386.c (ix86_multiplication_cost): Fix
13058 multiplication cost for TARGET_AVX512DQ.
13060 2018-02-08 Marek Polacek <polacek@redhat.com>
13062 PR tree-optimization/84238
13063 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13066 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13068 PR tree-optimization/84265
13069 * tree-vect-stmts.c (vectorizable_store): Don't treat
13070 VMAT_CONTIGUOUS accesses as grouped.
13071 (vectorizable_load): Likewise.
13073 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13075 PR tree-optimization/81635
13076 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13077 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13078 (test_round_for_mask): New functions.
13079 (wide_int_cc_tests): Call test_round_for_mask.
13080 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13081 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13082 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13083 range returned by get_range_info.
13085 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13088 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13089 * symtab.c: Include builtins.h
13090 (symtab_node::output_to_lto_symbol_table_p): Move here
13091 from lto-streamer-out.c:output_symbol_p.
13092 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13093 (output_symbol_p): Move all logic to symtab.c
13094 (produce_symtab): Update.
13096 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13098 * config/s390/s390-opts.h (enum indirect_branch): Define.
13099 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13100 (s390_indirect_branch_via_thunk)
13101 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13102 (enum s390_indirect_branch_type): Define.
13103 * config/s390/s390.c (struct s390_frame_layout, struct
13104 machine_function): Remove.
13105 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13106 (indirect_branch_table_label_no, indirect_branch_table_name):
13108 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13109 (enum s390_indirect_branch_option): Define.
13110 (s390_return_addr_from_memory): New function.
13111 (s390_handle_string_attribute): New function.
13112 (s390_attribute_table): Add new attribute handler.
13113 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13114 (s390_indirect_branch_via_thunk): New function.
13115 (s390_indirect_branch_via_inline_thunk): New function.
13116 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13117 sibling call optimization for non z10 compiles.
13118 (s390_emit_call): Force indirect branch target to be a single
13119 register. Add r1 clobber for non-z10 compiles.
13120 (s390_emit_epilogue): Emit return jump via return_use expander.
13121 (s390_reorg): Handle JUMP_INSNs as execute targets.
13122 (s390_option_override_internal): Perform validity checks for the
13123 new command line options.
13124 (s390_indirect_branch_attrvalue): New function.
13125 (s390_indirect_branch_settings): New function.
13126 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13127 (s390_output_indirect_thunk_function): New function.
13128 (s390_code_end): Implement target hook.
13129 (s390_case_values_threshold): Implement target hook.
13130 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13132 * config/s390/s390.h (struct s390_frame_layout)
13133 (struct machine_function): Move here from s390.c.
13134 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13135 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13136 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13137 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13138 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13139 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13140 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13141 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13142 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13143 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13144 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13145 (mnemonic attribute): Add values which aren't recognized
13147 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13148 pattern for branch conversion. Fix mnemonic attribute.
13149 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13150 indirect branch via thunk if requested.
13151 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13152 ("*indirect_jump"): Disable for branch conversion using out of
13154 ("indirect_jump_via_thunk<mode>_z10")
13155 ("indirect_jump_via_thunk<mode>")
13156 ("indirect_jump_via_inlinethunk<mode>_z10")
13157 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13158 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13159 ("casesi_jump_via_inlinethunk<mode>_z10")
13160 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13161 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13162 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13163 ("*indirect2_jump"): Disable for branch conversion.
13164 ("casesi_jump"): Turn into expander and expand patterns for branch
13166 ("return_use"): New expander.
13167 ("*return"): Emit return via thunk and rename it to ...
13168 ("*return<mode>"): ... this one.
13169 * config/s390/s390.opt: Add new options and and enum for the
13172 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13174 * lra-constraints.c (match_reload): Unconditionally use
13175 gen_lowpart_SUBREG, rather than selecting between that
13176 and equivalent gen_rtx_SUBREG code.
13178 2018-02-08 Richard Biener <rguenther@suse.de>
13180 PR tree-optimization/84233
13181 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13182 changed flag instead of boguously re-using phi_inserted.
13184 2018-02-08 Martin Jambor <mjambor@suse.cz>
13186 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13187 static local variables.
13189 2018-02-08 Richard Biener <rguenther@suse.de>
13191 PR tree-optimization/84278
13192 * tree-vect-stmts.c (vectorizable_store): When looking for
13193 smaller vector types to perform grouped strided loads/stores
13194 make sure the mode is supported by the target.
13195 (vectorizable_load): Likewise.
13197 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13199 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13200 Increase LDP/STP opportunities by adding adjacent callee-saves.
13202 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13204 PR rtl-optimization/84068
13205 PR rtl-optimization/83459
13206 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13208 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13210 PR tree-optimization/84224
13211 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13212 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13213 non-zero arguments.
13215 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13218 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13219 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13220 count, remove one USE.
13222 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13224 * doc/install.texi (Configuration): Document the
13225 --with-long-double-format={ibm,ieee} PowerPC configuration
13229 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13230 Convert from define_expand to be define_insn_and_split. Rework
13231 float/double/_Float128 conversions to QI/HI/SImode to work with
13232 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13233 conversions to QI/HImode types did a store and then a load to
13234 truncate the value. For conversions to VSX registers, don't split
13235 the insn, instead emit the code directly. Use the code iterator
13236 any_fix to combine signed and unsigned conversions.
13237 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13238 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13239 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13240 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13241 (fix_<mode>di2_hw): Likewise.
13242 (fixuns_<mode>di2_hw): Likewise.
13243 (fix_<mode>si2_hw): Likewise.
13244 (fixuns_<mode>si2_hw): Likewise.
13245 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13246 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13247 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13248 fix<uns>_trunc<SFDF:mode>si2_p8.
13249 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13250 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13251 (fix<uns>_<mode>_mem): Likewise.
13252 (fctiw<u>z_<mode>_mem): Likewise.
13253 (fix<uns>_<mode>_mem): Likewise.
13254 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13255 the register allocator from doing a direct move to the GPRs to do
13256 a store, and instead use the ISA 3.0 store byte/half-word from
13257 vector register instruction. For IEEE 128-bit floating point,
13258 also optimize stores of 32-bit ints.
13259 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13261 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13263 * genextract.c (push_pathstr_operand): New function to support
13265 (walk_rtx): Call push_pathstr_operand.
13266 (print_path): Support [a-zA-Z].
13268 2018-02-07 Richard Biener <rguenther@suse.de>
13270 PR tree-optimization/84037
13271 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13272 (cse_and_gimplify_to_preheader): Declare.
13273 (vect_get_place_in_interleaving_chain): Likewise.
13274 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13276 (_loop_vec_info::~_loop_vec_info): Delete it.
13277 (cse_and_gimplify_to_preheader): New function.
13278 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13279 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13280 (vectorizable_load): Likewise. For grouped stores always base
13281 the IV on the first element.
13282 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13283 condition before gimplifying.
13285 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13287 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13288 *DIV_EXPR and *MOD_EXPR.
13290 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13293 * config/i386/i386.c (ix86_option_override_internal): Mask out
13294 the CF_SET bit when checking -fcf-protection.
13296 2018-02-07 Tom de Vries <tom@codesourcery.com>
13299 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13302 2018-02-07 Richard Biener <rguenther@suse.de>
13304 PR tree-optimization/84204
13305 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13308 PR tree-optimization/84205
13309 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13310 special-case isl_ast_op_zdiv_r.
13312 PR tree-optimization/84223
13313 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13314 Only add conditions from within the region.
13315 (gather_bbs::after_dom_children): Adjust.
13317 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13320 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13321 * config/avr/avr.md: Only post-reload split REG-REG moves if
13322 either register is GENERAL_REG_P.
13324 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13326 PR tree-optimization/84235
13327 * tree-ssa-scopedtables.c
13328 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13329 if the subtraction is performed in floating point type where NaNs are
13330 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13331 build 1. Formatting fix.
13333 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13336 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13337 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13338 and skip it regardless of bb boundaries. Use CALL_P macro,
13339 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13340 unnecessarily, formatting fix.
13342 2018-02-06 Michael Collison <michael.collison@arm.com>
13344 * config/arm/thumb2.md:
13345 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13346 (*thumb_mov_notscc): Ditto.
13348 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13351 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13352 unsigned_fix, not "s".
13354 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13356 * configure.ac (gcc_fn_eh_frame_ro): New function.
13357 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13358 correct .eh_frame permissions.
13359 * configure: Regenerate.
13361 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13363 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13364 irrelevant options.
13366 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13368 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13369 Display warning message for -mno-speculate-indirect-jumps.
13371 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13373 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13375 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13377 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13379 PR tree-optimization/84225
13380 * tree-eh.c (find_trapping_overflow): Only call
13381 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13383 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13386 * config/i386/i386.c: Reimplement the check of possible options
13387 -mibt/-mshstk conbination. Change error messages.
13388 * doc/invoke.texi: Fix a typo: remove extra '='.
13390 2018-02-06 Marek Polacek <polacek@redhat.com>
13392 PR tree-optimization/84228
13393 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13395 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13398 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13399 emitted arch directives.
13400 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13401 __ARM_FEATURE_COPROC before changing architectures.
13403 2018-02-06 Richard Biener <rguenther@suse.de>
13405 * config/i386/i386.c (print_reg): Fix typo.
13406 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13408 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13410 * configure: Regenerate.
13412 2018-02-05 Martin Sebor <msebor@redhat.com>
13414 PR tree-optimization/83369
13415 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13418 2018-02-05 Martin Liska <mliska@suse.cz>
13420 * doc/invoke.texi: Cherry-pick upstream r323995.
13422 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13424 * ira.c (ira_init_register_move_cost): Adjust comment.
13426 2018-02-05 Martin Liska <mliska@suse.cz>
13428 PR gcov-profile/84137
13429 * doc/gcov.texi: Fix typo in documentation.
13431 2018-02-05 Martin Liska <mliska@suse.cz>
13433 PR gcov-profile/83879
13434 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13435 using dlopen functionality.
13437 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13439 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13440 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13441 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13442 _mm_maskz_range_round_ss): New intrinsics.
13443 (__builtin_ia32_rangesd128_round)
13444 (__builtin_ia32_rangess128_round): Remove.
13445 (__builtin_ia32_rangesd128_mask_round,
13446 __builtin_ia32_rangess128_mask_round): New builtins.
13447 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13448 __builtin_ia32_rangess128_round): Remove.
13449 (__builtin_ia32_rangesd128_mask_round,
13450 __builtin_ia32_rangess128_mask_round): New builtins.
13451 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13452 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13453 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13454 "<round_saeonly_constraint>")): Changed to ...
13455 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13456 "<round_saeonly_scalar_constraint>")): ... this.
13457 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13458 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13459 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13460 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13461 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13463 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13465 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13467 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13468 Remove all values except native, 8540 and 8548.
13470 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13472 * config/i386/i386.c (ix86_output_function_return): Pass
13473 INVALID_REGNUM, instead of -1, as invalid register number to
13474 indirect_thunk_name and output_indirect_thunk.
13476 2018-02-02 Julia Koval <julia.koval@intel.com>
13478 * config.gcc: Add -march=icelake.
13479 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13480 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13481 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13482 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13483 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13484 (processor_target_table): Add icelake.
13485 (ix86_option_override_internal): Handle new PTAs.
13486 (get_builtin_code_for_version): Handle icelake.
13487 (M_INTEL_COREI7_ICELAKE): New.
13488 (fold_builtin_cpu): Handle icelake.
13489 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13490 * doc/invoke.texi: Add -march=icelake.
13492 2018-02-02 Julia Koval <julia.koval@intel.com>
13494 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13495 to wide_int_bitmask.
13496 * wide-int-bitmask.h: New.
13498 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13501 * config/i386/i386.md: Replace Pmode with word_mode in
13502 builtin_setjmp_setup and builtin_longjmp to support x32.
13504 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13508 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13510 (rs6000_supported_cpu_names): New static variable.
13511 (linux_cpu_translation_table): Likewise.
13512 (elf_platform) <cpu>: Define new static variable and use it.
13513 Translate kernel AT_PLATFORM name to canonical name if needed.
13514 Error if platform name is unknown.
13516 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13519 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13521 2018-02-01 Jeff Law <law@redhat.com>
13524 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13525 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13526 the scratch if RELEASE_VIA_POP is false.
13527 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13528 If we have to save a temporary register, decrement SIZE appropriately.
13529 Pass new arguments to release_scratch_register_on_entry.
13530 (ix86_adjust_stack_and_probe): Likewise.
13531 (ix86_emit_probe_stack_range): Pass new arguments to
13532 release_scratch_register_on_entry.
13534 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13536 PR rtl-optimization/84157
13537 * combine.c (change_zero_ext): Use REG_P predicate in
13538 front of HARD_REGISTER_P predicate.
13540 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13542 * config/avr/avr.c (avr_option_override): Move disabling of
13543 -fdelete-null-pointer-checks to...
13544 * common/config/avr/avr-common.c (avr_option_optimization_table):
13547 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13549 PR tree-optimization/81635
13550 * tree-data-ref.c (split_constant_offset_1): For types that
13551 wrap on overflow, try to use range info to prove that wrapping
13554 2018-02-01 Renlin Li <renlin.li@arm.com>
13557 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13558 TAILCALL_ADDR_REGS.
13559 (aarch64_register_move_cost): Likewise.
13560 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13561 TAILCALL_ADDR_REGS.
13562 (REG_CLASS_NAMES): Likewise.
13563 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13564 TAILCALL_ADDR_REGS. Remove IP registers.
13565 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13567 2018-02-01 Richard Biener <rguenther@suse.de>
13569 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13570 for specifying RPO order and allow NULL for that.
13571 * domwalk.c (dom_walker::dom_walker): Likewise.
13572 (dom_walker::walk): Handle NULL RPO order.
13573 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13575 (rewrite_update_dom_walker): Likewise.
13576 (mark_def_dom_walker): Likewise.
13578 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13580 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13581 (aarch64_maybe_expand_sve_subreg_move): Declare.
13582 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13583 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13585 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13586 that are semantically a reverse operation.
13587 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13588 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13589 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13591 (aarch64_can_change_mode_class): For big-endian, forbid changes
13592 between two SVE modes if they have different element sizes.
13594 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13596 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13597 the TImode handling for big-endian targets.
13599 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13601 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13602 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13604 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13605 Remove BSWAP handing for big-endian targets and use the form of
13606 LD1RQ appropariate for the mode.
13608 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13610 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13611 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13612 duplicated element.
13614 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13617 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13618 check for operands that need to go through aarch64_sve_reload_be.
13620 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13622 PR tree-optimization/81661
13623 PR tree-optimization/84117
13624 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13625 * tree-eh.c: Include gimplify.h.
13626 (find_trapping_overflow, replace_trapping_overflow,
13627 rewrite_to_non_trapping_overflow): New functions.
13628 * tree-vect-loop.c: Include tree-eh.h.
13629 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13630 * tree-data-ref.c: Include tree-eh.h.
13631 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13633 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13635 PR rtl-optimization/84123
13636 * combine.c (change_zero_ext): Check if hard register satisfies
13637 can_change_dest_mode before calling gen_lowpart_SUBREG.
13639 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13642 * ira.c (ira_init_register_move_cost): Remove assert.
13644 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13646 PR rtl-optimization/84071
13647 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13648 * doc/tm.texi: Regenerate.
13650 2018-01-31 Richard Biener <rguenther@suse.de>
13652 PR tree-optimization/84132
13653 * tree-data-ref.c (analyze_miv_subscript): Properly
13654 check whether evolution_function_is_affine_multivariate_p
13655 before calling gcd_of_steps_may_divide_p.
13657 2018-01-31 Julia Koval <julia.koval@intel.com>
13660 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13661 * config/i386/i386.md (rdpid_rex64) New.
13662 (rdpid): Make 32bit only.
13664 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13667 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13668 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13670 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13673 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13675 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13677 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13679 PR rtl-optimization/84071
13680 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13681 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13683 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13685 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13686 (arc_attribute_table): Add 'aux' attribute.
13687 (arc_in_small_data_p): Consider aux like variables.
13688 (arc_is_aux_reg_p): New function.
13689 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13690 (arc_get_aux_arg): New function.
13691 (prepare_move_operands): Handle aux-register access.
13692 (arc_handle_aux_attribute): New function.
13693 * doc/extend.texi (ARC Variable attributes): Add subsection.
13695 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13697 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13698 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13699 (arc_attribute_table): Add 'uncached' attribute.
13700 (arc_print_operand): Print '.di' flag for uncached memory
13702 (arc_in_small_data_p): Do not consider for small data the uncached
13704 (arc_is_uncached_mem_p): New function.
13705 * config/arc/predicates.md (compact_store_memory_operand): Check
13706 for uncached memory accesses.
13707 (nonvol_nonimm_operand): Likewise.
13708 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13710 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13713 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13714 falign-loops=): Add Optimization flag.
13716 2018-01-30 Jeff Law <law@redhat.com>
13719 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13720 INT_REGISTERS_SAVED. Check it prior to calling
13721 get_scratch_register_on_entry.
13722 (ix86_adjust_stack_and_probe): Similarly.
13723 (ix86_emit_probe_stack_range): Similarly.
13724 (ix86_expand_prologue): Corresponding changes.
13726 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13729 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13730 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13732 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13735 * lra-constraints.c (curr_insn_transform): Process AND in the
13738 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13740 PR rtl-optimization/83986
13741 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13742 dependence against last_pending_memory_flush in addition to
13743 pending_jump_insns.
13745 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13747 PR tree-optimization/81611
13748 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13751 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13754 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13757 2018-01-30 Richard Biener <rguenther@suse.de>
13758 Jakub Jelinek <jakub@redhat.com>
13760 PR tree-optimization/84111
13761 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13762 inner loops added during recursion, as they don't have up-to-date
13765 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13768 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13769 (can_inline_edge_by_limits_p): ... here.
13770 (can_early_inline_edge_p, check_callers,
13771 update_caller_keys, update_callee_keys, recursive_inlining,
13772 add_new_edges_to_heap, speculation_useful_p,
13773 inline_small_functions,
13774 inline_small_functions, flatten_function,
13775 inline_to_all_callers_1): Update.
13777 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13779 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13782 2018-01-30 Richard Biener <rguenther@suse.de>
13784 PR tree-optimization/83008
13785 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13786 invariant and constant vector uses in stmts when they need
13787 more than one stmt.
13789 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13792 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13793 * configure: Regenerate.
13795 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13797 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13799 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13800 Use gen_rtx_REG rather than gen_lowpart.
13802 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13804 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13805 rather than 0 when creating partial subregs.
13807 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13809 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13812 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13815 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13816 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13817 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13818 flags. This restores the settings used before the 2017-07-24.
13819 Turning off pre increment/decrement/modify allows IVOPTS to
13820 optimize DF/SF loops where the index is an int.
13822 2018-01-29 Richard Biener <rguenther@suse.de>
13823 Kelvin Nilsen <kelvin@gcc.gnu.org>
13826 * tree-vect-stmts.c (vectorizable_call): Don't call
13827 targetm.vectorize_builtin_md_vectorized_function if callee is
13830 2018-01-22 Carl Love <cel@us.ibm.com>
13832 * doc/extend.tex: Fix typo in second arg in
13833 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13835 2018-01-29 Richard Biener <rguenther@suse.de>
13837 PR tree-optimization/84086
13838 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13839 (flush_ssaname_freelist): When SSA names were released reset
13840 the SCEV hash table.
13842 2018-01-29 Richard Biener <rguenther@suse.de>
13844 PR tree-optimization/84057
13845 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13846 removed paths when removing edges.
13848 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13850 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13851 -mfunction-return=@var{choice}.
13853 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13855 PR diagnostic/84034
13856 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13857 Handle CR like TAB.
13858 (layout::print_source_line): Likewise.
13859 (test_get_line_width_without_trailing_whitespace): Add test cases.
13861 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13863 PR middle-end/84040
13864 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13867 2018-01-26 Jim Wilson <jimw@sifive.com>
13869 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13871 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13874 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13876 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13877 and CMP + SUB-immediate -> SUBS.
13879 2018-01-26 Martin Sebor <msebor@redhat.com>
13881 PR tree-optimization/83896
13882 * tree-ssa-strlen.c (get_string_len): Rename...
13883 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13884 Avoid assuming length is constant.
13885 (handle_char_store): Use HOST_WIDE_INT for string length.
13887 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13890 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13891 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13893 2018-01-26 Richard Biener <rguenther@suse.de>
13895 PR rtl-optimization/84003
13896 * dse.c (record_store): Only record redundant stores when
13897 the earlier store aliases at least all accesses the later one does.
13899 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13901 PR rtl-optimization/83985
13902 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13903 REG_CFA_RESTORE insns.
13904 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13905 REG_CFA_RESTORE insns here.
13908 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13909 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13911 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13913 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13914 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13915 (arc_init): Likewise.
13916 (arc_override_options): Likewise.
13917 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13919 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13921 * config/arc/arc.h (TARGET_DBNZ): Define.
13922 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13923 properly set the tune attribute.
13924 (dbnz): Use TARGET_DBNZ guard.
13925 * config/arc/arc.opt (mtune): Add core3 option.
13927 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13929 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13930 recognize new pic like addresses.
13931 (arc_delegitimize_address): Clean up.
13933 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13935 * config/arc/arc-arches.def: Option mrf16 valid for all
13937 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13938 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13939 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13940 * config/arc/arc-tables.opt: Regenerate.
13941 * config/arc/arc.c (arc_conditional_register_usage): Handle
13942 reduced register file case.
13943 (arc_file_start): Set must have build attributes.
13944 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13945 mrf16 option value.
13946 * config/arc/arc.opt (mrf16): Add new option.
13947 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13948 * config/arc/genmultilib.awk: Handle new mrf16 option.
13949 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13950 * config/arc/t-multilib: Regenerate.
13951 * doc/invoke.texi (ARC Options): Document mrf16 option.
13953 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13955 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13956 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13957 (arc_attribute_table): Add 'secure_call' attribute.
13958 (arc_print_operand): Print secure call operand.
13959 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13961 (arc_is_secure_call_p): New function. * config/arc/arc.md
13962 (call_i): Add support for sjli instruction.
13963 (call_value_i): Likewise.
13964 * config/arc/constraints.md (Csc): New constraint.
13966 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13967 John Eric Martin <John.Martin@emmicro-us.com>
13969 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13970 * config/arc/arc.c (_arc_jli_section): New struct.
13971 (arc_jli_section): New type.
13972 (rc_jli_sections): New static variable.
13973 (arc_handle_jli_attribute): New function.
13974 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13975 (arc_file_end): New function.
13976 (TARGET_ASM_FILE_END): Define.
13977 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13978 (arc_add_jli_section): New function.
13979 (jli_call_scan): Likewise.
13980 (arc_reorg): Call jli_call_scan.
13981 (arc_output_addsi): Remove 'S' from printing asm operand.
13982 (arc_is_jli_call_p): New function.
13983 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13985 (movhi_insn): Likewise.
13986 (movsi_insn): Likewise.
13987 (movsi_set_cc_insn): Likewise.
13988 (loadqi_update): Likewise.
13989 (load_zeroextendqisi_update): Likewise.
13990 (load_signextendqisi_update): Likewise.
13991 (loadhi_update): Likewise.
13992 (load_zeroextendhisi_update): Likewise.
13993 (load_signextendhisi_update): Likewise.
13994 (loadsi_update): Likewise.
13995 (loadsf_update): Likewise.
13996 (movsicc_insn): Likewise.
13997 (bset_insn): Likewise.
13998 (bxor_insn): Likewise.
13999 (bclr_insn): Likewise.
14000 (bmsk_insn): Likewise.
14001 (bicsi3_insn): Likewise.
14002 (cmpsi_cc_c_insn): Likewise.
14003 (movsi_ne): Likewise.
14004 (movsi_cond_exec): Likewise.
14005 (clrsbsi2): Likewise.
14006 (norm_f): Likewise.
14016 (call_i): Remove 'S' asm letter, add jli instruction.
14017 (call_value_i): Likewise.
14018 * config/arc/arc.op (mjli-always): New option.
14019 * config/arc/constraints.md (Cji): New constraint.
14020 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14022 (subsf3_fpx): Likewise.
14023 (mulsf3_fpx): Likewise.
14024 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14026 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14027 function attrbutes.
14028 * doc/invoke.texi (ARC): Document mjli-always option.
14030 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14032 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14033 avoid addition with 0 and use incw and decw where possible.
14035 2018-01-26 Richard Biener <rguenther@suse.de>
14037 PR tree-optimization/81082
14038 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14039 association if it requires casting to unsigned.
14040 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14041 from fold_plusminus_mult_expr to catch important cases late when
14042 range info is available.
14044 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14046 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14047 * configure.ac (hidden_linkonce): New test.
14048 * configure: Regenerate.
14049 * config.in: Regenerate.
14051 2018-01-26 Julia Koval <julia.koval@intel.com>
14053 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14054 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14055 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14056 _mm_mask_bitshuffle_epi64_mask): Fix type.
14057 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14058 USI_FTYPE_V4DI_V4DI_USI): Remove.
14059 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14060 __builtin_ia32_vpshufbitqmb256_mask,
14061 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14062 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14063 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14065 2018-01-26 Alan Modra <amodra@gmail.com>
14068 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14069 UNSPEC_VBPERMQ. Sort other unspecs.
14071 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14073 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14075 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14077 PR middle-end/83055
14078 * predict.c (drop_profile): Do not push/pop cfun; update also
14080 (handle_missing_profiles): Fix logic looking for zero profiles.
14082 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14084 PR middle-end/83977
14085 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14086 on functions with #pragma omp declare simd or functions with simd
14088 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14089 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14090 Remove trailing \n from warning_at calls.
14092 2018-01-25 Tom de Vries <tom@codesourcery.com>
14095 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14096 for neutered workers.
14098 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14101 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14102 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14104 2018-01-24 Jeff Law <law@redhat.com>
14107 * i386.c (get_probe_interval): Move to earlier point.
14108 (ix86_compute_frame_layout): If -fstack-clash-protection and
14109 the frame is larger than the probe interval, then use pushes
14110 to save registers rather than reg->mem moves.
14111 (ix86_expand_prologue): Remove conditional for int_registers_saved
14114 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14117 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14118 min/max for never referenced object.
14120 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14122 PR middle-end/83977
14123 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14125 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14126 attributes from DECL_ATTRIBUTES (decl) without affecting
14127 DECL_ATTRIBUTES (current_function_decl).
14128 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14129 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14131 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14133 PR tree-optimization/83979
14134 * fold-const.c (fold_comparison): Use constant_boolean_node
14135 instead of boolean_{true,false}_node.
14137 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14139 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14142 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14144 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14145 Simplify the clause that sets the length attribute.
14146 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14147 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14148 clause that sets the length attribute.
14149 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14151 2018-01-24 Tom de Vries <tom@codesourcery.com>
14154 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14155 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14156 Add strict parameter.
14157 (prevent_branch_around_nothing): Insert dummy insn between branch to
14158 label and label with no ptx insn inbetween.
14159 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14161 2018-01-24 Tom de Vries <tom@codesourcery.com>
14164 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14165 for neutered threads in warp.
14166 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14168 2018-01-24 Richard Biener <rguenther@suse.de>
14170 PR tree-optimization/83176
14171 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14174 2018-01-24 Richard Biener <rguenther@suse.de>
14176 PR tree-optimization/82819
14177 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14178 code generating pluses that are no-ops in the target precision.
14180 2018-01-24 Richard Biener <rguenther@suse.de>
14182 PR middle-end/84000
14183 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14185 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14187 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14188 to merge probabilities.
14189 * predict.c (probably_never_executed): Also mark as cold functions
14190 with global 0 profile and guessed local profile.
14191 * profile-count.c (profile_probability::combine_with_count): New
14193 * profile-count.h (profile_probability::operator*,
14194 profile_probability::operator*=, profile_probability::operator/,
14195 profile_probability::operator/=): Reduce precision to adjusted
14196 and set value to guessed on contradictory divisions.
14197 (profile_probability::combine_with_freq): Remove.
14198 (profile_probability::combine_wiht_count): Declare.
14199 (profile_count::force_nonzero):: Set to adjusted.
14200 (profile_count::probability_in):: Set quality to adjusted.
14201 * tree-ssa-tail-merge.c (replace_block_by): Use
14202 combine_with_count.
14204 2018-01-23 Andrew Waterman <andrew@sifive.com>
14205 Jim Wilson <jimw@sifive.com>
14207 * config/riscv/riscv.c (riscv_stack_boundary): New.
14208 (riscv_option_override): Set riscv_stack_boundary. Handle
14209 riscv_preferred_stack_boundary_arg.
14210 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14211 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14212 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14213 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14214 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14215 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14217 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14220 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14221 of struct ix86_frame.
14222 (ix86_expand_epilogue): Likewise. Add a local variable for
14223 the reg_save_offset field in struct ix86_frame.
14225 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14227 PR tree-optimization/82604
14228 * tree-loop-distribution.c (enum partition_kind): New enum item
14229 PKIND_PARTIAL_MEMSET.
14230 (partition_builtin_p): Support above new enum item.
14231 (generate_code_for_partition): Ditto.
14232 (compute_access_range): Differentiate cases that equality can be
14233 proven at all loops, the innermost loops or no loops.
14234 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14235 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14236 (finalize_partitions, distribute_loop): Don't fuse partition of
14237 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14238 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14239 parloop is enabled.
14241 2018-01-23 Martin Liska <mliska@suse.cz>
14243 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14244 order to ignore the predictor.
14245 (PRED_POLYMORPHIC_CALL): Likewise.
14246 (PRED_RECURSIVE_CALL): Likewise.
14248 2018-01-23 Martin Liska <mliska@suse.cz>
14250 * tree-profile.c (tree_profiling): Print function header to
14251 aware reader which function we are working on.
14252 * value-prof.c (gimple_find_values_to_profile): Do not print
14253 not interesting value histograms.
14255 2018-01-23 Martin Liska <mliska@suse.cz>
14257 * profile-count.h (enum profile_quality): Add
14258 profile_uninitialized as the first value. Do not number values
14259 as they are zero based.
14260 (profile_count::verify): Update sanity check.
14261 (profile_probability::verify): Likewise.
14263 2018-01-23 Nathan Sidwell <nathan@acm.org>
14265 * doc/invoke.texi (ffor-scope): Deprecate.
14267 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14269 PR tree-optimization/83510
14270 * domwalk.c (set_all_edges_as_executable): New function.
14271 (dom_walker::dom_walker): Convert bool param
14272 "skip_unreachable_blocks" to enum reachability. Move setup of
14273 edge flags to set_all_edges_as_executable and only do it when
14274 reachability is REACHABLE_BLOCKS.
14275 * domwalk.h (enum dom_walker::reachability): New enum.
14276 (dom_walker::dom_walker): Convert bool param
14277 "skip_unreachable_blocks" to enum reachability.
14278 (set_all_edges_as_executable): New decl.
14279 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14280 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14282 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14283 but converting true to REACHABLE_BLOCKS.
14284 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14286 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14287 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14288 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14290 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14291 if check_all_array_refs will be called.
14293 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14295 * tree.c (selftest::test_location_wrappers): Add more test
14298 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14300 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14301 (selftest::test_bit_in_range): Likewise.
14303 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14306 * doc/sourcebuild.texi (vect_float): Say that the selector
14307 only describes the situation when -funsafe-math-optimizations is on.
14308 (vect_float_strict): Document.
14310 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14312 PR tree-optimization/83965
14313 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14314 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14315 instead of checking only for a reduction.
14316 (vect_recog_widen_sum_pattern): Likewise.
14318 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14320 * predict.c (probably_never_executed): Only use precise profile info.
14321 (compute_function_frequency): Skip after inlining hack since we now
14322 have quality checking.
14324 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14326 * profile-count.h (profile_probability::very_unlikely,
14327 profile_probability::unlikely, profile_probability::even): Set
14328 precision to guessed.
14330 2018-01-23 Richard Biener <rguenther@suse.de>
14332 PR tree-optimization/83963
14333 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14334 Properly terminate dominator walk when crossing the exit edge not
14335 when visiting its source block.
14337 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14340 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14341 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14343 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14345 PR tree-optimization/83957
14346 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14347 semicolon after for body surrounded by braces.
14349 PR tree-optimization/83081
14350 * profile-count.h (profile_probability::split): New method.
14351 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14352 Use profile_probability::split.
14353 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14354 when splitting a single conditional jump into 2.
14356 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14358 PR tree-optimization/69452
14359 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14362 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14364 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14365 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14366 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14368 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14370 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14372 * config/rl78/rl78.md (movdi): New define_expand.
14373 * config/rl78/rl78.c (rl78_split_movdi): New function.
14375 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14378 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14380 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14381 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14382 128-bit to produce an UNSPEC move to get the double word with the
14383 signbit and then a shift directly to do signbit.
14384 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14385 implementation with a new version that just does either a direct
14386 move or a regular move. Move memory interface to separate insns.
14387 Move insns so they are next to the expander.
14388 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14389 with signbit move. Split big and little endian case.
14390 (signbit<mode>2_dm_mem_le): Likewise.
14391 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14392 (signbit<mode>2_dm2): Likewise.
14394 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14396 * config/rl78/rl78.md (anddi3): New define_expand.
14398 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14400 * config/rl78/rl78.md (umindi3): New define_expand.
14402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14404 * config/rl78/rl78.md (smindi3): New define_expand.
14406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14408 * config/rl78/rl78.md (smaxdi3): New define_expand.
14410 2018-01-22 Carl Love <cel@us.ibm.com>
14412 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14413 LVX_V1TI): Add macro expansion.
14414 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14415 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14416 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14417 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14418 Change check to determine if the instruction is a byte reversing
14419 entry. Fix typo in comment.
14420 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14421 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14422 Add def_builtin calls for new builtins.
14423 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14424 Add define_insn expansion.
14426 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14428 * config/rl78/rl78.md (umaxdi3): New define_expand.
14430 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14432 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14433 for non-QImode registers.
14435 2018-01-22 Richard Biener <rguenther@suse.de>
14437 PR tree-optimization/83963
14438 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14439 including the loop exit block.
14440 (scop_detection::merge_sese): Likewise.
14441 (scop_detection::add_scop): Do it here instead.
14443 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14445 * doc/sourcebuild.texi (arm_softfloat): Document.
14447 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14450 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14451 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14452 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14454 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14455 David Edelsohn <dje.gcc@gmail.com>
14458 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14459 Change "crset eq" to "crset 2".
14460 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14461 (*call_indirect_aix<mode>_nospec): Likewise.
14462 (*call_value_indirect_aix<mode>_nospec): Likewise.
14463 (*call_indirect_elfv2<mode>_nospec): Likewise.
14464 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14465 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14466 change assembly output from . to $.
14467 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14468 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14469 (*tablejump<mode>_internal1_nospec): Likewise.
14471 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14474 * config/sh/sh_optimize_sett_clrt.cc:
14475 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14477 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14479 PR tree-optimization/83940
14480 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14481 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14482 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14483 use it to pass back the definition type.
14484 (vect_check_store_rhs): Likewise rhs_dt_out.
14485 (vect_build_gather_load_calls): Add a mask_dt argument and use
14486 it instead of a call to vect_is_simple_use.
14487 (vectorizable_store): Update calls to vect_check_load_store_mask
14488 and vect_check_store_rhs. Use the dt returned by the latter instead
14489 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14490 instead of calls to vect_is_simple_use. Pass the scalar rather
14491 than the vector operand to vect_is_simple_use when handling
14492 second and subsequent copies of an rhs value.
14493 (vectorizable_load): Update calls to vect_check_load_store_mask
14494 and vect_build_gather_load_calls. Use the cached mask_dt and
14495 gs_info.offset_dt instead of calls to vect_is_simple_use.
14497 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14499 PR middle-end/83945
14500 * tree-emutls.c: Include gimplify.h.
14501 (lower_emutls_2): New function.
14502 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14503 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14504 it before further processing.
14507 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14508 UINTVAL (trueop1) instead of INTVAL (op1).
14510 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14514 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14515 INCOMING_FRAME_SP_OFFSET if not defined.
14516 (scan_trace): Add ENTRY argument. If true and
14517 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14518 emit a note to adjust the CFA offset.
14519 (create_cfi_notes): Adjust scan_trace callers.
14520 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14521 INCOMING_FRAME_SP_OFFSET in the CIE.
14522 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14523 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14525 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14526 * doc/tm.texi: Regenerated.
14528 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14530 PR rtl-optimization/83147
14531 * lra-constraints.c (remove_inheritance_pseudos): Use
14532 lra_substitute_pseudo_within_insn.
14534 2018-01-19 Tom de Vries <tom@codesourcery.com>
14535 Cesar Philippidis <cesar@codesourcery.com>
14538 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14540 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14543 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14544 spaces for function labels.
14546 2018-01-19 Martin Liska <mliska@suse.cz>
14548 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14549 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14550 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14551 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14552 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14553 (PRED_CONST_RETURN): Change from 69 to 65.
14554 (PRED_NULL_RETURN): Change from 91 to 71.
14555 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14556 (PRED_LOOP_GUARD): Change from 66 to 73.
14558 2018-01-19 Martin Liska <mliska@suse.cz>
14560 * predict.c (predict_insn_def): Add new assert.
14561 (struct branch_predictor): Change type to signed integer.
14562 (test_prediction_value_range): Amend test to cover
14563 PROB_UNINITIALIZED.
14564 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14565 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14566 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14567 (PRED_LOOP_IV_COMPARE): Likewise.
14568 * predict.h (PROB_UNINITIALIZED): Define new constant.
14570 2018-01-19 Martin Liska <mliska@suse.cz>
14572 * predict.c (dump_prediction): Add new format for
14573 analyze_brprob.py script which is enabled with -details
14575 * profile-count.h (precise_p): New function.
14577 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14579 PR tree-optimization/83922
14580 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14581 there are no statements that need masking.
14582 (vect_active_double_reduction_p): New function.
14583 (vect_analyze_loop_operations): Use it when handling phis that
14584 are not in the loop header.
14586 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14588 PR tree-optimization/83914
14589 * tree-vect-loop.c (vectorizable_induction): Don't convert
14590 init_expr or apply the peeling adjustment for inductions
14591 that are nested within the vectorized loop.
14593 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14595 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14598 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14602 * function.h (gimplify_parameters): Add gimple_seq * argument.
14603 * function.c: Include gimple.h and options.h.
14604 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14605 for the added local temporaries if needed.
14606 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14607 if there are any parameter cleanups, wrap whole body into a
14608 try/finally with the cleanups.
14610 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14613 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14614 Use GET_MODE_CLASS for scalar floating point.
14616 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14620 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14621 Fix call of call_cgraph_insertion_hooks.
14623 2018-01-18 Martin Sebor <msebor@redhat.com>
14625 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14627 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14630 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14633 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14636 * common.opt: (-ffile-prefix-map): New option.
14637 * opts.c (common_handle_option): Defer it.
14638 * opts-global.c (handle_common_deferred_options): Handle it.
14639 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14640 * file-prefix-map.h: New file.
14641 (remap_debug_filename, add_debug_prefix_map): ...here.
14642 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14643 * final.c (debug_prefix_map, add_debug_prefix_map
14644 remap_debug_filename): Move to...
14645 * file-prefix-map.c: New file.
14646 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14647 generalize, get rid of alloca(), use strrchr() instead of strchr().
14648 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14649 Implement in terms of add_prefix_map().
14650 (remap_macro_filename, remap_debug_filename): Implement in term of
14652 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14653 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14654 * dbxout.c: Include file-prefix-map.h.
14655 * varasm.c: Likewise.
14656 * vmsdbgout.c: Likewise.
14657 * xcoffout.c: Likewise.
14658 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14659 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14660 * doc/invoke.texi (-ffile-prefix-map): Document.
14661 (-fdebug-prefix-map): Update description.
14663 2018-01-18 Martin Liska <mliska@suse.cz>
14665 * config/i386/i386.c (indirect_thunk_name): Document that also
14667 (output_indirect_thunk): Document why both instructions
14668 (pause and lfence) are generated.
14670 2018-01-18 Richard Biener <rguenther@suse.de>
14672 PR tree-optimization/83887
14673 * graphite-scop-detection.c
14674 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14675 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14676 (scop_detection::merge_sese): Re-implement with a flood-fill
14677 algorithm that properly finds a SESE region if it exists.
14679 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14682 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14683 pointer_diff optimizations use view_convert instead of convert.
14685 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14687 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14688 Generate different code for -mno-speculate-indirect-jumps.
14689 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14690 (*call_indirect_aix<mode>): Disable for
14691 -mno-speculate-indirect-jumps.
14692 (*call_indirect_aix<mode>_nospec): New define_insn.
14693 (*call_value_indirect_aix<mode>): Disable for
14694 -mno-speculate-indirect-jumps.
14695 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14696 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14697 -mno-speculate-indirect-jumps.
14698 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14700 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14702 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14703 long double type, set the flags for noting the default long double
14704 type, even if we don't pass or return a long double type.
14706 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14709 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14712 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14714 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14715 support for merge[hl].
14716 (fold_mergehl_helper): New helper function.
14717 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14718 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14719 (altivec_vmrglw_direct): Add xxmrglw insn.
14721 2018-01-17 Andrew Waterman <andrew@sifive.com>
14723 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14724 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14726 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14729 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14730 call the lto_location_cache before reading the
14731 DECL_SOURCE_LOCATION of the types.
14733 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14734 Richard Sandiford <richard.sandiford@linaro.org>
14736 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14737 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14738 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14739 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14740 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14742 * config/aarch64/constraints.md (aarch64_movti_operand):
14744 * config/aarch64/predicates.md (Uti): Add new constraint.
14746 2018-01-17 Carl Love <cel@us.ibm.com>
14748 * config/rs6000/vsx.md (define_expand xl_len_r,
14749 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14750 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14752 (define_expand, define_insn): Move the shift left from the
14753 define_insn to the define_expand for lxvl and stxvl instructions.
14754 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14755 and XL_LEN_R definitions to PURE.
14757 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14759 * config/i386/i386.c (indirect_thunk_name): Declare regno
14760 as unsigned int. Compare regno with INVALID_REGNUM.
14761 (output_indirect_thunk): Ditto.
14762 (output_indirect_thunk_function): Ditto.
14763 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14764 in the call to output_indirect_thunk_function.
14766 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14768 PR middle-end/83884
14769 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14770 rather than the size of inner_type to determine the stack slot size
14771 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14773 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14776 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14779 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14781 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14782 endian Linux systems to optionally enable multilibs for selecting
14783 the long double type if the user configured an explicit type.
14784 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14785 have no long double multilibs if not defined.
14786 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14787 warn if the user used -mabi={ieee,ibm}longdouble and we built
14788 multilibs for long double.
14789 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14790 appropriate multilib option.
14791 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14793 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14794 for building long double multilibs.
14795 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14797 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14799 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14802 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14804 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14807 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14810 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14813 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14815 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14816 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14818 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14820 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14821 different rtl trees depending on TARGET_64BIT.
14822 (rs6000_gen_lvx): Likewise.
14824 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14826 * config/visium/visium.md (nop): Tweak comment.
14827 (hazard_nop): Likewise.
14829 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14831 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14832 -mspeculate-indirect-jumps.
14833 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14834 for -mno-speculate-indirect-jumps.
14835 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14836 (*call_value_indirect_elfv2<mode>): Disable for
14837 -mno-speculate-indirect-jumps.
14838 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14839 (indirect_jump): Emit different RTL for
14840 -mno-speculate-indirect-jumps.
14841 (*indirect_jump<mode>): Disable for
14842 -mno-speculate-indirect-jumps.
14843 (*indirect_jump<mode>_nospec): New define_insn.
14844 (tablejump): Emit different RTL for
14845 -mno-speculate-indirect-jumps.
14846 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14847 (tablejumpsi_nospec): New define_expand.
14848 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14849 (tablejumpdi_nospec): New define_expand.
14850 (*tablejump<mode>_internal1): Disable for
14851 -mno-speculate-indirect-jumps.
14852 (*tablejump<mode>_internal1_nospec): New define_insn.
14853 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14856 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14858 * caller-save.c (insert_save): Drop unnecessary parameter. All
14861 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14862 Richard Biener <rguenth@suse.de>
14865 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14866 return early, inline manually is_gimple_sizepos. Make sure if we
14867 call gimplify_expr we don't end up with a gimple constant.
14868 * tree.c (variably_modified_type_p): Don't return true for
14869 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14870 * gimplify.h (is_gimple_sizepos): Remove.
14872 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14874 PR tree-optimization/83857
14875 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14876 vectorizable_live_operation for pure SLP statements.
14877 (vectorizable_live_operation): Handle PHIs.
14879 2018-01-16 Richard Biener <rguenther@suse.de>
14881 PR tree-optimization/83867
14882 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14883 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14885 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14888 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14889 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14890 If off is not INTEGER_CST, issue a may not be aligned warning
14891 rather than isn't aligned. Use isn%'t rather than isn't.
14892 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14894 <case MULT_EXPR>: Improve the case when bottom and one of the
14895 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14896 operand, in that case check if the other operand is multiple of
14897 bottom divided by the INTEGER_CST operand.
14899 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14902 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14903 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14904 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14905 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14906 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14907 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14908 (pa_function_arg_size): New function.
14910 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14912 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14913 in a separate statement.
14915 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14917 PR tree-optimization/83847
14918 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14919 group gathers and scatters.
14921 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14923 PR rtl-optimization/86620
14924 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14926 PR rtl-optimization/83213
14927 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14928 to last if both are JUMP_INSNs.
14930 PR tree-optimization/83843
14931 * gimple-ssa-store-merging.c
14932 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14933 store_immediate_info for bswap/nop orig_stores.
14935 2018-01-15 Andrew Waterman <andrew@sifive.com>
14937 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14939 <UDIV>: Increase cost if !TARGET_DIV.
14941 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14943 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14944 (define_attr "cr_logical_3op"): New.
14945 (cceq_ior_compare): Adjust.
14946 (cceq_ior_compare_complement): Adjust.
14947 (*cceq_rev_compare): Adjust.
14948 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14949 (is_cracked_insn): Adjust.
14950 (insn_must_be_first_in_group): Adjust.
14951 * config/rs6000/40x.md: Adjust.
14952 * config/rs6000/440.md: Adjust.
14953 * config/rs6000/476.md: Adjust.
14954 * config/rs6000/601.md: Adjust.
14955 * config/rs6000/603.md: Adjust.
14956 * config/rs6000/6xx.md: Adjust.
14957 * config/rs6000/7450.md: Adjust.
14958 * config/rs6000/7xx.md: Adjust.
14959 * config/rs6000/8540.md: Adjust.
14960 * config/rs6000/cell.md: Adjust.
14961 * config/rs6000/e300c2c3.md: Adjust.
14962 * config/rs6000/e500mc.md: Adjust.
14963 * config/rs6000/e500mc64.md: Adjust.
14964 * config/rs6000/e5500.md: Adjust.
14965 * config/rs6000/e6500.md: Adjust.
14966 * config/rs6000/mpc.md: Adjust.
14967 * config/rs6000/power4.md: Adjust.
14968 * config/rs6000/power5.md: Adjust.
14969 * config/rs6000/power6.md: Adjust.
14970 * config/rs6000/power7.md: Adjust.
14971 * config/rs6000/power8.md: Adjust.
14972 * config/rs6000/power9.md: Adjust.
14973 * config/rs6000/rs64.md: Adjust.
14974 * config/rs6000/titan.md: Adjust.
14976 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14978 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14979 ix86_indirect_branch_register logic.
14981 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14983 * config/i386/constraints.md (Bs): Update
14984 ix86_indirect_branch_register check. Don't check
14985 ix86_indirect_branch_register with GOT_memory_operand.
14987 * config/i386/predicates.md (GOT_memory_operand): Don't check
14988 ix86_indirect_branch_register here.
14989 (GOT32_symbol_operand): Likewise.
14991 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14993 * config/i386/predicates.md (constant_call_address_operand):
14994 Rewrite ix86_indirect_branch_register logic.
14995 (sibcall_insn_operand): Likewise.
14997 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14999 * config/i386/constraints.md (Bs): Replace
15000 ix86_indirect_branch_thunk_register with
15001 ix86_indirect_branch_register.
15003 * config/i386/i386.md (indirect_jump): Likewise.
15004 (tablejump): Likewise.
15005 (*sibcall_memory): Likewise.
15006 (*sibcall_value_memory): Likewise.
15007 Peepholes of indirect call and jump via memory: Likewise.
15008 * config/i386/i386.opt: Likewise.
15009 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15010 (GOT_memory_operand): Likewise.
15011 (call_insn_operand): Likewise.
15012 (sibcall_insn_operand): Likewise.
15013 (GOT32_symbol_operand): Likewise.
15015 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15017 PR middle-end/83837
15018 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15019 type rather than type addr's type points to.
15020 (expand_omp_atomic_mutex): Likewise.
15021 (expand_omp_atomic): Likewise.
15023 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15026 * config/i386/i386.c (output_indirect_thunk_function): Use
15027 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15028 for __x86_return_thunk.
15030 2018-01-15 Richard Biener <rguenther@suse.de>
15032 PR middle-end/83850
15033 * expmed.c (extract_bit_field_1): Fix typo.
15035 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15038 * config/arm/iterators.md (VF): New mode iterator.
15039 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15040 Remove integer-related logic from pattern.
15041 (neon_vabd<mode>_3): Likewise.
15043 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15045 PR middle-end/82694
15046 * common.opt (fstrict-overflow): No longer an alias.
15047 (fwrapv-pointer): New option.
15048 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15049 also for pointer types based on flag_wrapv_pointer.
15050 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15051 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15052 opts->x_flag_wrapv got set.
15053 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15054 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15055 POINTER_TYPE_OVERFLOW_UNDEFINED.
15056 * match.pd: Likewise in address comparison pattern.
15057 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15059 2018-01-15 Richard Biener <rguenther@suse.de>
15062 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15063 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15064 Reset type names to their identifier if their TYPE_DECL doesn't
15065 have linkage (and thus is used for ODR and devirt).
15066 (save_debug_info_for_decl): Remove.
15067 (save_debug_info_for_type): Likewise.
15068 (add_tree_to_fld_list): Adjust.
15069 * tree-pretty-print.c (dump_generic_node): Make dumping of
15070 type names more robust.
15072 2018-01-15 Richard Biener <rguenther@suse.de>
15074 * BASE-VER: Bump to 8.0.1.
15076 2018-01-14 Martin Sebor <msebor@redhat.com>
15079 * builtins.c (check_access): Avoid warning when the no-warning bit
15082 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15084 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15085 * ira-color (allocno_hard_regs_compare): Likewise.
15087 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15090 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15091 Use .pushsection/.popsection.
15093 2018-01-14 Martin Sebor <msebor@redhat.com>
15096 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15098 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15100 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15101 entry from extra_headers.
15102 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15103 extra_headers, make the list bitwise identical to the i?86-*-* one.
15105 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15107 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15108 -mcmodel=large with -mindirect-branch=thunk,
15109 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15110 -mfunction-return=thunk-extern.
15111 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15112 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15113 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15115 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15117 * config/i386/i386.c (print_reg): Print the name of the full
15118 integer register without '%'.
15119 (ix86_print_operand): Handle 'V'.
15120 * doc/extend.texi: Document 'V' modifier.
15122 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15124 * config/i386/constraints.md (Bs): Disallow memory operand for
15125 -mindirect-branch-register.
15127 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15128 (GOT_memory_operand): Likewise.
15129 (call_insn_operand): Likewise.
15130 (sibcall_insn_operand): Likewise.
15131 (GOT32_symbol_operand): Likewise.
15132 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15133 for -mindirect-branch-register.
15134 (tablejump): Likewise.
15135 (*sibcall_memory): Likewise.
15136 (*sibcall_value_memory): Likewise.
15137 Disallow peepholes of indirect call and jump via memory for
15138 -mindirect-branch-register.
15139 (*call_pop): Replace m with Bw.
15140 (*call_value_pop): Likewise.
15141 (*sibcall_pop_memory): Replace m with Bs.
15142 * config/i386/i386.opt (mindirect-branch-register): New option.
15143 * doc/invoke.texi: Document -mindirect-branch-register option.
15145 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15147 * config/i386/i386-protos.h (ix86_output_function_return): New.
15148 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15149 set function_return_type.
15150 (indirect_thunk_name): Add ret_p to indicate thunk for function
15152 (output_indirect_thunk_function): Pass false to
15153 indirect_thunk_name.
15154 (ix86_output_indirect_branch_via_reg): Likewise.
15155 (ix86_output_indirect_branch_via_push): Likewise.
15156 (output_indirect_thunk_function): Create alias for function
15157 return thunk if regno < 0.
15158 (ix86_output_function_return): New function.
15159 (ix86_handle_fndecl_attribute): Handle function_return.
15160 (ix86_attribute_table): Add function_return.
15161 * config/i386/i386.h (machine_function): Add
15162 function_return_type.
15163 * config/i386/i386.md (simple_return_internal): Use
15164 ix86_output_function_return.
15165 (simple_return_internal_long): Likewise.
15166 * config/i386/i386.opt (mfunction-return=): New option.
15167 (indirect_branch): Mention -mfunction-return=.
15168 * doc/extend.texi: Document function_return function attribute.
15169 * doc/invoke.texi: Document -mfunction-return= option.
15171 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15173 * config/i386/i386-opts.h (indirect_branch): New.
15174 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15175 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15176 with local indirect jump when converting indirect call and jump.
15177 (ix86_set_indirect_branch_type): New.
15178 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15179 (indirectlabelno): New.
15180 (indirect_thunk_needed): Likewise.
15181 (indirect_thunk_bnd_needed): Likewise.
15182 (indirect_thunks_used): Likewise.
15183 (indirect_thunks_bnd_used): Likewise.
15184 (INDIRECT_LABEL): Likewise.
15185 (indirect_thunk_name): Likewise.
15186 (output_indirect_thunk): Likewise.
15187 (output_indirect_thunk_function): Likewise.
15188 (ix86_output_indirect_branch_via_reg): Likewise.
15189 (ix86_output_indirect_branch_via_push): Likewise.
15190 (ix86_output_indirect_branch): Likewise.
15191 (ix86_output_indirect_jmp): Likewise.
15192 (ix86_code_end): Call output_indirect_thunk_function if needed.
15193 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15195 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15196 (ix86_attribute_table): Add indirect_branch.
15197 * config/i386/i386.h (machine_function): Add indirect_branch_type
15198 and has_local_indirect_jump.
15199 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15201 (tablejump): Likewise.
15202 (*indirect_jump): Use ix86_output_indirect_jmp.
15203 (*tablejump_1): Likewise.
15204 (simple_return_indirect_internal): Likewise.
15205 * config/i386/i386.opt (mindirect-branch=): New option.
15206 (indirect_branch): New.
15209 (thunk-inline): Likewise.
15210 (thunk-extern): Likewise.
15211 * doc/extend.texi: Document indirect_branch function attribute.
15212 * doc/invoke.texi: Document -mindirect-branch= option.
15214 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15217 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15219 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15221 * ipa-inline.c (want_inline_small_function_p): Return false if
15222 inlining has already failed with CIF_FINAL_ERROR.
15223 (update_caller_keys): Call want_inline_small_function_p before
15225 (update_callee_keys): Likewise.
15227 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15229 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15231 (rs6000_quadword_masked_address_p): Likewise.
15232 (quad_aligned_load_p): Likewise.
15233 (quad_aligned_store_p): Likewise.
15234 (const_load_sequence_p): Add comment to describe the outer-most loop.
15235 (mimic_memory_attributes_and_flags): New function.
15236 (rs6000_gen_stvx): Likewise.
15237 (replace_swapped_aligned_store): Likewise.
15238 (rs6000_gen_lvx): Likewise.
15239 (replace_swapped_aligned_load): Likewise.
15240 (replace_swapped_load_constant): Capitalize argument name in
15241 comment describing this function.
15242 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15243 and stores that access quad-word aligned addresses and replace
15244 with stvx or lvx instructions when appropriate.
15245 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15246 New function prototype.
15247 (rs6000_quadword_masked_address_p): Likewise.
15248 (rs6000_gen_lvx): Likewise.
15249 (rs6000_gen_stvx): Likewise.
15250 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15251 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15252 when memory address is aligned.
15253 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15254 this split to select lvx instruction when memory address is aligned.
15255 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15256 instruction when memory address is aligned.
15257 (*vsx_le_perm_load_v16qi): Likewise.
15258 (four unnamed splitters): Modify to select the stvx instruction
15259 when memory is aligned.
15261 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15263 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15264 which appears in the queue multiple times.
15266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15267 Alan Hayward <alan.hayward@arm.com>
15268 David Sherwood <david.sherwood@arm.com>
15270 * tree-vectorizer.h (vec_lower_bound): New structure.
15271 (_loop_vec_info): Add check_nonzero and lower_bounds.
15272 (LOOP_VINFO_CHECK_NONZERO): New macro.
15273 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15274 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15275 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15276 fields. Make seg_len the distance travelled, not including the
15278 (dr_direction_indicator): Declare.
15279 (dr_zero_step_indicator): Likewise.
15280 (dr_known_forward_stride_p): Likewise.
15281 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15283 (runtime_alias_check_p): Allow runtime alias checks with
15285 (operator ==): Compare access_size and align.
15286 (prune_runtime_alias_test_list): Rework for new distinction between
15287 the access_size and seg_len.
15288 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15290 (get_segment_min_max): New function.
15291 (create_intersect_range_checks): Use it.
15292 (dr_step_indicator): New function.
15293 (dr_direction_indicator): Likewise.
15294 (dr_zero_step_indicator): Likewise.
15295 (dr_known_forward_stride_p): Likewise.
15296 * tree-loop-distribution.c (data_ref_segment_size): Return
15297 DR_STEP * (niters - 1).
15298 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15300 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15301 (vect_preserves_scalar_order_p): New function, split out from...
15302 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15303 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15304 (vect_vfa_access_size): New function.
15305 (vect_vfa_align): Likewise.
15306 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15307 (dump_lower_bound): New function.
15308 (vect_check_lower_bound): Likewise.
15309 (vect_small_gap_p): Likewise.
15310 (vectorizable_with_step_bound_p): Likewise.
15311 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15312 depencies if the vectorization factor is 1. Convert the checks
15313 for nonzero steps into checks on the bounds of DR_STEP. Try using
15314 a bunds check for variable steps if the minimum required step is
15315 relatively small. Update calls to the dr_with_seg_len
15316 constructor and to vect_compile_time_alias.
15317 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15319 (vect_loop_versioning): Call it.
15320 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15322 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15325 Alan Hayward <alan.hayward@arm.com>
15326 David Sherwood <david.sherwood@arm.com>
15328 * doc/sourcebuild.texi (vect_scatter_store): Document.
15329 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15331 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15333 * genopinit.c (main): Add supports_vec_scatter_store and
15334 supports_vec_scatter_store_cached to target_optabs.
15335 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15336 IFN_MASK_SCATTER_STORE.
15337 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15339 * internal-fn.h (internal_store_fn_p): Declare.
15340 (internal_fn_stored_value_index): Likewise.
15341 * internal-fn.c (scatter_store_direct): New macro.
15342 (expand_scatter_store_optab_fn): New function.
15343 (direct_scatter_store_optab_supported_p): New macro.
15344 (internal_store_fn_p): New function.
15345 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15346 IFN_MASK_SCATTER_STORE.
15347 (internal_fn_mask_index): Likewise.
15348 (internal_fn_stored_value_index): New function.
15349 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15350 for scatter stores.
15351 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15352 * optabs-query.c (supports_vec_scatter_store_p): New function.
15353 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15354 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15355 true for scatter stores.
15356 (vect_gather_scatter_fn_p): Handle scatter stores too.
15357 (vect_check_gather_scatter): Consider using scatter stores if
15358 supports_vec_scatter_store_p.
15359 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15360 scatter stores too.
15361 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15362 internal_fn_stored_value_index.
15363 (check_load_store_masking): Handle scatter stores too.
15364 (vect_get_store_rhs): Make public.
15365 (vectorizable_call): Use internal_store_fn_p.
15366 (vectorizable_store): Handle scatter store internal functions.
15367 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15368 when deciding whether the end of the group has been reached.
15369 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15370 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15371 (mask_scatter_store<mode>): New insns.
15373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15374 Alan Hayward <alan.hayward@arm.com>
15375 David Sherwood <david.sherwood@arm.com>
15377 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15378 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15379 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15381 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15382 Use vect_truncate_gather_scatter_offset if we can't treat the
15383 operation as a normal gather load or scatter store.
15384 (get_group_load_store_type): Take the gather_scatter_info
15385 as argument. Try using a gather load or scatter store for
15386 single-element groups.
15387 (get_load_store_type): Update calls to get_group_load_store_type
15388 and vect_use_strided_gather_scatters_p.
15390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15391 Alan Hayward <alan.hayward@arm.com>
15392 David Sherwood <david.sherwood@arm.com>
15394 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15395 optional tree argument.
15396 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15398 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15399 but continue to use the current value as a fallback.
15400 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15401 to compare the updates.
15402 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15403 (get_load_store_type): Use it when handling a strided access.
15404 (vect_get_strided_load_store_ops): New function.
15405 (vect_get_data_ptr_increment): Likewise.
15406 (vectorizable_load): Handle strided gather loads. Always pass
15407 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15410 Alan Hayward <alan.hayward@arm.com>
15411 David Sherwood <david.sherwood@arm.com>
15413 * doc/md.texi (gather_load@var{m}): Document.
15414 (mask_gather_load@var{m}): Likewise.
15415 * genopinit.c (main): Add supports_vec_gather_load and
15416 supports_vec_gather_load_cached to target_optabs.
15417 * optabs-tree.c (init_tree_optimization_optabs): Use
15418 ggc_cleared_alloc to allocate target_optabs.
15419 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15420 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15422 * internal-fn.h (internal_load_fn_p): Declare.
15423 (internal_gather_scatter_fn_p): Likewise.
15424 (internal_fn_mask_index): Likewise.
15425 (internal_gather_scatter_fn_supported_p): Likewise.
15426 * internal-fn.c (gather_load_direct): New macro.
15427 (expand_gather_load_optab_fn): New function.
15428 (direct_gather_load_optab_supported_p): New macro.
15429 (direct_internal_fn_optab): New function.
15430 (internal_load_fn_p): Likewise.
15431 (internal_gather_scatter_fn_p): Likewise.
15432 (internal_fn_mask_index): Likewise.
15433 (internal_gather_scatter_fn_supported_p): Likewise.
15434 * optabs-query.c (supports_at_least_one_mode_p): New function.
15435 (supports_vec_gather_load_p): Likewise.
15436 * optabs-query.h (supports_vec_gather_load_p): Declare.
15437 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15438 and memory_type field.
15439 (NUM_PATTERNS): Bump to 15.
15440 * tree-vect-data-refs.c: Include internal-fn.h.
15441 (vect_gather_scatter_fn_p): New function.
15442 (vect_describe_gather_scatter_call): Likewise.
15443 (vect_check_gather_scatter): Try using internal functions for
15444 gather loads. Recognize existing calls to a gather load function.
15445 (vect_analyze_data_refs): Consider using gather loads if
15446 supports_vec_gather_load_p.
15447 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15448 (vect_get_gather_scatter_offset_type): Likewise.
15449 (vect_convert_mask_for_vectype): Likewise.
15450 (vect_add_conversion_to_patterm): Likewise.
15451 (vect_try_gather_scatter_pattern): Likewise.
15452 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15453 (vect_vect_recog_func_ptrs): Add it.
15454 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15455 internal_fn_mask_index and internal_gather_scatter_fn_p.
15456 (check_load_store_masking): Take the gather_scatter_info as an
15457 argument and handle gather loads.
15458 (vect_get_gather_scatter_ops): New function.
15459 (vectorizable_call): Check internal_load_fn_p.
15460 (vectorizable_load): Likewise. Handle gather load internal
15462 (vectorizable_store): Update call to check_load_store_masking.
15463 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15464 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15465 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15466 (aarch64_gather_scale_operand_d): New predicates.
15467 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15468 (mask_gather_load<mode>): New insns.
15470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15471 Alan Hayward <alan.hayward@arm.com>
15472 David Sherwood <david.sherwood@arm.com>
15474 * optabs.def (fold_left_plus_optab): New optab.
15475 * doc/md.texi (fold_left_plus_@var{m}): Document.
15476 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15477 * internal-fn.c (fold_left_direct): Define.
15478 (expand_fold_left_optab_fn): Likewise.
15479 (direct_fold_left_optab_supported_p): Likewise.
15480 * fold-const-call.c (fold_const_fold_left): New function.
15481 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15482 * tree-parloops.c (valid_reduction_p): New function.
15483 (gather_scalar_reductions): Use it.
15484 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15485 (vect_finish_replace_stmt): Declare.
15486 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15487 (needs_fold_left_reduction_p): New function, split out from...
15488 (vect_is_simple_reduction): ...here. Accept reductions that
15489 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15490 (vect_force_simple_reduction): Also store the reduction type in
15491 the assignment's STMT_VINFO_REDUC_TYPE.
15492 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15493 (merge_with_identity): New function.
15494 (vect_expand_fold_left): Likewise.
15495 (vectorize_fold_left_reduction): Likewise.
15496 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15497 scalar phi in place for it. Check for target support and reject
15498 cases that would reassociate the operation. Defer the transform
15499 phase to vectorize_fold_left_reduction.
15500 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15501 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15502 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15506 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15507 call to ifc_temp_var.
15509 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15510 Alan Hayward <alan.hayward@arm.com>
15511 David Sherwood <david.sherwood@arm.com>
15513 * target.def (legitimize_address_displacement): Take the original
15514 offset as a poly_int.
15515 * targhooks.h (default_legitimize_address_displacement): Update
15517 * targhooks.c (default_legitimize_address_displacement): Likewise.
15518 * doc/tm.texi: Regenerate.
15519 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15520 as an argument, moving assert of ad->disp == ad->disp_term to...
15521 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15522 Try calling targetm.legitimize_address_displacement before expanding
15523 the address rather than afterwards, and adjust for the new interface.
15524 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15525 Match the new hook interface. Handle SVE addresses.
15526 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15527 new hook interface.
15529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15531 * Makefile.in (OBJS): Add early-remat.o.
15532 * target.def (select_early_remat_modes): New hook.
15533 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15534 * doc/tm.texi: Regenerate.
15535 * targhooks.h (default_select_early_remat_modes): Declare.
15536 * targhooks.c (default_select_early_remat_modes): New function.
15537 * timevar.def (TV_EARLY_REMAT): New timevar.
15538 * passes.def (pass_early_remat): New pass.
15539 * tree-pass.h (make_pass_early_remat): Declare.
15540 * early-remat.c: New file.
15541 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15543 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15546 Alan Hayward <alan.hayward@arm.com>
15547 David Sherwood <david.sherwood@arm.com>
15549 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15550 vfm1 with a bound_epilog parameter.
15551 (vect_do_peeling): Update calls accordingly, and move the prologue
15552 call earlier in the function. Treat the base bound_epilog as 0 for
15553 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15554 this base when peeling for gaps.
15555 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15556 with fully-masked loops.
15557 (vect_estimate_min_profitable_iters): Handle the single peeled
15558 iteration in that case.
15560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15561 Alan Hayward <alan.hayward@arm.com>
15562 David Sherwood <david.sherwood@arm.com>
15564 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15565 single-element interleaving even if the size is not a power of 2.
15566 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15567 accesses for single-element interleaving if the group size is
15570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15571 Alan Hayward <alan.hayward@arm.com>
15572 David Sherwood <david.sherwood@arm.com>
15574 * doc/md.texi (fold_extract_last_@var{m}): Document.
15575 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15576 * optabs.def (fold_extract_last_optab): New optab.
15577 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15578 * internal-fn.c (fold_extract_direct): New macro.
15579 (expand_fold_extract_optab_fn): Likewise.
15580 (direct_fold_extract_optab_supported_p): Likewise.
15581 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15582 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15583 EXTRACT_LAST_REDUCTION.
15584 (get_initial_def_for_reduction): Do not create an initial vector
15585 for EXTRACT_LAST_REDUCTION reductions.
15586 (vectorizable_reduction): Leave the scalar phi in place for
15587 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15588 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15589 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15590 transform phase to vectorizable_condition.
15591 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15593 (vect_finish_stmt_generation): ...here.
15594 (vect_finish_replace_stmt): New function.
15595 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15596 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15598 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 (extract_last_@var{m}): Document.
15605 * optabs.def (extract_last_optab): New optab.
15606 * internal-fn.def (EXTRACT_LAST): New internal function.
15607 * internal-fn.c (cond_unary_direct): New macro.
15608 (expand_cond_unary_optab_fn): Likewise.
15609 (direct_cond_unary_optab_supported_p): Likewise.
15610 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15611 loops using EXTRACT_LAST.
15612 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15613 (extract_last_<mode>): ...this optab.
15614 (vec_extract<mode><Vel>): Update accordingly.
15616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15617 Alan Hayward <alan.hayward@arm.com>
15618 David Sherwood <david.sherwood@arm.com>
15620 * target.def (empty_mask_is_expensive): New hook.
15621 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15622 * doc/tm.texi: Regenerate.
15623 * targhooks.h (default_empty_mask_is_expensive): Declare.
15624 * targhooks.c (default_empty_mask_is_expensive): New function.
15625 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15626 if the target says that empty masks are expensive.
15627 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15629 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15632 Alan Hayward <alan.hayward@arm.com>
15633 David Sherwood <david.sherwood@arm.com>
15635 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15636 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15637 (vect_use_loop_mask_for_alignment_p): New function.
15638 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15639 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15640 niters_skip argument. Make sure that the first niters_skip elements
15641 of the first iteration are inactive.
15642 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15643 Update call to vect_set_loop_masks_directly.
15644 (get_misalign_in_elems): New function, split out from...
15645 (vect_gen_prolog_loop_niters): ...here.
15646 (vect_update_init_of_dr): Take a code argument that specifies whether
15647 the adjustment should be added or subtracted.
15648 (vect_update_init_of_drs): Likewise.
15649 (vect_prepare_for_masked_peels): New function.
15650 (vect_do_peeling): Skip prologue peeling if we're using a mask
15651 instead. Update call to vect_update_inits_of_drs.
15652 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15654 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15655 alignment. Do not include the number of peeled iterations in
15656 the minimum threshold in that case.
15657 (vectorizable_induction): Adjust the start value down by
15658 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15659 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15660 Take the number of skipped iterations into account when calculating
15662 * tree-vect-stmts.c (vect_gen_while_not): New function.
15664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15665 Alan Hayward <alan.hayward@arm.com>
15666 David Sherwood <david.sherwood@arm.com>
15668 * doc/sourcebuild.texi (vect_fully_masked): Document.
15669 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15670 default value to 0.
15671 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15673 (vect_analyze_loop_2): ...here. Don't check the vectorization
15674 factor against the number of loop iterations if the loop is
15677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15678 Alan Hayward <alan.hayward@arm.com>
15679 David Sherwood <david.sherwood@arm.com>
15681 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15682 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15683 (dump_groups): Update accordingly.
15684 (iv_use::mem_type): New member variable.
15685 (address_p): New function.
15686 (record_use): Add a mem_type argument and initialize the new
15688 (record_group_use): Add a mem_type argument. Use address_p.
15689 Remove obsolete null checks of base_object. Update call to record_use.
15690 (find_interesting_uses_op): Update call to record_group_use.
15691 (find_interesting_uses_cond): Likewise.
15692 (find_interesting_uses_address): Likewise.
15693 (get_mem_type_for_internal_fn): New function.
15694 (find_address_like_use): Likewise.
15695 (find_interesting_uses_stmt): Try find_address_like_use before
15696 calling find_interesting_uses_op.
15697 (addr_offset_valid_p): Use the iv mem_type field as the type
15698 of the addressed memory.
15699 (add_autoinc_candidates): Likewise.
15700 (get_address_cost): Likewise.
15701 (split_small_address_groups_p): Use address_p.
15702 (split_address_groups): Likewise.
15703 (add_iv_candidate_for_use): Likewise.
15704 (autoinc_possible_for_pair): Likewise.
15705 (rewrite_groups): Likewise.
15706 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15707 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15708 (get_alias_ptr_type_for_ptr_address): New function.
15709 (rewrite_use_address): Rewrite address uses in calls that were
15710 identified by find_address_like_use.
15712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15713 Alan Hayward <alan.hayward@arm.com>
15714 David Sherwood <david.sherwood@arm.com>
15716 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15718 * gimple-expr.h (is_gimple_addressable: Likewise.
15719 * gimple-expr.c (is_gimple_address): Likewise.
15720 * internal-fn.c (expand_call_mem_ref): New function.
15721 (expand_mask_load_optab_fn): Use it.
15722 (expand_mask_store_optab_fn): Likewise.
15724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15725 Alan Hayward <alan.hayward@arm.com>
15726 David Sherwood <david.sherwood@arm.com>
15728 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15729 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15730 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15731 (cond_umax@var{mode}): Document.
15732 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15733 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15734 (cond_umin_optab, cond_umax_optab): New optabs.
15735 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15736 (COND_IOR, COND_XOR): New internal functions.
15737 * internal-fn.h (get_conditional_internal_fn): Declare.
15738 * internal-fn.c (cond_binary_direct): New macro.
15739 (expand_cond_binary_optab_fn): Likewise.
15740 (direct_cond_binary_optab_supported_p): Likewise.
15741 (get_conditional_internal_fn): New function.
15742 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15743 Cope with reduction statements that are vectorized as calls rather
15745 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15746 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15747 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15748 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15749 (UNSPEC_COND_EOR): New unspecs.
15750 (optab): Add mappings for them.
15751 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15752 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
15759 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15760 * internal-fn.def (WHILE_ULT): New internal function.
15761 * internal-fn.h (direct_internal_fn_supported_p): New override
15762 that takes two types as argument.
15763 * internal-fn.c (while_direct): New macro.
15764 (expand_while_optab_fn): New function.
15765 (convert_optab_supported_p): Likewise.
15766 (direct_while_optab_supported_p): New macro.
15767 * wide-int.h (wi::udiv_ceil): New function.
15768 * tree-vectorizer.h (rgroup_masks): New structure.
15769 (vec_loop_masks): New typedef.
15770 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15771 and fully_masked_p.
15772 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15773 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15774 (vect_max_vf): New function.
15775 (slpeel_make_loop_iterate_ntimes): Delete.
15776 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15777 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15778 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15779 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15780 internal-fn.h, stor-layout.h and optabs-query.h.
15781 (vect_set_loop_mask): New function.
15782 (add_preheader_seq): Likewise.
15783 (add_header_seq): Likewise.
15784 (interleave_supported_p): Likewise.
15785 (vect_maybe_permute_loop_masks): Likewise.
15786 (vect_set_loop_masks_directly): Likewise.
15787 (vect_set_loop_condition_masked): Likewise.
15788 (vect_set_loop_condition_unmasked): New function, split out from
15789 slpeel_make_loop_iterate_ntimes.
15790 (slpeel_make_loop_iterate_ntimes): Rename to..
15791 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15792 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15793 (vect_do_peeling): Update call accordingly.
15794 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15796 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15797 mask_compare_type, can_fully_mask_p and fully_masked_p.
15798 (release_vec_loop_masks): New function.
15799 (_loop_vec_info): Use it to free the loop masks.
15800 (can_produce_all_loop_masks_p): New function.
15801 (vect_get_max_nscalars_per_iter): Likewise.
15802 (vect_verify_full_masking): Likewise.
15803 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15804 retries, and free the mask rgroups before retrying. Check loop-wide
15805 reasons for disallowing fully-masked loops. Make the final decision
15806 about whether use a fully-masked loop or not.
15807 (vect_estimate_min_profitable_iters): Do not assume that peeling
15808 for the number of iterations will be needed for fully-masked loops.
15809 (vectorizable_reduction): Disable fully-masked loops.
15810 (vectorizable_live_operation): Likewise.
15811 (vect_halve_mask_nunits): New function.
15812 (vect_double_mask_nunits): Likewise.
15813 (vect_record_loop_mask): Likewise.
15814 (vect_get_loop_mask): Likewise.
15815 (vect_transform_loop): Handle the case in which the final loop
15816 iteration might handle a partial vector. Call vect_set_loop_condition
15817 instead of slpeel_make_loop_iterate_ntimes.
15818 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15819 (check_load_store_masking): New function.
15820 (prepare_load_store_mask): Likewise.
15821 (vectorizable_store): Handle fully-masked loops.
15822 (vectorizable_load): Likewise.
15823 (supportable_widening_operation): Use vect_halve_mask_nunits for
15825 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15826 (vect_gen_while): New function.
15827 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15828 (aarch64_uqdec<mode>): New insn.
15830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15831 Alan Hayward <alan.hayward@arm.com>
15832 David Sherwood <david.sherwood@arm.com>
15834 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15835 (reduc_xor_scal_optab): New optabs.
15836 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15837 (reduc_xor_scal_@var{m}): Document.
15838 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15839 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15840 internal functions.
15841 * fold-const-call.c (fold_const_call): Handle them.
15842 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15843 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15844 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15845 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15846 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15847 (UNSPEC_XORV): New unspecs.
15848 (optab): Add entries for them.
15849 (BITWISEV): New int iterator.
15850 (bit_reduc_op): New int attributes.
15852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15853 Alan Hayward <alan.hayward@arm.com>
15854 David Sherwood <david.sherwood@arm.com>
15856 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15857 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15858 * optabs.def (vec_shl_insert_optab): New optab.
15859 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15860 (duplicate_and_interleave): Likewise.
15861 * tree-vect-loop.c: Include internal-fn.h.
15862 (neutral_op_for_slp_reduction): New function, split out from
15863 get_initial_defs_for_reduction.
15864 (get_initial_def_for_reduction): Handle option 2 for variable-length
15865 vectors by loading the neutral value into a vector and then shifting
15866 the initial value into element 0.
15867 (get_initial_defs_for_reduction): Replace the code argument with
15868 the neutral value calculated by neutral_op_for_slp_reduction.
15869 Use gimple_build_vector for constant-length vectors.
15870 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15871 but the first group_size elements have a neutral value.
15872 Use duplicate_and_interleave otherwise.
15873 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15874 Update call to get_initial_defs_for_reduction. Handle SLP
15875 reductions for variable-length vectors by creating one vector
15876 result for each scalar result, with the elements associated
15877 with other scalar results stubbed out with the neutral value.
15878 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15879 Require IFN_VEC_SHL_INSERT for double reductions on
15880 variable-length vectors, or SLP reductions that have
15881 a neutral value. Require can_duplicate_and_interleave_p
15882 support for variable-length unchained SLP reductions if there
15883 is no neutral value, such as for MIN/MAX reductions. Also require
15884 the number of vector elements to be a multiple of the number of
15885 SLP statements when doing variable-length unchained SLP reductions.
15886 Update call to vect_create_epilog_for_reduction.
15887 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15888 and remove initial values.
15889 (duplicate_and_interleave): Make public.
15890 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15891 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15894 Alan Hayward <alan.hayward@arm.com>
15895 David Sherwood <david.sherwood@arm.com>
15897 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15898 (can_duplicate_and_interleave_p): New function.
15899 (vect_get_and_check_slp_defs): Take the vector of statements
15900 rather than just the current one. Remove excess parentheses.
15901 Restriction rejectinon of vect_constant_def and vect_external_def
15902 for variable-length vectors to boolean types, or types for which
15903 can_duplicate_and_interleave_p is false.
15904 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15905 (duplicate_and_interleave): New function.
15906 (vect_get_constant_vectors): Use gimple_build_vector for
15907 constant-length vectors and suitable variable-length constant
15908 vectors. Use duplicate_and_interleave for other variable-length
15909 vectors. Don't defer the update when inserting new statements.
15911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15912 Alan Hayward <alan.hayward@arm.com>
15913 David Sherwood <david.sherwood@arm.com>
15915 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15916 min_profitable_iters doesn't go negative.
15918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15919 Alan Hayward <alan.hayward@arm.com>
15920 David Sherwood <david.sherwood@arm.com>
15922 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15923 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15924 * optabs.def (vec_mask_load_lanes_optab): New optab.
15925 (vec_mask_store_lanes_optab): Likewise.
15926 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15927 (MASK_STORE_LANES): Likewise.
15928 * internal-fn.c (mask_load_lanes_direct): New macro.
15929 (mask_store_lanes_direct): Likewise.
15930 (expand_mask_load_optab_fn): Handle masked operations.
15931 (expand_mask_load_lanes_optab_fn): New macro.
15932 (expand_mask_store_optab_fn): Handle masked operations.
15933 (expand_mask_store_lanes_optab_fn): New macro.
15934 (direct_mask_load_lanes_optab_supported_p): Likewise.
15935 (direct_mask_store_lanes_optab_supported_p): Likewise.
15936 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15938 (vect_load_lanes_supported): Likewise.
15939 * tree-vect-data-refs.c (strip_conversion): New function.
15940 (can_group_stmts_p): Likewise.
15941 (vect_analyze_data_ref_accesses): Use it instead of checking
15942 for a pair of assignments.
15943 (vect_store_lanes_supported): Take a masked_p parameter.
15944 (vect_load_lanes_supported): Likewise.
15945 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15946 vect_store_lanes_supported and vect_load_lanes_supported.
15947 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15948 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15949 parameter. Don't allow gaps for masked accesses.
15950 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15951 and vect_load_lanes_supported.
15952 (get_load_store_type): Take a masked_p parameter and update
15953 call to get_group_load_store_type.
15954 (vectorizable_store): Update call to get_load_store_type.
15955 Handle IFN_MASK_STORE_LANES.
15956 (vectorizable_load): Update call to get_load_store_type.
15957 Handle IFN_MASK_LOAD_LANES.
15959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15960 Alan Hayward <alan.hayward@arm.com>
15961 David Sherwood <david.sherwood@arm.com>
15963 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15965 * config/aarch64/aarch64-protos.h
15966 (aarch64_sve_struct_memory_operand_p): Declare.
15967 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15968 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15969 (VPRED, vpred): Handle SVE structure modes.
15970 * config/aarch64/constraints.md (Utx): New constraint.
15971 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15972 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15973 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15974 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15975 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15976 structure modes. Split into pieces after RA.
15977 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15978 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15980 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15981 SVE structure modes.
15982 (aarch64_classify_address): Likewise.
15983 (sizetochar): Move earlier in file.
15984 (aarch64_print_operand): Handle SVE register lists.
15985 (aarch64_array_mode): New function.
15986 (aarch64_sve_struct_memory_operand_p): Likewise.
15987 (TARGET_ARRAY_MODE): Redefine.
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 * target.def (array_mode): New target hook.
15994 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15995 * doc/tm.texi: Regenerate.
15996 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15997 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15998 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15999 targetm.array_mode.
16000 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16004 Alan Hayward <alan.hayward@arm.com>
16005 David Sherwood <david.sherwood@arm.com>
16007 * fold-const.c (fold_binary_loc): Check the argument types
16008 rather than the result type when testing for a vector operation.
16010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16012 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16013 * doc/tm.texi: Regenerate.
16015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16016 Alan Hayward <alan.hayward@arm.com>
16017 David Sherwood <david.sherwood@arm.com>
16019 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16020 (sve): Document new AArch64 extension.
16021 * doc/md.texi (w): Extend the description of the AArch64
16022 constraint to include SVE vectors.
16023 (Upl, Upa): Document new AArch64 predicate constraints.
16024 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16026 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16027 (msve-vector-bits=): New option.
16028 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16029 SVE when these are disabled.
16030 (sve): New extension.
16031 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16032 modes. Adjust their number of units based on aarch64_sve_vg.
16033 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16034 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16035 aarch64_addr_query_type.
16036 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16037 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16038 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16039 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16040 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16041 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16042 (aarch64_simd_imm_zero_p): Delete.
16043 (aarch64_check_zero_based_sve_index_immediate): Declare.
16044 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16045 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16046 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16047 (aarch64_sve_float_mul_immediate_p): Likewise.
16048 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16049 rather than an rtx.
16050 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16051 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16052 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16053 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16054 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16055 (aarch64_regmode_natural_size): Likewise.
16056 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16057 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16059 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16060 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16061 for VG and the SVE predicate registers.
16062 (V_ALIASES): Add a "z"-prefixed alias.
16063 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16064 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16065 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16066 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16067 (REG_CLASS_NAMES): Add entries for them.
16068 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16069 and the predicate registers.
16070 (aarch64_sve_vg): Declare.
16071 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16072 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16073 (REGMODE_NATURAL_SIZE): Define.
16074 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16076 * config/aarch64/aarch64.c: Include cfgrtl.h.
16077 (simd_immediate_info): Add a constructor for series vectors,
16078 and an associated step field.
16079 (aarch64_sve_vg): New variable.
16080 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16081 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16082 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16083 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16084 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16085 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16086 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16087 (aarch64_get_mask_mode): New functions.
16088 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16089 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16090 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16091 predicate modes and predicate registers. Explicitly restrict
16092 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16093 to store a vector mode if it is recognized by
16094 aarch64_classify_vector_mode.
16095 (aarch64_regmode_natural_size): New function.
16096 (aarch64_hard_regno_caller_save_mode): Return the original mode
16098 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16099 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16100 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16101 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16103 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16104 does not overlap dest if the function is frame-related. Handle
16106 (aarch64_split_add_offset): New function.
16107 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16108 them aarch64_add_offset.
16109 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16110 and update call to aarch64_sub_sp.
16111 (aarch64_add_cfa_expression): New function.
16112 (aarch64_expand_prologue): Pass extra temporary registers to the
16113 functions above. Handle the case in which we need to emit new
16114 DW_CFA_expressions for registers that were originally saved
16115 relative to the stack pointer, but now have to be expressed
16116 relative to the frame pointer.
16117 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16119 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16120 IP0 and IP1 values for SVE frames.
16121 (aarch64_expand_vec_series): New function.
16122 (aarch64_expand_sve_widened_duplicate): Likewise.
16123 (aarch64_expand_sve_const_vector): Likewise.
16124 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16125 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16126 into the register, rather than emitting a SET directly.
16127 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16128 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16129 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16130 (offset_9bit_signed_scaled_p): New functions.
16131 (aarch64_replicate_bitmask_imm): New function.
16132 (aarch64_bitmask_imm): Use it.
16133 (aarch64_cannot_force_const_mem): Reject expressions involving
16134 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16135 (aarch64_classify_index): Handle SVE indices, by requiring
16136 a plain register index with a scale that matches the element size.
16137 (aarch64_classify_address): Handle SVE addresses. Assert that
16138 the mode of the address is VOIDmode or an integer mode.
16139 Update call to aarch64_classify_symbol.
16140 (aarch64_classify_symbolic_expression): Update call to
16141 aarch64_classify_symbol.
16142 (aarch64_const_vec_all_in_range_p): New function.
16143 (aarch64_print_vector_float_operand): Likewise.
16144 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16145 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16146 and the FP immediates 1.0 and 0.5.
16147 (aarch64_print_address_internal): Handle SVE addresses.
16148 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16149 (aarch64_regno_regclass): Handle predicate registers.
16150 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16152 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16153 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16154 (aarch64_convert_sve_vector_bits): New function.
16155 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16156 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16157 rather than an rtx.
16158 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16159 Handle SVE vector and predicate modes. Accept VL-based constants
16160 that need only one temporary register, and VL offsets that require
16161 no temporary registers.
16162 (aarch64_conditional_register_usage): Mark the predicate registers
16163 as fixed if SVE isn't available.
16164 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16165 Return true for SVE vector and predicate modes.
16166 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16167 rather than an unsigned int. Handle SVE modes.
16168 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16170 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16172 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16173 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16174 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16175 (aarch64_sve_float_mul_immediate_p): New functions.
16176 (aarch64_sve_valid_immediate): New function.
16177 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16178 Explicitly reject structure modes. Check for INDEX constants.
16179 Handle PTRUE and PFALSE constants.
16180 (aarch64_check_zero_based_sve_index_immediate): New function.
16181 (aarch64_simd_imm_zero_p): Delete.
16182 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16183 vector modes. Accept constants in the range of CNT[BHWD].
16184 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16185 ask for an Advanced SIMD mode.
16186 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16187 (aarch64_simd_vector_alignment): Handle SVE predicates.
16188 (aarch64_vectorize_preferred_vector_alignment): New function.
16189 (aarch64_simd_vector_alignment_reachable): Use it instead of
16191 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16192 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16194 (MAX_VECT_LEN): Delete.
16195 (expand_vec_perm_d): Add a vec_flags field.
16196 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16197 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16198 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16200 (aarch64_evpc_rev): Rename to...
16201 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16202 (aarch64_evpc_rev_global): New function.
16203 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16204 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16206 (aarch64_evpc_sve_tbl): New function.
16207 (aarch64_expand_vec_perm_const_1): Update after rename of
16208 aarch64_evpc_rev. Handle SVE permutes too, trying
16209 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16210 than aarch64_evpc_tbl.
16211 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16212 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16213 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16214 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16215 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16216 (aarch64_expand_sve_vcond): New functions.
16217 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16218 of aarch64_vector_mode_p.
16219 (aarch64_dwarf_poly_indeterminate_value): New function.
16220 (aarch64_compute_pressure_classes): Likewise.
16221 (aarch64_can_change_mode_class): Likewise.
16222 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16223 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16224 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16225 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16226 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16227 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16228 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16229 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16231 (Dn, Dl, Dr): Accept const as well as const_vector.
16232 (Dz): Likewise. Compare against CONST0_RTX.
16233 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16234 of "vector" where appropriate.
16235 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16236 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16237 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16238 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16239 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16240 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16241 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16242 (v_int_equiv): Extend to SVE modes.
16243 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16245 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16246 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16247 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16248 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16249 (SVE_COND_FP_CMP): New int iterators.
16250 (perm_hilo): Handle the new unpack unspecs.
16251 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16253 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16254 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16255 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16256 (aarch64_equality_operator, aarch64_constant_vector_operand)
16257 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16258 (aarch64_sve_nonimmediate_operand): Likewise.
16259 (aarch64_sve_general_operand): Likewise.
16260 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16261 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16262 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16263 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16264 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16265 (aarch64_sve_float_arith_immediate): Likewise.
16266 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16267 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16268 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16269 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16270 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16271 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16272 (aarch64_sve_float_arith_operand): Likewise.
16273 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16274 (aarch64_sve_float_mul_operand): Likewise.
16275 (aarch64_sve_vec_perm_operand): Likewise.
16276 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16277 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16278 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16279 as well as const_vector.
16280 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16281 in file. Use CONST0_RTX and CONSTM1_RTX.
16282 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16283 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16284 Use aarch64_simd_imm_zero.
16285 * config/aarch64/aarch64-sve.md: New file.
16286 * config/aarch64/aarch64.md: Include it.
16287 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16288 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16289 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16290 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16291 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16292 (sve): New attribute.
16293 (enabled): Disable instructions with the sve attribute unless
16295 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16296 aarch64_expand_mov_immediate.
16297 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16298 CNT[BHSD] immediates.
16299 (movti): Split CONST_POLY_INT moves into two halves.
16300 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16301 Split additions that need a temporary here if the destination
16302 is the stack pointer.
16303 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16304 (*add<mode>3_poly_1): New instruction.
16305 (set_clobber_cc): New expander.
16307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16309 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16310 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16311 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16312 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16313 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16314 Change innermode from fixed_mode_size to machine_mode.
16315 (simplify_subreg): Update call accordingly. Handle a constant-sized
16316 subreg of a variable-length CONST_VECTOR.
16318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16319 Alan Hayward <alan.hayward@arm.com>
16320 David Sherwood <david.sherwood@arm.com>
16322 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16323 (add_offset_to_base): New function, split out from...
16324 (create_mem_ref): ...here. When handling a scale other than 1,
16325 check first whether the address is valid without the offset.
16326 Add it into the base if so, leaving the index and scale as-is.
16328 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16331 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16332 fold_for_warn before checking if arg2 is INTEGER_CST.
16334 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16336 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16337 (store_multiple_operation): Delete.
16338 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16339 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16340 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16341 guarded by TARGET_STRING.
16342 (rs6000_output_load_multiple): Delete.
16343 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16344 OPTION_MASK_STRING / TARGET_STRING handling.
16345 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16346 (const rs6000_opt_masks) <"string">: Change mask to 0.
16347 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16348 (MASK_STRING): Delete.
16349 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16351 (load_multiple): Delete.
16358 (store_multiple): Delete.
16365 (movmemsi_8reg): Delete.
16366 (corresponding unnamed define_insn): Delete.
16367 (movmemsi_6reg): Delete.
16368 (corresponding unnamed define_insn): Delete.
16369 (movmemsi_4reg): Delete.
16370 (corresponding unnamed define_insn): Delete.
16371 (movmemsi_2reg): Delete.
16372 (corresponding unnamed define_insn): Delete.
16373 (movmemsi_1reg): Delete.
16374 (corresponding unnamed define_insn): Delete.
16375 * config/rs6000/rs6000.opt (mno-string): New.
16376 (mstring): Replace by deprecation warning stub.
16377 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16379 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16381 * regrename.c (regrename_do_replace): If replacing the same
16382 reg multiple times, try to reuse last created gen_raw_REG.
16385 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16386 main to workaround a bug in GDB.
16388 2018-01-12 Tom de Vries <tom@codesourcery.com>
16391 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16393 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16395 PR rtl-optimization/80481
16396 * ira-color.c (get_cap_member): New function.
16397 (allocnos_conflict_by_live_ranges_p): Use it.
16398 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16399 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16401 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16404 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16405 (*saddl_se_1): Ditto.
16406 (*ssubsi_1): Ditto.
16407 (*ssubl_se_1): Ditto.
16409 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16411 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16412 rather than wi::to_widest for DR_INITs.
16413 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16414 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16415 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16417 (vect_analyze_group_access_1): Note that here.
16419 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16421 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16422 polynomial type sizes.
16424 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16426 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16427 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16428 (gimple_add_tmp_var): Likewise.
16430 2018-01-12 Martin Liska <mliska@suse.cz>
16432 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16433 (gimple_alloc_sizes): Likewise.
16434 (dump_gimple_statistics): Use PRIu64 in printf format.
16435 * gimple.h: Change uint64_t to int.
16437 2018-01-12 Martin Liska <mliska@suse.cz>
16439 * tree-core.h: Use uint64_t instead of int.
16440 * tree.c (tree_node_counts): Likewise.
16441 (tree_node_sizes): Likewise.
16442 (dump_tree_statistics): Use PRIu64 in printf format.
16444 2018-01-12 Martin Liska <mliska@suse.cz>
16446 * Makefile.in: As qsort_chk is implemented in vec.c, add
16447 vec.o to linkage of gencfn-macros.
16448 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16449 passing the info to record_node_allocation_statistics.
16450 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16452 * ggc-common.c (struct ggc_usage): Add operator== and use
16453 it in operator< and compare function.
16454 * mem-stats.h (struct mem_usage): Likewise.
16455 * vec.c (struct vec_usage): Remove operator< and compare
16456 function. Can be simply inherited.
16458 2018-01-12 Martin Jambor <mjambor@suse.cz>
16461 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16462 * tree-ssa-math-opts.c: Include domwalk.h.
16463 (convert_mult_to_fma_1): New function.
16464 (fma_transformation_info): New type.
16465 (fma_deferring_state): Likewise.
16466 (cancel_fma_deferring): New function.
16467 (result_of_phi): Likewise.
16468 (last_fma_candidate_feeds_initial_phi): Likewise.
16469 (convert_mult_to_fma): Added deferring logic, split actual
16470 transformation to convert_mult_to_fma_1.
16471 (math_opts_dom_walker): New type.
16472 (math_opts_dom_walker::after_dom_children): New method, body moved
16473 here from pass_optimize_widening_mul::execute, added deferring logic
16475 (pass_optimize_widening_mul::execute): Moved most of code to
16476 math_opts_dom_walker::after_dom_children.
16477 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16478 * config/i386/i386.c (ix86_option_override_internal): Added
16479 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16481 2018-01-12 Richard Biener <rguenther@suse.de>
16484 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16485 inline instance vars.
16487 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16490 * config/rx/rx.c (rx_is_restricted_memory_address):
16491 Handle SUBREG case.
16493 2018-01-12 Richard Biener <rguenther@suse.de>
16495 PR tree-optimization/80846
16496 * target.def (split_reduction): New target hook.
16497 * targhooks.c (default_split_reduction): New function.
16498 * targhooks.h (default_split_reduction): Declare.
16499 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16500 target requests first reduce vectors by combining low and high
16502 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16503 (get_vectype_for_scalar_type_and_size): Export.
16504 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16505 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16506 * doc/tm.texi: Regenerate.
16507 * config/i386/i386.c (ix86_split_reduction): Implement
16508 TARGET_VECTORIZE_SPLIT_REDUCTION.
16510 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16513 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16514 in PIC mode except for TARGET_VXWORKS_RTP.
16515 * config/sparc/sparc.c: Include cfgrtl.h.
16516 (TARGET_INIT_PIC_REG): Define.
16517 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16518 (sparc_pic_register_p): New predicate.
16519 (sparc_legitimate_address_p): Use it.
16520 (sparc_legitimize_pic_address): Likewise.
16521 (sparc_delegitimize_address): Likewise.
16522 (sparc_mode_dependent_address_p): Likewise.
16523 (gen_load_pcrel_sym): Remove 4th parameter.
16524 (load_got_register): Adjust call to above. Remove obsolete stuff.
16525 (sparc_expand_prologue): Do not call load_got_register here.
16526 (sparc_flat_expand_prologue): Likewise.
16527 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16528 (sparc_use_pseudo_pic_reg): New function.
16529 (sparc_init_pic_reg): Likewise.
16530 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16531 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16533 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16535 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16536 Add item for branch_cost.
16538 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16540 PR rtl-optimization/83565
16541 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16542 not extend the result to a larger mode for rotate operations.
16543 (num_sign_bit_copies1): Likewise.
16545 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16548 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16550 Use values-Xc.o for -pedantic.
16551 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16553 2018-01-12 Martin Liska <mliska@suse.cz>
16556 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16558 (possible_polymorphic_call_targets): Use it.
16559 (ipa_devirt): Likewise.
16561 2018-01-12 Martin Liska <mliska@suse.cz>
16563 * profile-count.h (enum profile_quality): Use 0 as invalid
16564 enum value of profile_quality.
16566 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16568 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16569 -mext-string options.
16571 2018-01-12 Richard Biener <rguenther@suse.de>
16573 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16574 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16575 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16577 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16579 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16581 * configure.ac (--with-long-double-format): Add support for the
16582 configuration option to change the default long double format on
16584 * config.gcc (powerpc*-linux*-*): Likewise.
16585 * configure: Regenerate.
16586 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16587 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16588 used without modification.
16590 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16592 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16593 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16594 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16595 MISC_BUILTIN_SPEC_BARRIER.
16596 (rs6000_init_builtins): Likewise.
16597 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16599 (speculation_barrier): New define_insn.
16600 * doc/extend.texi: Document __builtin_speculation_barrier.
16602 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16605 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16606 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16607 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16609 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16610 integral modes instead of "ss" and "sd".
16611 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16612 vectors with 32-bit and 64-bit elements.
16613 (vecdupssescalarmodesuffix): New mode attribute.
16614 (vec_dup<mode>): Use it.
16616 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16619 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16620 frame if argument is passed on stack.
16622 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16625 * ree.c (combine_reaching_defs): Optimize also
16626 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16627 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16629 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16631 PR middle-end/83189
16632 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16634 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16636 PR middle-end/83718
16637 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16638 after they are computed.
16640 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16642 PR tree-optimization/83695
16643 * gimple-loop-linterchange.cc
16644 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16645 reset cached scev information after interchange.
16646 (pass_linterchange::execute): Remove call to scev_reset_htab.
16648 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16650 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16651 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16652 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16653 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16654 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16655 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16656 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16657 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16658 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16659 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16660 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16661 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16662 (V_lane_reg): Likewise.
16663 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16665 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16666 (vfmal_lane_low<mode>_intrinsic,
16667 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16668 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16669 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16670 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16671 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16672 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16674 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16676 * config/arm/arm-cpus.in (fp16fml): New feature.
16677 (ALL_SIMD): Add fp16fml.
16678 (armv8.2-a): Add fp16fml as an option.
16679 (armv8.3-a): Likewise.
16680 (armv8.4-a): Add fp16fml as part of fp16.
16681 * config/arm/arm.h (TARGET_FP16FML): Define.
16682 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16684 * config/arm/arm-modes.def (V2HF): Define.
16685 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16686 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16687 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16688 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16689 vfmsl_low, vfmsl_high): New set of builtins.
16690 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16691 (vfml_op): New code attribute.
16692 (VFMLHALVES): New int iterator.
16693 (VFML, VFMLSEL): New mode attributes.
16694 (V_reg): Define mapping for V2HF.
16695 (V_hi, V_lo): New mode attributes.
16696 (VF_constraint): Likewise.
16697 (vfml_half, vfml_half_selector): New int attributes.
16698 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16700 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16701 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16703 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16704 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16705 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16706 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16708 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16709 Document new effective target and option set.
16711 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16713 * config/arm/arm-cpus.in (armv8_4): New feature.
16714 (ARMv8_4a): New fgroup.
16715 (armv8.4-a): New arch.
16716 * config/arm/arm-tables.opt: Regenerate.
16717 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16718 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16719 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16720 Add matching rules for -march=armv8.4-a and extensions.
16721 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16723 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16726 * config/rx/rx.md (BW): New mode attribute.
16727 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16729 2018-01-11 Richard Biener <rguenther@suse.de>
16731 PR tree-optimization/83435
16732 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16733 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16734 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16736 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16737 Alan Hayward <alan.hayward@arm.com>
16738 David Sherwood <david.sherwood@arm.com>
16740 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16742 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16743 (aarch64_print_address_internal): Use it to check for a zero offset.
16745 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16746 Alan Hayward <alan.hayward@arm.com>
16747 David Sherwood <david.sherwood@arm.com>
16749 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16750 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16751 Return a poly_int64 rather than a HOST_WIDE_INT.
16752 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16753 rather than a HOST_WIDE_INT.
16754 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16755 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16756 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16757 final_offset from HOST_WIDE_INT to poly_int64.
16758 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16759 to_constant when getting the number of units in an Advanced SIMD
16761 (aarch64_builtin_vectorized_function): Check for a constant number
16763 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16765 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16766 attribute instead of GET_MODE_NUNITS.
16767 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16768 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16769 GET_MODE_SIZE for fixed-size registers.
16770 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16771 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16772 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16773 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16774 (aarch64_print_operand, aarch64_print_address_internal)
16775 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16776 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16777 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16778 Handle polynomial GET_MODE_SIZE.
16779 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16780 wider than SImode without modification.
16781 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16782 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16783 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16784 passing and returning SVE modes.
16785 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16786 rather than GEN_INT.
16787 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16788 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16789 (aarch64_allocate_and_probe_stack_space): Likewise.
16790 (aarch64_layout_frame): Cope with polynomial offsets.
16791 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16792 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16793 polynomial offsets.
16794 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16795 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16796 poly_int64 rather than a HOST_WIDE_INT.
16797 (aarch64_get_separate_components, aarch64_process_components)
16798 (aarch64_expand_prologue, aarch64_expand_epilogue)
16799 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16800 (aarch64_anchor_offset): New function, split out from...
16801 (aarch64_legitimize_address): ...here.
16802 (aarch64_builtin_vectorization_cost): Handle polynomial
16803 TYPE_VECTOR_SUBPARTS.
16804 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16806 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16807 number of elements from the PARALLEL rather than the mode.
16808 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16809 rather than GET_MODE_BITSIZE.
16810 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16811 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16812 (aarch64_expand_vec_perm_const_1): Handle polynomial
16813 d->perm.length () and d->perm elements.
16814 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16815 Apply to_constant to d->perm elements.
16816 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16817 polynomial CONST_VECTOR_NUNITS.
16818 (aarch64_move_pointer): Take amount as a poly_int64 rather
16820 (aarch64_progress_pointer): Avoid temporary variable.
16821 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16822 the mode attribute instead of GET_MODE.
16824 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16825 Alan Hayward <alan.hayward@arm.com>
16826 David Sherwood <david.sherwood@arm.com>
16828 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16829 x exists before using it.
16830 (aarch64_add_constant_internal): Rename to...
16831 (aarch64_add_offset_1): ...this. Replace regnum with separate
16832 src and dest rtxes. Handle the case in which they're different,
16833 including when the offset is zero. Replace scratchreg with an rtx.
16834 Use 2 additions if there is no spare register into which we can
16835 move a 16-bit constant.
16836 (aarch64_add_constant): Delete.
16837 (aarch64_add_offset): Replace reg with separate src and dest
16838 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16839 Use aarch64_add_offset_1.
16840 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16841 an rtx rather than an int. Take the delta as a poly_int64
16842 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16843 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16844 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16845 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16846 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16847 and aarch64_add_sp.
16848 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16849 aarch64_add_constant.
16851 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16853 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16854 Use scalar_float_mode.
16856 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16858 * config/aarch64/aarch64-simd.md
16859 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16860 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16861 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16862 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16863 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16864 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16865 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16866 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16867 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16868 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16870 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16873 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16874 targ_options->x_arm_arch_string is non NULL.
16876 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16878 * config/aarch64/aarch64.h
16879 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16881 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16884 * expmed.c (emit_store_flag_force): Swap if const op0
16885 and change VOIDmode to mode of op0.
16887 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16889 PR rtl-optimization/83761
16890 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16891 than bytes to mode_for_size.
16893 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16895 PR middle-end/83189
16896 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16897 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16900 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16902 PR middle-end/83575
16903 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16904 when in layout mode.
16905 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16906 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16909 2018-01-10 Michael Collison <michael.collison@arm.com>
16911 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16912 * config/aarch64/aarch64-option-extension.def: Add
16913 AARCH64_OPT_EXTENSION of 'fp16fml'.
16914 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16915 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16916 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16917 * config/aarch64/constraints.md (Ui7): New constraint.
16918 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16919 (VFMLA_SEL_W): Ditto.
16922 (VFMLA16_LOW): New int iterator.
16923 (VFMLA16_HIGH): Ditto.
16924 (UNSPEC_FMLAL): New unspec.
16925 (UNSPEC_FMLSL): Ditto.
16926 (UNSPEC_FMLAL2): Ditto.
16927 (UNSPEC_FMLSL2): Ditto.
16928 (f16mac): New code attribute.
16929 * config/aarch64/aarch64-simd-builtins.def
16930 (aarch64_fmlal_lowv2sf): Ditto.
16931 (aarch64_fmlsl_lowv2sf): Ditto.
16932 (aarch64_fmlalq_lowv4sf): Ditto.
16933 (aarch64_fmlslq_lowv4sf): Ditto.
16934 (aarch64_fmlal_highv2sf): Ditto.
16935 (aarch64_fmlsl_highv2sf): Ditto.
16936 (aarch64_fmlalq_highv4sf): Ditto.
16937 (aarch64_fmlslq_highv4sf): Ditto.
16938 (aarch64_fmlal_lane_lowv2sf): Ditto.
16939 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16940 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16941 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16942 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16943 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16944 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16945 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16946 (aarch64_fmlal_lane_highv2sf): Ditto.
16947 (aarch64_fmlsl_lane_highv2sf): Ditto.
16948 (aarch64_fmlal_laneq_highv2sf): Ditto.
16949 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16950 (aarch64_fmlalq_lane_highv4sf): Ditto.
16951 (aarch64_fmlsl_lane_highv4sf): Ditto.
16952 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16953 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16954 * config/aarch64/aarch64-simd.md:
16955 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16956 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16957 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16958 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16959 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16960 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16961 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16962 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16963 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16964 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16965 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16966 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16967 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16968 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16969 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16970 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16971 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16972 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16973 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16974 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16975 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16976 (vfmlsl_low_u32): Ditto.
16977 (vfmlalq_low_u32): Ditto.
16978 (vfmlslq_low_u32): Ditto.
16979 (vfmlal_high_u32): Ditto.
16980 (vfmlsl_high_u32): Ditto.
16981 (vfmlalq_high_u32): Ditto.
16982 (vfmlslq_high_u32): Ditto.
16983 (vfmlal_lane_low_u32): Ditto.
16984 (vfmlsl_lane_low_u32): Ditto.
16985 (vfmlal_laneq_low_u32): Ditto.
16986 (vfmlsl_laneq_low_u32): Ditto.
16987 (vfmlalq_lane_low_u32): Ditto.
16988 (vfmlslq_lane_low_u32): Ditto.
16989 (vfmlalq_laneq_low_u32): Ditto.
16990 (vfmlslq_laneq_low_u32): Ditto.
16991 (vfmlal_lane_high_u32): Ditto.
16992 (vfmlsl_lane_high_u32): Ditto.
16993 (vfmlal_laneq_high_u32): Ditto.
16994 (vfmlsl_laneq_high_u32): Ditto.
16995 (vfmlalq_lane_high_u32): Ditto.
16996 (vfmlslq_lane_high_u32): Ditto.
16997 (vfmlalq_laneq_high_u32): Ditto.
16998 (vfmlslq_laneq_high_u32): Ditto.
16999 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17000 (AARCH64_FL_FOR_ARCH8_4): New.
17001 (AARCH64_ISA_F16FML): New ISA flag.
17002 (TARGET_F16FML): New feature flag for fp16fml.
17003 (doc/invoke.texi): Document new fp16fml option.
17005 2018-01-10 Michael Collison <michael.collison@arm.com>
17007 * config/aarch64/aarch64-builtins.c:
17008 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17009 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17010 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17011 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17012 (AARCH64_ISA_SHA3): New ISA flag.
17013 (TARGET_SHA3): New feature flag for sha3.
17014 * config/aarch64/iterators.md (sha512_op): New int attribute.
17015 (CRYPTO_SHA512): New int iterator.
17016 (UNSPEC_SHA512H): New unspec.
17017 (UNSPEC_SHA512H2): Ditto.
17018 (UNSPEC_SHA512SU0): Ditto.
17019 (UNSPEC_SHA512SU1): Ditto.
17020 * config/aarch64/aarch64-simd-builtins.def
17021 (aarch64_crypto_sha512hqv2di): New builtin.
17022 (aarch64_crypto_sha512h2qv2di): Ditto.
17023 (aarch64_crypto_sha512su0qv2di): Ditto.
17024 (aarch64_crypto_sha512su1qv2di): Ditto.
17025 (aarch64_eor3qv8hi): Ditto.
17026 (aarch64_rax1qv2di): Ditto.
17027 (aarch64_xarqv2di): Ditto.
17028 (aarch64_bcaxqv8hi): Ditto.
17029 * config/aarch64/aarch64-simd.md:
17030 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17031 (aarch64_crypto_sha512su0qv2di): Ditto.
17032 (aarch64_crypto_sha512su1qv2di): Ditto.
17033 (aarch64_eor3qv8hi): Ditto.
17034 (aarch64_rax1qv2di): Ditto.
17035 (aarch64_xarqv2di): Ditto.
17036 (aarch64_bcaxqv8hi): Ditto.
17037 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17038 (vsha512h2q_u64): Ditto.
17039 (vsha512su0q_u64): Ditto.
17040 (vsha512su1q_u64): Ditto.
17041 (veor3q_u16): Ditto.
17042 (vrax1q_u64): Ditto.
17043 (vxarq_u64): Ditto.
17044 (vbcaxq_u16): Ditto.
17045 * config/arm/types.md (crypto_sha512): New type attribute.
17046 (crypto_sha3): Ditto.
17047 (doc/invoke.texi): Document new sha3 option.
17049 2018-01-10 Michael Collison <michael.collison@arm.com>
17051 * config/aarch64/aarch64-builtins.c:
17052 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17053 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17054 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17055 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17056 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17057 (AARCH64_ISA_SM4): New ISA flag.
17058 (TARGET_SM4): New feature flag for sm4.
17059 * config/aarch64/aarch64-simd-builtins.def
17060 (aarch64_sm3ss1qv4si): Ditto.
17061 (aarch64_sm3tt1aq4si): Ditto.
17062 (aarch64_sm3tt1bq4si): Ditto.
17063 (aarch64_sm3tt2aq4si): Ditto.
17064 (aarch64_sm3tt2bq4si): Ditto.
17065 (aarch64_sm3partw1qv4si): Ditto.
17066 (aarch64_sm3partw2qv4si): Ditto.
17067 (aarch64_sm4eqv4si): Ditto.
17068 (aarch64_sm4ekeyqv4si): Ditto.
17069 * config/aarch64/aarch64-simd.md:
17070 (aarch64_sm3ss1qv4si): Ditto.
17071 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17072 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17073 (aarch64_sm4eqv4si): Ditto.
17074 (aarch64_sm4ekeyqv4si): Ditto.
17075 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17076 (sm3part_op): Ditto.
17077 (CRYPTO_SM3TT): Ditto.
17078 (CRYPTO_SM3PART): Ditto.
17079 (UNSPEC_SM3SS1): New unspec.
17080 (UNSPEC_SM3TT1A): Ditto.
17081 (UNSPEC_SM3TT1B): Ditto.
17082 (UNSPEC_SM3TT2A): Ditto.
17083 (UNSPEC_SM3TT2B): Ditto.
17084 (UNSPEC_SM3PARTW1): Ditto.
17085 (UNSPEC_SM3PARTW2): Ditto.
17086 (UNSPEC_SM4E): Ditto.
17087 (UNSPEC_SM4EKEY): Ditto.
17088 * config/aarch64/constraints.md (Ui2): New constraint.
17089 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17090 * config/arm/types.md (crypto_sm3): New type attribute.
17091 (crypto_sm4): Ditto.
17092 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17093 (vsm3tt1aq_u32): Ditto.
17094 (vsm3tt1bq_u32): Ditto.
17095 (vsm3tt2aq_u32): Ditto.
17096 (vsm3tt2bq_u32): Ditto.
17097 (vsm3partw1q_u32): Ditto.
17098 (vsm3partw2q_u32): Ditto.
17099 (vsm4eq_u32): Ditto.
17100 (vsm4ekeyq_u32): Ditto.
17101 (doc/invoke.texi): Document new sm4 option.
17103 2018-01-10 Michael Collison <michael.collison@arm.com>
17105 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17106 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17107 (AARCH64_FL_FOR_ARCH8_4): New.
17108 (AARCH64_FL_V8_4): New flag.
17109 (doc/invoke.texi): Document new armv8.4-a option.
17111 2018-01-10 Michael Collison <michael.collison@arm.com>
17113 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17114 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17115 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17116 * config/aarch64/aarch64-option-extension.def: Add
17117 AARCH64_OPT_EXTENSION of 'sha2'.
17118 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17119 (crypto): Disable sha2 and aes if crypto disabled.
17120 (crypto): Enable aes and sha2 if enabled.
17121 (simd): Disable sha2 and aes if simd disabled.
17122 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17124 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17125 (TARGET_SHA2): New feature flag for sha2.
17126 (TARGET_AES): New feature flag for aes.
17127 * config/aarch64/aarch64-simd.md:
17128 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17129 conditional on TARGET_AES.
17130 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17131 (aarch64_crypto_sha1hsi): Make pattern conditional
17133 (aarch64_crypto_sha1hv4si): Ditto.
17134 (aarch64_be_crypto_sha1hv4si): Ditto.
17135 (aarch64_crypto_sha1su1v4si): Ditto.
17136 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17137 (aarch64_crypto_sha1su0v4si): Ditto.
17138 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17139 (aarch64_crypto_sha256su0v4si): Ditto.
17140 (aarch64_crypto_sha256su1v4si): Ditto.
17141 (doc/invoke.texi): Document new aes and sha2 options.
17143 2018-01-10 Martin Sebor <msebor@redhat.com>
17145 PR tree-optimization/83781
17146 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17149 2018-01-11 Martin Sebor <msebor@gmail.com>
17150 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17152 PR tree-optimization/83501
17153 PR tree-optimization/81703
17155 * tree-ssa-strlen.c (get_string_cst): Rename...
17156 (get_string_len): ...to this. Handle global constants.
17157 (handle_char_store): Adjust.
17159 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17160 Jim Wilson <jimw@sifive.com>
17162 * config/riscv/riscv-protos.h (riscv_output_return): New.
17163 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17164 (riscv_attribute_table, riscv_output_return),
17165 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17166 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17167 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17169 (riscv_expand_prologue): Add early return for naked function.
17170 (riscv_expand_epilogue): Likewise.
17171 (riscv_function_ok_for_sibcall): Return false for naked function.
17172 (riscv_set_current_function): New.
17173 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17174 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17175 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17176 * doc/extend.texi (RISC-V Function Attributes): New.
17178 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17180 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17181 check for 128-bit long double before checking TCmode.
17182 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17183 128-bit long doubles before checking TFmode or TCmode.
17184 (FLOAT128_IBM_P): Likewise.
17186 2018-01-10 Martin Sebor <msebor@redhat.com>
17188 PR tree-optimization/83671
17189 * builtins.c (c_strlen): Unconditionally return zero for the empty
17191 Use -Warray-bounds for warnings.
17192 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17193 for non-constant array indices with COMPONENT_REF, arrays of
17194 arrays, and pointers to arrays.
17195 (gimple_fold_builtin_strlen): Determine and set length range for
17196 non-constant character arrays.
17198 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17200 PR middle-end/81897
17201 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17204 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17206 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17208 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17211 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17212 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17213 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17214 indexed_or_indirect_operand predicate.
17215 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17216 (*vsx_le_perm_load_v8hi): Likewise.
17217 (*vsx_le_perm_load_v16qi): Likewise.
17218 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17219 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17220 (*vsx_le_perm_store_v8hi): Likewise.
17221 (*vsx_le_perm_store_v16qi): Likewise.
17222 (eight unnamed splitters): Likewise.
17224 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17226 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17227 * config/rs6000/emmintrin.h: Likewise.
17228 * config/rs6000/mmintrin.h: Likewise.
17229 * config/rs6000/xmmintrin.h: Likewise.
17231 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17234 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17236 * tree.c (tree_nop_conversion): Return true for location wrapper
17238 (maybe_wrap_with_location): New function.
17239 (selftest::check_strip_nops): New function.
17240 (selftest::test_location_wrappers): New function.
17241 (selftest::tree_c_tests): Call it.
17242 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17243 (maybe_wrap_with_location): New decl.
17244 (EXPR_LOCATION_WRAPPER_P): New macro.
17245 (location_wrapper_p): New inline function.
17246 (tree_strip_any_location_wrapper): New inline function.
17248 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17251 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17252 stack_realign_offset for the largest alignment of stack slot
17254 (ix86_find_max_used_stack_alignment): New function.
17255 (ix86_finalize_stack_frame_flags): Use it. Set
17256 max_used_stack_alignment if we don't realign stack.
17257 * config/i386/i386.h (machine_function): Add
17258 max_used_stack_alignment.
17260 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17262 * config/arm/arm.opt (-mbranch-cost): New option.
17263 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17266 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17269 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17270 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17272 2018-01-10 Richard Biener <rguenther@suse.de>
17275 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17276 early out so it also covers the case where we have a non-NULL
17279 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17281 PR tree-optimization/83753
17282 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17283 for non-strided grouped accesses if the number of elements is 1.
17285 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17288 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17289 * i386.h (TARGET_USE_GATHER): Define.
17290 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17292 2018-01-10 Martin Liska <mliska@suse.cz>
17295 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17296 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17298 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17299 CLEANUP_NO_PARTITIONING is not set.
17301 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17303 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17304 for vectors, as a partial revert of r254296.
17305 * rtl.h (const_vec_p): Delete.
17306 (const_vec_duplicate_p): Don't test for vector CONSTs.
17307 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17308 * expmed.c (make_tree): Likewise.
17311 * common.md (E, F): Use CONSTANT_P instead of checking for
17313 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17314 checking for CONST_VECTOR.
17316 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17318 PR middle-end/83575
17319 * predict.c (force_edge_cold): Handle in more sane way edges
17320 with no prediction.
17322 2018-01-09 Carl Love <cel@us.ibm.com>
17324 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17326 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17327 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17328 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17329 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17330 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17331 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17332 * config/rs6000/rs6000-protos.h: Add extern defition for
17333 rs6000_generate_float2_double_code.
17334 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17336 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17337 (float2_v2df): Add define_expand.
17339 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17342 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17343 op_mode in the force_to_mode call.
17345 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17347 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17348 instead of checking each element individually.
17349 (aarch64_evpc_uzp): Likewise.
17350 (aarch64_evpc_zip): Likewise.
17351 (aarch64_evpc_ext): Likewise.
17352 (aarch64_evpc_rev): Likewise.
17353 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17354 instead of checking each element individually. Return true without
17356 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17357 whether all selected elements come from the same input, instead of
17358 checking each element individually. Remove calls to gen_rtx_REG,
17359 start_sequence and end_sequence and instead assert that no rtl is
17362 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17364 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17365 order of HIGH and CONST checks.
17367 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17369 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17370 if the destination isn't an SSA_NAME.
17372 2018-01-09 Richard Biener <rguenther@suse.de>
17374 PR tree-optimization/83668
17375 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17377 (canonicalize_loop_form): ... here, renamed from ...
17378 (canonicalize_loop_closed_ssa_form): ... this and amended to
17379 swap successor edges for loop exit blocks to make us use
17380 the RPO order we need for initial schedule generation.
17382 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17384 PR tree-optimization/64811
17385 * match.pd: When optimizing comparisons with Inf, avoid
17386 introducing or losing exceptions from comparisons with NaN.
17388 2018-01-09 Martin Liska <mliska@suse.cz>
17391 * asan.c (shadow_mem_size): Add gcc_assert.
17393 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17395 Don't save registers in main().
17398 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17399 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17400 * config/avr/avr.c (avr_set_current_function): Don't error if
17401 naked, OS_task or OS_main are specified at the same time.
17402 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17404 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17406 * common/config/avr/avr-common.c (avr_option_optimization_table):
17407 Switch on -mmain-is-OS_task for optimizing compilations.
17409 2018-01-09 Richard Biener <rguenther@suse.de>
17411 PR tree-optimization/83572
17412 * graphite.c: Include cfganal.h.
17413 (graphite_transform_loops): Connect infinite loops to exit
17414 and remove fake edges at the end.
17416 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17418 * ipa-inline.c (edge_badness): Revert accidental checkin.
17420 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17423 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17424 symbols; not inline clones.
17426 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17429 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17430 hard registers. Formatting fixes.
17432 PR preprocessor/83722
17433 * gcc.c (try_generate_repro): Pass
17434 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17435 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17438 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17439 Kito Cheng <kito.cheng@gmail.com>
17441 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17442 (riscv_leaf_function_p): Delete.
17443 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17445 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17447 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17449 (do_ifelse): New function.
17450 (do_isel): New function.
17451 (do_sub3): New function.
17452 (do_add3): New function.
17453 (do_load_mask_compare): New function.
17454 (do_overlap_load_compare): New function.
17455 (expand_compare_loop): New function.
17456 (expand_block_compare): Call expand_compare_loop() when appropriate.
17457 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17458 option description.
17459 (-mblock-compare-inline-loop-limit): New option.
17461 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17464 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17465 Reverse order of second and third operands in first alternative.
17466 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17467 of first and second elements in UNSPEC_VPERMR vector.
17468 (altivec_expand_vec_perm_le): Likewise.
17470 2018-01-08 Jeff Law <law@redhat.com>
17472 PR rtl-optimizatin/81308
17473 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17474 (process_switch): If group_case_labels makes a change, then set
17476 (pass_convert_switch::execute): If a switch is converted, then
17477 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17479 PR rtl-optimization/81308
17480 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17483 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17485 PR target/83663 - Revert r255946
17486 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17487 generation for cases where splatting a value is not useful.
17488 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17489 across a vec_duplicate and a paradoxical subreg forming a vector
17490 mode to a vec_concat.
17492 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17494 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17495 -march=armv8.3-a variants.
17496 * config/arm/t-multilib: Likewise.
17497 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17499 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17501 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17503 (cceq_ior_compare_complement): Give it a name so I can use it, and
17504 change boolean_or_operator predicate to boolean_operator so it can
17505 be used to generate a crand.
17506 (eqne): New code iterator.
17507 (bd/bd_neg): New code_attrs.
17508 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17509 a single define_insn.
17510 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17511 decrement (bdnzt/bdnzf/bdzt/bdzf).
17512 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17513 with the new names of the branch decrement patterns, and added the
17514 names of the branch decrement conditional patterns.
17516 2018-01-08 Richard Biener <rguenther@suse.de>
17518 PR tree-optimization/83563
17519 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17522 2018-01-08 Richard Biener <rguenther@suse.de>
17524 PR middle-end/83713
17525 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17527 2018-01-08 Richard Biener <rguenther@suse.de>
17529 PR tree-optimization/83685
17530 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17531 references to abnormals.
17533 2018-01-08 Richard Biener <rguenther@suse.de>
17536 * dwarf2out.c (output_indirect_strings): Handle empty
17537 skeleton_debug_str_hash.
17538 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17540 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17542 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17543 (emit_store_direct): Likewise.
17544 (arc_trampoline_adjust_address): Likewise.
17545 (arc_asm_trampoline_template): New function.
17546 (arc_initialize_trampoline): Use asm_trampoline_template.
17547 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17548 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17549 * config/arc/arc.md (flush_icache): Delete pattern.
17551 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17553 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17554 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17557 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17560 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17561 by not USED_FOR_TARGET.
17562 (make_pass_resolve_sw_modes): Likewise.
17564 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17566 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17569 2018-01-08 Richard Biener <rguenther@suse.de>
17571 PR middle-end/83580
17572 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17574 2018-01-08 Richard Biener <rguenther@suse.de>
17576 PR middle-end/83517
17577 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17579 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17581 PR middle-end/81897
17582 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17583 basic blocks with a small number of successors.
17584 (convert_control_dep_chain_into_preds): Improve handling of
17586 (dump_predicates): Split apart into...
17587 (dump_pred_chain): ...here...
17588 (dump_pred_info): ...and here.
17589 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17590 (can_chain_union_be_invalidated_p): Improve check for invalidation
17592 (uninit_uses_cannot_happen): Avoid unnecessary if
17593 convert_control_dep_chain_into_preds yielded nothing.
17595 2018-01-06 Martin Sebor <msebor@redhat.com>
17597 PR tree-optimization/83640
17598 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17599 subtracting negative offset from size.
17600 (builtin_access::overlap): Adjust offset bounds of the access to fall
17601 within the size of the object if possible.
17603 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17605 PR rtl-optimization/83699
17606 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17607 extract_bit_field_as_subreg to cases in which the extracted
17608 value is also a vector.
17610 * lra-constraints.c (process_alt_operands): Test for the equivalence
17611 substitutions when detecting a possible reload cycle.
17613 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17616 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17617 by default if flag_selective_schedling{,2}. Formatting fixes.
17619 PR rtl-optimization/83682
17620 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17621 if it has non-VECTOR_MODE element mode.
17622 (vec_duplicate_p): Likewise.
17624 PR middle-end/83694
17625 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17626 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17628 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17631 * config/i386/i386-builtin.def
17632 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17633 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17634 Require also OPTION_MASK_ISA_AVX512F in addition to
17635 OPTION_MASK_ISA_GFNI.
17636 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17637 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17638 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17639 to OPTION_MASK_ISA_GFNI.
17640 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17641 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17642 OPTION_MASK_ISA_AVX512BW.
17643 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17644 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17645 addition to OPTION_MASK_ISA_GFNI.
17646 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17647 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17648 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17649 to OPTION_MASK_ISA_GFNI.
17650 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17651 a requirement for all ISAs rather than any of them with a few
17653 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17655 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17656 bitmasks to be enabled with 3 exceptions, instead of requiring any
17657 enabled ISA with lots of exceptions.
17658 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17659 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17660 Change avx512bw in isa attribute to avx512f.
17661 * config/i386/sgxintrin.h: Add license boilerplate.
17662 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17663 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17664 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17665 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17667 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17668 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17669 temporarily sse2 rather than sse if not enabled already.
17672 * config/i386/sse.md (VI248_VLBW): Rename to ...
17673 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17674 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17675 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17676 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17677 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17678 mode iterator instead of VI248_VLBW.
17680 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17682 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17683 (record_modified): Skip clobbers; add debug output.
17684 (param_change_prob): Use sreal frequencies.
17686 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17688 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17689 punt for user-aligned variables.
17691 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17693 * tree-chrec.c (chrec_contains_symbols): Return true for
17696 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17699 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17700 of (x|y) == x for BICS pattern.
17702 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17704 PR tree-optimization/83605
17705 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17706 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17709 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17711 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17712 * config/epiphany/rtems.h: New file.
17714 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17715 Uros Bizjak <ubizjak@gmail.com>
17718 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17719 QIreg_operand instead of register_operand predicate.
17720 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17721 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17722 comments instead of -fmitigate[-_]rop.
17724 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17727 * cgraphunit.c (symbol_table::compile): Switch to text_section
17728 before calling assembly_start debug hook.
17729 * run-rtl-passes.c (run_rtl_passes): Likewise.
17732 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17734 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17735 range_int_cst_p rather than !symbolic_range_p before calling
17736 extract_range_from_multiplicative_op_1.
17738 2018-01-04 Jeff Law <law@redhat.com>
17740 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17741 redundant test in assertion.
17743 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17745 * doc/rtl.texi: Document machine_mode wrapper classes.
17747 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17749 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17750 using tree_to_uhwi.
17752 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17754 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17755 the VEC_PERM_EXPR fold to fail.
17757 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17760 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17761 to switched_sections.
17763 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17766 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17767 test for d.testing.
17769 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17772 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17773 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17775 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17778 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17779 is BLKmode and bitpos not zero or mode change is needed.
17781 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17784 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17787 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17790 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17791 instead of MULT rtx. Update all corresponding splitters.
17792 (*saddl_se): Ditto.
17793 (*ssub<modesuffix>): Ditto.
17794 (*ssubl_se): Ditto.
17795 (*cmp_sadd_di): Update split patterns.
17796 (*cmp_sadd_si): Ditto.
17797 (*cmp_sadd_sidi): Ditto.
17798 (*cmp_ssub_di): Ditto.
17799 (*cmp_ssub_si): Ditto.
17800 (*cmp_ssub_sidi): Ditto.
17801 * config/alpha/predicates.md (const23_operand): New predicate.
17802 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17803 Look for ASHIFT, not MULT inner operand.
17804 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17806 2018-01-04 Martin Liska <mliska@suse.cz>
17808 PR gcov-profile/83669
17809 * gcov.c (output_intermediate_file): Add version to intermediate
17811 * doc/gcov.texi: Document new field 'version' in intermediate
17812 file format. Fix location of '-k' option of gcov command.
17814 2018-01-04 Martin Liska <mliska@suse.cz>
17817 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17819 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17821 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17823 2018-01-03 Martin Sebor <msebor@redhat.com>
17825 PR tree-optimization/83655
17826 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17827 checking calls with invalid arguments.
17829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17831 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17832 (vectorizable_mask_load_store): Delete.
17833 (vectorizable_call): Return false for masked loads and stores.
17834 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17835 instead of gimple_assign_rhs1.
17836 (vectorizable_load): Handle IFN_MASK_LOAD.
17837 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17841 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17843 (vectorizable_mask_load_store): ...here.
17844 (vectorizable_load): ...and here.
17846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17848 * tree-vect-stmts.c (vect_build_all_ones_mask)
17849 (vect_build_zero_merge_argument): New functions, split out from...
17850 (vectorizable_load): ...here.
17852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17854 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17856 (vectorizable_mask_load_store): ...here.
17857 (vectorizable_store): ...and here.
17859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17861 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17863 (vectorizable_mask_load_store): ...here.
17865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17867 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17868 (vect_model_store_cost): Take a vec_load_store_type instead of a
17870 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17871 (vect_model_store_cost): Take a vec_load_store_type instead of a
17873 (vectorizable_mask_load_store): Update accordingly.
17874 (vectorizable_store): Likewise.
17875 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17879 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17880 IFN_MASK_LOAD calls here rather than...
17881 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17884 Alan Hayward <alan.hayward@arm.com>
17885 David Sherwood <david.sherwood@arm.com>
17887 * expmed.c (extract_bit_field_1): For vector extracts,
17888 fall back to extract_bit_field_as_subreg if vec_extract
17891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17892 Alan Hayward <alan.hayward@arm.com>
17893 David Sherwood <david.sherwood@arm.com>
17895 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17896 they are variable or constant sized.
17897 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17898 slots for constant-sized data.
17900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17901 Alan Hayward <alan.hayward@arm.com>
17902 David Sherwood <david.sherwood@arm.com>
17904 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17905 handling COND_EXPRs with boolean comparisons, try to find a better
17906 basis for the mask type than the boolean itself.
17908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17910 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17911 is calculated and how it can be overridden.
17912 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17913 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17915 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17919 Alan Hayward <alan.hayward@arm.com>
17920 David Sherwood <david.sherwood@arm.com>
17922 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17923 Remove the mode argument.
17924 (aarch64_simd_valid_immediate): Remove the mode and inverse
17926 * config/aarch64/iterators.md (bitsize): New iterator.
17927 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17928 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17929 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17930 aarch64_simd_valid_immediate.
17931 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17932 (aarch64_reg_or_bic_imm): Likewise.
17933 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17934 with an insn_type enum and msl with a modifier_type enum.
17935 Replace element_width with a scalar_mode. Change the shift
17936 to unsigned int. Add constructors for scalar_float_mode and
17937 scalar_int_mode elements.
17938 (aarch64_vect_float_const_representable_p): Delete.
17939 (aarch64_can_const_movi_rtx_p)
17940 (aarch64_simd_scalar_immediate_valid_for_move)
17941 (aarch64_simd_make_constant): Update call to
17942 aarch64_simd_valid_immediate.
17943 (aarch64_advsimd_valid_immediate_hs): New function.
17944 (aarch64_advsimd_valid_immediate): Likewise.
17945 (aarch64_simd_valid_immediate): Remove mode and inverse
17946 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17947 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17948 and aarch64_float_const_representable_p on the result.
17949 (aarch64_output_simd_mov_immediate): Remove mode argument.
17950 Update call to aarch64_simd_valid_immediate and use of
17951 simd_immediate_info.
17952 (aarch64_output_scalar_simd_mov_immediate): Update call
17955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17956 Alan Hayward <alan.hayward@arm.com>
17957 David Sherwood <david.sherwood@arm.com>
17959 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17960 (mode_nunits): Likewise CONST_MODE_NUNITS.
17961 * machmode.def (ADJUST_NUNITS): Document.
17962 * genmodes.c (mode_data::need_nunits_adj): New field.
17963 (blank_mode): Update accordingly.
17964 (adj_nunits): New variable.
17965 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17967 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17968 listed in adj_nunits.
17969 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17970 listed in adj_nunits. Don't emit case statements for such modes.
17971 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17972 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17973 nothing if adj_nunits is nonnull.
17974 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17975 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17976 (emit_mode_fbit): Update use of print_maybe_const_decl.
17977 (emit_move_size): Likewise. Treat the array as non-const
17979 (emit_mode_adjustments): Handle adj_nunits.
17981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17983 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17984 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17985 (VECTOR_MODES): Use it.
17986 (make_vector_modes): Take the prefix as an argument.
17988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17989 Alan Hayward <alan.hayward@arm.com>
17990 David Sherwood <david.sherwood@arm.com>
17992 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17993 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17994 for MODE_VECTOR_BOOL.
17995 * machmode.def (VECTOR_BOOL_MODE): Document.
17996 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17997 (make_vector_bool_mode): New function.
17998 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18000 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18001 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18003 * stor-layout.c (int_mode_for_mode): Likewise.
18004 * tree.c (build_vector_type_for_mode): Likewise.
18005 * varasm.c (output_constant_pool_2): Likewise.
18006 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18007 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18008 for MODE_VECTOR_BOOL.
18009 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18010 of mode class checks.
18011 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18012 instead of a list of mode class checks.
18013 (expand_vector_scalar_condition): Likewise.
18014 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18017 Alan Hayward <alan.hayward@arm.com>
18018 David Sherwood <david.sherwood@arm.com>
18020 * machmode.h (mode_size): Change from unsigned short to
18022 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18023 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18024 or if measurement_type is not polynomial.
18025 (fixed_size_mode::includes_p): Check for constant-sized modes.
18026 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18027 return a poly_uint16 rather than an unsigned short.
18028 (emit_mode_size): Change the type of mode_size from unsigned short
18029 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18030 (emit_mode_adjustments): Cope with polynomial vector sizes.
18031 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18033 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18035 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18036 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18037 * caller-save.c (setup_save_areas): Likewise.
18038 (replace_reg_with_saved_mem): Likewise.
18039 * calls.c (emit_library_call_value_1): Likewise.
18040 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18041 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18042 (gen_lowpart_for_combine): Likewise.
18043 * convert.c (convert_to_integer_1): Likewise.
18044 * cse.c (equiv_constant, cse_insn): Likewise.
18045 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18046 (cselib_subst_to_values): Likewise.
18047 * dce.c (word_dce_process_block): Likewise.
18048 * df-problems.c (df_word_lr_mark_ref): Likewise.
18049 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18050 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18051 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18052 (rtl_for_decl_location): Likewise.
18053 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18054 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18055 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18056 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18057 (expand_expr_real_1): Likewise.
18058 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18059 (pad_below): Likewise.
18060 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18061 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18062 * ira.c (get_subreg_tracking_sizes): Likewise.
18063 * ira-build.c (ira_create_allocno_objects): Likewise.
18064 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18065 (ira_sort_regnos_for_alter_reg): Likewise.
18066 * ira-costs.c (record_operand_costs): Likewise.
18067 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18068 (resolve_simple_move): Likewise.
18069 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18070 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18071 (lra_constraints): Likewise.
18072 (CONST_POOL_OK_P): Reject variable-sized modes.
18073 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18074 (add_pseudo_to_slot, lra_spill): Likewise.
18075 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18076 * optabs-query.c (get_best_extraction_insn): Likewise.
18077 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18078 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18079 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18080 * recog.c (offsettable_address_addr_space_p): Likewise.
18081 * regcprop.c (maybe_mode_change): Likewise.
18082 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18083 * regrename.c (build_def_use): Likewise.
18084 * regstat.c (dump_reg_info): Likewise.
18085 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18086 (find_reloads, find_reloads_subreg_address): Likewise.
18087 * reload1.c (eliminate_regs_1): Likewise.
18088 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18089 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18090 (simplify_binary_operation_1, simplify_subreg): Likewise.
18091 * targhooks.c (default_function_arg_padding): Likewise.
18092 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18093 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18094 (verify_gimple_assign_ternary): Likewise.
18095 * tree-inline.c (estimate_move_cost): Likewise.
18096 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18097 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18098 (get_address_cost_ainc): Likewise.
18099 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18100 (vect_supportable_dr_alignment): Likewise.
18101 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18102 (vectorizable_reduction): Likewise.
18103 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18104 (vectorizable_operation, vectorizable_load): Likewise.
18105 * tree.c (build_same_sized_truth_vector_type): Likewise.
18106 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18107 * var-tracking.c (emit_note_insn_var_location): Likewise.
18108 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18109 (ADDR_VEC_ALIGN): Likewise.
18111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18112 Alan Hayward <alan.hayward@arm.com>
18113 David Sherwood <david.sherwood@arm.com>
18115 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18117 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18118 or if measurement_type is polynomial.
18119 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18120 * combine.c (make_extraction): Likewise.
18121 * dse.c (find_shift_sequence): Likewise.
18122 * dwarf2out.c (mem_loc_descriptor): Likewise.
18123 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18124 (extract_bit_field, extract_low_bits): Likewise.
18125 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18126 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18127 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18128 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18129 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18130 * reload.c (find_reloads): Likewise.
18131 * reload1.c (alter_reg): Likewise.
18132 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18133 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18134 * tree-if-conv.c (predicate_mem_writes): Likewise.
18135 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18136 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18137 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18138 * valtrack.c (dead_debug_insert_temp): Likewise.
18139 * varasm.c (mergeable_constant_section): Likewise.
18140 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18143 Alan Hayward <alan.hayward@arm.com>
18144 David Sherwood <david.sherwood@arm.com>
18146 * expr.c (expand_assignment): Cope with polynomial mode sizes
18147 when assigning to a CONCAT.
18149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18150 Alan Hayward <alan.hayward@arm.com>
18151 David Sherwood <david.sherwood@arm.com>
18153 * machmode.h (mode_precision): Change from unsigned short to
18155 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18157 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18158 or if measurement_type is not polynomial.
18159 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18160 in which the mode is already known to be a scalar_int_mode.
18161 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18162 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18164 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18165 for GET_MODE_PRECISION.
18166 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18167 for GET_MODE_PRECISION.
18168 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18170 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18171 (expand_field_assignment, make_extraction): Likewise.
18172 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18173 (get_last_value): Likewise.
18174 * convert.c (convert_to_integer_1): Likewise.
18175 * cse.c (cse_insn): Likewise.
18176 * expr.c (expand_expr_real_1): Likewise.
18177 * lra-constraints.c (simplify_operand_subreg): Likewise.
18178 * optabs-query.c (can_atomic_load_p): Likewise.
18179 * optabs.c (expand_atomic_load): Likewise.
18180 (expand_atomic_store): Likewise.
18181 * ree.c (combine_reaching_defs): Likewise.
18182 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18183 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18184 * tree.h (type_has_mode_precision_p): Likewise.
18185 * ubsan.c (instrument_si_overflow): Likewise.
18187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18188 Alan Hayward <alan.hayward@arm.com>
18189 David Sherwood <david.sherwood@arm.com>
18191 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18192 polynomial numbers of units.
18193 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18194 (valid_vector_subparts_p): New function.
18195 (build_vector_type): Remove temporary shim and take the number
18196 of units as a poly_uint64 rather than an int.
18197 (build_opaque_vector_type): Take the number of units as a
18198 poly_uint64 rather than an int.
18199 * tree.c (build_vector_from_ctor): Handle polynomial
18200 TYPE_VECTOR_SUBPARTS.
18201 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18202 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18203 (build_vector_from_val): If the number of units is variable,
18204 use build_vec_duplicate_cst for constant operands and
18205 VEC_DUPLICATE_EXPR otherwise.
18206 (make_vector_type): Remove temporary is_constant ().
18207 (build_vector_type, build_opaque_vector_type): Take the number of
18208 units as a poly_uint64 rather than an int.
18209 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18211 * cfgexpand.c (expand_debug_expr): Likewise.
18212 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18213 (store_constructor, expand_expr_real_1): Likewise.
18214 (const_scalar_mask_from_tree): Likewise.
18215 * fold-const-call.c (fold_const_reduction): Likewise.
18216 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18217 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18218 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18219 (fold_relational_const): Likewise.
18220 (native_interpret_vector): Likewise. Change the size from an
18221 int to an unsigned int.
18222 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18223 TYPE_VECTOR_SUBPARTS.
18224 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18225 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18226 duplicating a non-constant operand into a variable-length vector.
18227 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18228 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18229 * ipa-icf.c (sem_variable::equals): Likewise.
18230 * match.pd: Likewise.
18231 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18232 * print-tree.c (print_node): Likewise.
18233 * stor-layout.c (layout_type): Likewise.
18234 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18235 * tree-cfg.c (verify_gimple_comparison): Likewise.
18236 (verify_gimple_assign_binary): Likewise.
18237 (verify_gimple_assign_ternary): Likewise.
18238 (verify_gimple_assign_single): Likewise.
18239 * tree-pretty-print.c (dump_generic_node): Likewise.
18240 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18241 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18242 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18243 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18244 (vect_shift_permute_load_chain): Likewise.
18245 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18246 (expand_vector_condition, optimize_vector_constructor): Likewise.
18247 (lower_vec_perm, get_compute_type): Likewise.
18248 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18249 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18250 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18251 (vect_recog_mask_conversion_pattern): Likewise.
18252 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18253 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18254 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18255 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18256 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18257 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18258 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18259 (supportable_widening_operation): Likewise.
18260 (supportable_narrowing_operation): Likewise.
18261 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18263 * varasm.c (output_constant): Likewise.
18265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18266 Alan Hayward <alan.hayward@arm.com>
18267 David Sherwood <david.sherwood@arm.com>
18269 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18270 so that both the length == 3 and length != 3 cases set up their
18271 own permute vectors. Add comments explaining why we know the
18272 number of elements is constant.
18273 (vect_permute_load_chain): Likewise.
18275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18276 Alan Hayward <alan.hayward@arm.com>
18277 David Sherwood <david.sherwood@arm.com>
18279 * machmode.h (mode_nunits): Change from unsigned char to
18281 (ONLY_FIXED_SIZE_MODES): New macro.
18282 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18283 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18284 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18286 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18287 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18288 or if measurement_type is not polynomial.
18289 * genmodes.c (ZERO_COEFFS): New macro.
18290 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18292 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18293 Use ZERO_COEFFS when emitting initializers.
18294 * data-streamer.h (bp_pack_poly_value): New function.
18295 (bp_unpack_poly_value): Likewise.
18296 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18297 for GET_MODE_NUNITS.
18298 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18299 for GET_MODE_NUNITS.
18300 * tree.c (make_vector_type): Remove temporary shim and make
18301 the real function take the number of units as a poly_uint64
18302 rather than an int.
18303 (build_vector_type_for_mode): Handle polynomial nunits.
18304 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18305 * emit-rtl.c (const_vec_series_p_1): Likewise.
18306 (gen_rtx_CONST_VECTOR): Likewise.
18307 * fold-const.c (test_vec_duplicate_folding): Likewise.
18308 * genrecog.c (validate_pattern): Likewise.
18309 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18310 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18311 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18312 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18313 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18314 * rtlanal.c (subreg_get_info): Likewise.
18315 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18316 (vect_grouped_load_supported): Likewise.
18317 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18318 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18319 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18320 (simplify_const_unary_operation, simplify_binary_operation_1)
18321 (simplify_const_binary_operation, simplify_ternary_operation)
18322 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18323 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18324 instead of CONST_VECTOR_NUNITS.
18325 * varasm.c (output_constant_pool_2): Likewise.
18326 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18327 explicit-encoded elements in the XVEC for variable-length vectors.
18329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18331 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18334 Alan Hayward <alan.hayward@arm.com>
18335 David Sherwood <david.sherwood@arm.com>
18337 * coretypes.h (fixed_size_mode): Declare.
18338 (fixed_size_mode_pod): New typedef.
18339 * builtins.h (target_builtins::x_apply_args_mode)
18340 (target_builtins::x_apply_result_mode): Change type to
18341 fixed_size_mode_pod.
18342 * builtins.c (apply_args_size, apply_result_size, result_vector)
18343 (expand_builtin_apply_args_1, expand_builtin_apply)
18344 (expand_builtin_return): Update accordingly.
18346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18348 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18349 * cselib.c (cselib_hash_rtx): Likewise.
18350 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18351 CONST_VECTOR encoding.
18353 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18354 Jeff Law <law@redhat.com>
18357 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18358 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18359 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18360 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18363 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18364 explicitly probe *sp in a noreturn function if there were any callee
18365 register saves or frame pointer is needed.
18367 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18370 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18371 BLKmode for ternary, binary or unary expressions.
18374 * var-tracking.c (delete_vta_debug_insn): New inline function.
18375 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18376 insns from get_insns () to NULL instead of each bb separately.
18377 Use delete_vta_debug_insn. No longer static.
18378 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18379 delete_vta_debug_insns callers.
18380 * rtl.h (delete_vta_debug_insns): Declare.
18381 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18382 instead of variable_tracking_main.
18384 2018-01-03 Martin Sebor <msebor@redhat.com>
18386 PR tree-optimization/83603
18387 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18388 arguments past the endof the argument list in functions declared
18389 without a prototype.
18390 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18391 Avoid checking when arguments are null.
18393 2018-01-03 Martin Sebor <msebor@redhat.com>
18396 * doc/extend.texi (attribute const): Fix a typo.
18397 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18398 issuing -Wsuggest-attribute for void functions.
18400 2018-01-03 Martin Sebor <msebor@redhat.com>
18402 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18403 offset_int::from instead of wide_int::to_shwi.
18404 (maybe_diag_overlap): Remove assertion.
18405 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18406 * gimple-ssa-sprintf.c (format_directive): Same.
18407 (parse_directive): Same.
18408 (sprintf_dom_walker::compute_format_length): Same.
18409 (try_substitute_return_value): Same.
18411 2018-01-03 Jeff Law <law@redhat.com>
18413 PR middle-end/83654
18414 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18415 non-constant residual for zero at runtime and avoid probing in
18416 that case. Reorganize code for trailing problem to mirror handling
18419 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18421 PR tree-optimization/83501
18422 * tree-ssa-strlen.c (get_string_cst): New.
18423 (handle_char_store): Call get_string_cst.
18425 2018-01-03 Martin Liska <mliska@suse.cz>
18427 PR tree-optimization/83593
18428 * tree-ssa-strlen.c: Include tree-cfg.h.
18429 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18430 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18431 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18433 (strlen_dom_walker::before_dom_children): Call
18434 gimple_purge_dead_eh_edges. Dump tranformation with details
18436 (strlen_dom_walker::before_dom_children): Update call by adding
18437 new argument cleanup_eh.
18438 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18440 2018-01-03 Martin Liska <mliska@suse.cz>
18443 * cif-code.def (VARIADIC_THUNK): New enum value.
18444 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18447 2018-01-03 Jan Beulich <jbeulich@suse.com>
18449 * sse.md (mov<mode>_internal): Tighten condition for when to use
18450 vmovdqu<ssescalarsize> for TI and OI modes.
18452 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18454 Update copyright years.
18456 2018-01-03 Martin Liska <mliska@suse.cz>
18459 * ipa-visibility.c (function_and_variable_visibility): Skip
18460 functions with noipa attribure.
18462 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18464 * gcc.c (process_command): Update copyright notice dates.
18465 * gcov-dump.c (print_version): Ditto.
18466 * gcov.c (print_version): Ditto.
18467 * gcov-tool.c (print_version): Ditto.
18468 * gengtype.c (create_file): Ditto.
18469 * doc/cpp.texi: Bump @copying's copyright year.
18470 * doc/cppinternals.texi: Ditto.
18471 * doc/gcc.texi: Ditto.
18472 * doc/gccint.texi: Ditto.
18473 * doc/gcov.texi: Ditto.
18474 * doc/install.texi: Ditto.
18475 * doc/invoke.texi: Ditto.
18477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18479 * vector-builder.h (vector_builder::m_full_nelts): Change from
18480 unsigned int to poly_uint64.
18481 (vector_builder::full_nelts): Update prototype accordingly.
18482 (vector_builder::new_vector): Likewise.
18483 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18484 (vector_builder::operator ==): Likewise.
18485 (vector_builder::finalize): Likewise.
18486 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18487 Take the number of elements as a poly_uint64 rather than an
18489 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18490 from unsigned int to poly_uint64.
18491 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18492 (vec_perm_indices::new_vector): Likewise.
18493 (vec_perm_indices::length): Likewise.
18494 (vec_perm_indices::nelts_per_input): Likewise.
18495 (vec_perm_indices::input_nelts): Likewise.
18496 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18497 number of elements per input as a poly_uint64 rather than an
18498 unsigned int. Use the original encoding for variable-length
18499 vectors, rather than clamping each individual element.
18500 For the second and subsequent elements in each pattern,
18501 clamp the step and base before clamping their sum.
18502 (vec_perm_indices::series_p): Handle polynomial element counts.
18503 (vec_perm_indices::all_in_range_p): Likewise.
18504 (vec_perm_indices_to_tree): Likewise.
18505 (vec_perm_indices_to_rtx): Likewise.
18506 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18507 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18508 (tree_vector_builder::new_binary_operation): Handle polynomial
18509 element counts. Return false if we need to know the number
18510 of elements at compile time.
18511 * fold-const.c (fold_vec_perm): Punt if the number of elements
18512 isn't known at compile time.
18514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18516 * vec-perm-indices.h (vec_perm_builder): Change element type
18517 from HOST_WIDE_INT to poly_int64.
18518 (vec_perm_indices::element_type): Update accordingly.
18519 (vec_perm_indices::clamp): Handle polynomial element_types.
18520 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18521 (vec_perm_indices::all_in_range_p): Likewise.
18522 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18524 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18525 polynomial vec_perm_indices element types.
18526 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18527 * fold-const.c (fold_vec_perm): Likewise.
18528 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18529 * tree-vect-generic.c (lower_vec_perm): Likewise.
18530 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18531 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18532 element type to HOST_WIDE_INT.
18534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18535 Alan Hayward <alan.hayward@arm.com>
18536 David Sherwood <david.sherwood@arm.com>
18538 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18539 rather than an int. Use plus_constant.
18540 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18541 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18544 Alan Hayward <alan.hayward@arm.com>
18545 David Sherwood <david.sherwood@arm.com>
18547 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18548 a HOST_WIDE_INT to a poly_int64.
18550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18551 Alan Hayward <alan.hayward@arm.com>
18552 David Sherwood <david.sherwood@arm.com>
18554 * calls.c (load_register_parameters): Cope with polynomial
18555 mode sizes. Require a constant size for BLKmode parameters
18556 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18557 forces a parameter to be padded at the lsb end in order to
18558 fill a complete number of words, require the parameter size
18559 to be ordered wrt UNITS_PER_WORD.
18561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18562 Alan Hayward <alan.hayward@arm.com>
18563 David Sherwood <david.sherwood@arm.com>
18565 * reload1.c (spill_stack_slot_width): Change element type
18566 from unsigned int to poly_uint64_pod.
18567 (alter_reg): Treat mode sizes as polynomial.
18569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18570 Alan Hayward <alan.hayward@arm.com>
18571 David Sherwood <david.sherwood@arm.com>
18573 * reload.c (complex_word_subreg_p): New function.
18574 (reload_inner_reg_of_subreg, push_reload): Use it.
18576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18577 Alan Hayward <alan.hayward@arm.com>
18578 David Sherwood <david.sherwood@arm.com>
18580 * lra-constraints.c (process_alt_operands): Reject matched
18581 operands whose sizes aren't ordered.
18582 (match_reload): Refer to this check here.
18584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18585 Alan Hayward <alan.hayward@arm.com>
18586 David Sherwood <david.sherwood@arm.com>
18588 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18589 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18596 Use plus_constant instead of gen_rtx_PLUS.
18598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18599 Alan Hayward <alan.hayward@arm.com>
18600 David Sherwood <david.sherwood@arm.com>
18602 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18603 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18604 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18605 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18606 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18607 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18608 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18609 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18610 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18611 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18613 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18614 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18615 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18616 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18617 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18618 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18619 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18620 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18621 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18622 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18624 * expr.c (emit_move_resolve_push): Treat the input and result
18625 of PUSH_ROUNDING as a poly_int64.
18626 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18627 (emit_push_insn): Likewise.
18628 * lra-eliminations.c (mark_not_eliminable): Likewise.
18629 * recog.c (push_operand): Likewise.
18630 * reload1.c (elimination_effects): Likewise.
18631 * rtlanal.c (nonzero_bits1): Likewise.
18632 * calls.c (store_one_arg): Likewise. Require the padding to be
18633 known at compile time.
18635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18636 Alan Hayward <alan.hayward@arm.com>
18637 David Sherwood <david.sherwood@arm.com>
18639 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18640 Use plus_constant instead of gen_rtx_PLUS.
18642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18643 Alan Hayward <alan.hayward@arm.com>
18644 David Sherwood <david.sherwood@arm.com>
18646 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18647 rather than an int.
18649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18650 Alan Hayward <alan.hayward@arm.com>
18651 David Sherwood <david.sherwood@arm.com>
18653 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18654 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18655 via stack temporaries. Treat the mode size as polynomial too.
18657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18658 Alan Hayward <alan.hayward@arm.com>
18659 David Sherwood <david.sherwood@arm.com>
18661 * expr.c (expand_expr_real_2): When handling conversions involving
18662 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18663 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18664 as a poly_uint64 too.
18666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18667 Alan Hayward <alan.hayward@arm.com>
18668 David Sherwood <david.sherwood@arm.com>
18670 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
18677 REGMODE_NATURAL_SIZE.
18678 * expmed.c (store_bit_field_1): Likewise.
18679 * expr.c (store_constructor): Likewise.
18680 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18681 and polynomial REGMODE_NATURAL_SIZE.
18682 (gen_lowpart_common): Likewise.
18683 * reginfo.c (record_subregs_of_mode): Likewise.
18684 * rtlanal.c (read_modify_subreg_p): Likewise.
18686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18687 Alan Hayward <alan.hayward@arm.com>
18688 David Sherwood <david.sherwood@arm.com>
18690 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18691 numbers of elements.
18693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18694 Alan Hayward <alan.hayward@arm.com>
18695 David Sherwood <david.sherwood@arm.com>
18697 * match.pd: Cope with polynomial numbers of vector elements.
18699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18700 Alan Hayward <alan.hayward@arm.com>
18701 David Sherwood <david.sherwood@arm.com>
18703 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18704 in a POINTER_PLUS_EXPR.
18706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18707 Alan Hayward <alan.hayward@arm.com>
18708 David Sherwood <david.sherwood@arm.com>
18710 * omp-simd-clone.c (simd_clone_subparts): New function.
18711 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18712 (ipa_simd_modify_function_body): Likewise.
18714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18715 Alan Hayward <alan.hayward@arm.com>
18716 David Sherwood <david.sherwood@arm.com>
18718 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18719 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18720 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18721 (expand_vector_condition, vector_element): Likewise.
18722 (subparts_gt): New function.
18723 (get_compute_type): Use subparts_gt.
18724 (count_type_subparts): Delete.
18725 (expand_vector_operations_1): Use subparts_gt instead of
18726 count_type_subparts.
18728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18729 Alan Hayward <alan.hayward@arm.com>
18730 David Sherwood <david.sherwood@arm.com>
18732 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18733 (vect_compile_time_alias): ...this new function. Do the calculation
18734 on poly_ints rather than trees.
18735 (vect_prune_runtime_alias_test_list): Update call accordingly.
18737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18738 Alan Hayward <alan.hayward@arm.com>
18739 David Sherwood <david.sherwood@arm.com>
18741 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18743 (vect_schedule_slp_instance): Likewise.
18745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18746 Alan Hayward <alan.hayward@arm.com>
18747 David Sherwood <david.sherwood@arm.com>
18749 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18750 constant and extern definitions for variable-length vectors.
18751 (vect_get_constant_vectors): Note that the number of units
18752 is known to be constant.
18754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18755 Alan Hayward <alan.hayward@arm.com>
18756 David Sherwood <david.sherwood@arm.com>
18758 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18759 of units as polynomial. Choose between WIDE and NARROW based
18762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18763 Alan Hayward <alan.hayward@arm.com>
18764 David Sherwood <david.sherwood@arm.com>
18766 * tree-vect-stmts.c (simd_clone_subparts): New function.
18767 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18770 Alan Hayward <alan.hayward@arm.com>
18771 David Sherwood <david.sherwood@arm.com>
18773 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18774 vectors as polynomial. Use build_index_vector for
18775 IFN_GOMP_SIMD_LANE.
18777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18778 Alan Hayward <alan.hayward@arm.com>
18779 David Sherwood <david.sherwood@arm.com>
18781 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18782 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18783 for variable-length vectors.
18784 (vectorizable_mask_load_store): Treat the number of units as
18785 polynomial, asserting that it is constant if the condition has
18786 already been enforced.
18787 (vectorizable_store, vectorizable_load): Likewise.
18789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18790 Alan Hayward <alan.hayward@arm.com>
18791 David Sherwood <david.sherwood@arm.com>
18793 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18794 of units as polynomial. Punt if we can't tell at compile time
18795 which vector contains the final result.
18797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18798 Alan Hayward <alan.hayward@arm.com>
18799 David Sherwood <david.sherwood@arm.com>
18801 * tree-vect-loop.c (vectorizable_induction): Treat the number
18802 of units as polynomial. Punt on SLP inductions. Use an integer
18803 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18804 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
18812 * tree.c (build_index_vector): New function.
18813 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18814 of units as polynomial, forcibly converting it to a constant if
18815 vectorizable_reduction has already enforced the condition.
18816 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18817 to create a {1,2,3,...} vector.
18818 (vectorizable_reduction): Treat the number of units as polynomial.
18819 Choose vectype_in based on the largest scalar element size rather
18820 than the smallest number of units. Enforce the restrictions
18823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18824 Alan Hayward <alan.hayward@arm.com>
18825 David Sherwood <david.sherwood@arm.com>
18827 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18828 number of units as polynomial.
18830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18831 Alan Hayward <alan.hayward@arm.com>
18832 David Sherwood <david.sherwood@arm.com>
18834 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18835 * target.def (autovectorize_vector_sizes): Return the vector sizes
18836 by pointer, using vector_sizes rather than a bitmask.
18837 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18838 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18839 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18841 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18842 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18843 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18844 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18845 * omp-general.c (omp_max_vf): Likewise.
18846 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18847 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18848 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18849 * tree-vect-slp.c (vect_slp_bb): Likewise.
18850 * doc/tm.texi: Regenerate.
18851 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18853 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18854 the vector size as a poly_uint64 rather than an unsigned int.
18855 (current_vector_size): Change from an unsigned int to a poly_uint64.
18856 (get_vectype_for_scalar_type): Update accordingly.
18857 * tree.h (build_truth_vector_type): Take the size and number of
18858 units as a poly_uint64 rather than an unsigned int.
18859 (build_vector_type): Add a temporary overload that takes
18860 the number of units as a poly_uint64 rather than an unsigned int.
18861 * tree.c (make_vector_type): Likewise.
18862 (build_truth_vector_type): Take the number of units as a poly_uint64
18863 rather than an unsigned int.
18865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18866 Alan Hayward <alan.hayward@arm.com>
18867 David Sherwood <david.sherwood@arm.com>
18869 * target.def (get_mask_mode): Take the number of units and length
18870 as poly_uint64s rather than unsigned ints.
18871 * targhooks.h (default_get_mask_mode): Update accordingly.
18872 * targhooks.c (default_get_mask_mode): Likewise.
18873 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18874 * doc/tm.texi: Regenerate.
18876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18877 Alan Hayward <alan.hayward@arm.com>
18878 David Sherwood <david.sherwood@arm.com>
18880 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18881 * omp-general.c (omp_max_vf): Likewise.
18882 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18883 (expand_omp_simd): Handle polynomial safelen.
18884 * omp-low.c (omplow_simd_context): Add a default constructor.
18885 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18886 (lower_rec_simd_input_clauses): Update accordingly.
18887 (lower_rec_input_clauses): Likewise.
18889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18890 Alan Hayward <alan.hayward@arm.com>
18891 David Sherwood <david.sherwood@arm.com>
18893 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18894 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18895 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18896 (vect_analyze_slp_cost): Likewise.
18897 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18898 (vect_model_load_cost): Likewise.
18900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18901 Alan Hayward <alan.hayward@arm.com>
18902 David Sherwood <david.sherwood@arm.com>
18904 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18905 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18906 from an unsigned int * to a poly_uint64_pod *.
18907 (calculate_unrolling_factor): New function.
18908 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18911 Alan Hayward <alan.hayward@arm.com>
18912 David Sherwood <david.sherwood@arm.com>
18914 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18915 from an unsigned int to a poly_uint64.
18916 (_loop_vec_info::slp_unrolling_factor): Likewise.
18917 (_loop_vec_info::vectorization_factor): Change from an int
18919 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18920 (vect_get_num_vectors): New function.
18921 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18922 (vect_get_num_copies): Use vect_get_num_vectors.
18923 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18924 to an unsigned int *.
18925 (vect_analyze_data_refs): Change min_vf from an int * to a
18927 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18928 than an unsigned HOST_WIDE_INT.
18929 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18930 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18931 to an unsigned int *.
18932 (vect_analyze_data_ref_dependences): Likewise.
18933 (vect_compute_data_ref_alignment): Handle polynomial vf.
18934 (vect_enhance_data_refs_alignment): Likewise.
18935 (vect_prune_runtime_alias_test_list): Likewise.
18936 (vect_shift_permute_load_chain): Likewise.
18937 (vect_supportable_dr_alignment): Likewise.
18938 (dependence_distance_ge_vf): Take the vectorization factor as a
18939 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18940 (vect_analyze_data_refs): Change min_vf from an int * to a
18942 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18943 vfm1 as a poly_uint64 rather than an int. Make the same change
18944 for the returned bound_scalar.
18945 (vect_gen_vector_loop_niters): Handle polynomial vf.
18946 (vect_do_peeling): Likewise. Update call to
18947 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18948 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18950 * tree-vect-loop.c (vect_determine_vectorization_factor)
18951 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18952 (vect_get_known_peeling_cost): Likewise.
18953 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18954 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18955 (vect_transform_loop): Likewise. Use the lowest possible VF when
18956 updating the upper bounds of the loop.
18957 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18958 rather than an int.
18959 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18960 polynomial unroll factors.
18961 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18962 (vect_make_slp_decision): Likewise.
18963 (vect_supported_load_permutation_p): Likewise, and polynomial
18965 (vect_analyze_slp_cost): Handle polynomial vf.
18966 (vect_slp_analyze_node_operations): Likewise.
18967 (vect_slp_analyze_bb_1): Likewise.
18968 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18969 than an unsigned HOST_WIDE_INT.
18970 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18971 (vectorizable_load): Handle polynomial vf.
18972 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18974 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18977 Alan Hayward <alan.hayward@arm.com>
18978 David Sherwood <david.sherwood@arm.com>
18980 * match.pd: Handle bit operations involving three constants
18981 and try to fold one pair.
18983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18985 * tree-vect-loop-manip.c: Include gimple-fold.h.
18986 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18987 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18988 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18989 Add a path that uses a step of VF instead of 1, but disable it
18991 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18992 and niters_no_overflow parameters. Update calls to
18993 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18994 Create a new SSA name if the latter choses to use a ste other
18995 than zero, and return it via niters_vector_mult_vf_var.
18996 * tree-vect-loop.c (vect_transform_loop): Update calls to
18997 vect_do_peeling, vect_gen_vector_loop_niters and
18998 slpeel_make_loop_iterate_ntimes.
18999 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19000 (vect_gen_vector_loop_niters): Update declarations after above changes.
19002 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19004 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19005 128-bit round to integer instructions.
19006 (ceil<mode>2): Likewise.
19007 (btrunc<mode>2): Likewise.
19008 (round<mode>2): Likewise.
19010 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19012 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19013 unaligned VSX load/store on P8/P9.
19014 (expand_block_clear): Allow the use of unaligned VSX
19015 load/store on P8/P9.
19017 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19019 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19021 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19022 swap associated with both a load and a store.
19024 2018-01-02 Andrew Waterman <andrew@sifive.com>
19026 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19027 * config/riscv/riscv.md (clear_cache): Use it.
19029 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19031 * web.c: Remove out-of-date comment.
19033 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19035 * expr.c (fixup_args_size_notes): Check that any existing
19036 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19037 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19038 (emit_single_push_insn): ...here.
19040 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19042 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19043 (const_vector_encoded_nelts): New function.
19044 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19045 (const_vector_int_elt, const_vector_elt): Declare.
19046 * emit-rtl.c (const_vector_int_elt_1): New function.
19047 (const_vector_elt): Likewise.
19048 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19049 of CONST_VECTOR_ELT.
19051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19053 * expr.c: Include rtx-vector-builder.h.
19054 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19055 directly on the tree encoding.
19056 (const_vector_from_tree): Likewise.
19057 * optabs.c: Include rtx-vector-builder.h.
19058 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19059 sequence of "u" values.
19060 * vec-perm-indices.c: Include rtx-vector-builder.h.
19061 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19062 directly on the vec_perm_indices encoding.
19064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19066 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19067 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19068 * rtx-vector-builder.h: New file.
19069 * rtx-vector-builder.c: Likewise.
19070 * rtl.h (rtx_def::u2): Add a const_vector field.
19071 (CONST_VECTOR_NPATTERNS): New macro.
19072 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19073 (CONST_VECTOR_DUPLICATE_P): Likewise.
19074 (CONST_VECTOR_STEPPED_P): Likewise.
19075 (CONST_VECTOR_ENCODED_ELT): Likewise.
19076 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19077 (unwrap_const_vec_duplicate): Likewise.
19078 (const_vec_series_p): Check for a non-duplicated vector encoding.
19079 Say that the function only returns true for integer vectors.
19080 * emit-rtl.c: Include rtx-vector-builder.h.
19081 (gen_const_vec_duplicate_1): Delete.
19082 (gen_const_vector): Call gen_const_vec_duplicate instead of
19083 gen_const_vec_duplicate_1.
19084 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19085 (gen_const_vec_duplicate): Use rtx_vector_builder.
19086 (gen_const_vec_series): Likewise.
19087 (gen_rtx_CONST_VECTOR): Likewise.
19088 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19089 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19090 Build a new vector rather than modifying a CONST_VECTOR in-place.
19091 (handle_special_swappables): Update call accordingly.
19092 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19093 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19094 Build a new vector rather than modifying a CONST_VECTOR in-place.
19095 (handle_special_swappables): Update call accordingly.
19097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19099 * simplify-rtx.c (simplify_const_binary_operation): Use
19100 CONST_VECTOR_ELT instead of XVECEXP.
19102 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19104 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19105 the selector elements to be different from the data elements
19106 if the selector is a VECTOR_CST.
19107 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19108 ssizetype for the selector.
19110 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19112 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19113 before testing each element individually.
19114 * tree-vect-generic.c (lower_vec_perm): Likewise.
19116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19118 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19119 * selftest-run-tests.c (selftest::run_tests): Call it.
19120 * vector-builder.h (vector_builder::operator ==): New function.
19121 (vector_builder::operator !=): Likewise.
19122 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19123 (vec_perm_indices::all_from_input_p): New function.
19124 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19125 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19126 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19127 instead of reading the VECTOR_CST directly. Detect whether both
19128 vector inputs are the same before constructing the vec_perm_indices,
19129 and update the number of inputs argument accordingly. Use the
19130 utility functions added above. Only construct sel2 if we need to.
19132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19134 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19135 the broadcast of the low byte.
19136 (expand_mult_highpart): Use an explicit encoding for the permutes.
19137 * optabs-query.c (can_mult_highpart_p): Likewise.
19138 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19139 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19140 (vectorizable_bswap): Likewise.
19141 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19142 explicit encoding for the power-of-2 permutes.
19143 (vect_permute_store_chain): Likewise.
19144 (vect_grouped_load_supported): Likewise.
19145 (vect_permute_load_chain): Likewise.
19147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19149 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19150 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19151 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19152 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19153 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19154 (vect_gen_perm_mask_any): Likewise.
19156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19158 * int-vector-builder.h: New file.
19159 * vec-perm-indices.h: Include int-vector-builder.h.
19160 (vec_perm_indices): Redefine as an int_vector_builder.
19161 (auto_vec_perm_indices): Delete.
19162 (vec_perm_builder): Redefine as a stand-alone class.
19163 (vec_perm_indices::vec_perm_indices): New function.
19164 (vec_perm_indices::clamp): Likewise.
19165 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19166 (vec_perm_indices::new_vector): New function.
19167 (vec_perm_indices::new_expanded_vector): Update for new
19168 vec_perm_indices class.
19169 (vec_perm_indices::rotate_inputs): New function.
19170 (vec_perm_indices::all_in_range_p): Operate directly on the
19171 encoded form, without computing elided elements.
19172 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19173 encoding. Update for new vec_perm_indices class.
19174 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19175 the given vec_perm_builder.
19176 (expand_vec_perm_var): Update vec_perm_builder constructor.
19177 (expand_mult_highpart): Use vec_perm_builder instead of
19178 auto_vec_perm_indices.
19179 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19180 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19181 or double series encoding as appropriate.
19182 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19183 vec_perm_indices instead of auto_vec_perm_indices.
19184 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19185 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19186 (vect_permute_store_chain): Likewise.
19187 (vect_grouped_load_supported): Likewise.
19188 (vect_permute_load_chain): Likewise.
19189 (vect_shift_permute_load_chain): Likewise.
19190 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19191 (vect_transform_slp_perm_load): Likewise.
19192 (vect_schedule_slp_instance): Likewise.
19193 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19194 (vectorizable_mask_load_store): Likewise.
19195 (vectorizable_bswap): Likewise.
19196 (vectorizable_store): Likewise.
19197 (vectorizable_load): Likewise.
19198 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19199 vec_perm_indices instead of auto_vec_perm_indices. Use
19200 tree_to_vec_perm_builder to read the vector from a tree.
19201 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19202 vec_perm_builder instead of a vec_perm_indices.
19203 (have_whole_vector_shift): Use vec_perm_builder and
19204 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19205 truncation to calc_vec_perm_mask_for_shift.
19206 (vect_create_epilog_for_reduction): Likewise.
19207 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19208 from auto_vec_perm_indices to vec_perm_indices.
19209 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19210 instead of changing individual elements.
19211 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19212 the vector in d.perm.
19213 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19214 from auto_vec_perm_indices to vec_perm_indices.
19215 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19216 instead of changing individual elements.
19217 (arm_vectorize_vec_perm_const): Use new_vector to install
19218 the vector in d.perm.
19219 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19220 Update vec_perm_builder constructor.
19221 (rs6000_expand_interleave): Likewise.
19222 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19223 (rs6000_expand_interleave): Likewise.
19225 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19227 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19228 to qimode could truncate the indices.
19229 * optabs.c (expand_vec_perm_var): Likewise.
19231 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19233 * Makefile.in (OBJS): Add vec-perm-indices.o.
19234 * vec-perm-indices.h: New file.
19235 * vec-perm-indices.c: Likewise.
19236 * target.h (vec_perm_indices): Replace with a forward class
19238 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19239 * optabs.h: Include vec-perm-indices.h.
19240 (expand_vec_perm): Delete.
19241 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19242 (expand_vec_perm_const): Declare.
19243 * target.def (vec_perm_const_ok): Replace with...
19244 (vec_perm_const): ...this new hook.
19245 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19246 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19247 * doc/tm.texi: Regenerate.
19248 * optabs.def (vec_perm_const): Delete.
19249 * doc/md.texi (vec_perm_const): Likewise.
19250 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19251 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19252 expand_vec_perm for constant permutation vectors. Assert that
19253 the mode of variable permutation vectors is the integer equivalent
19254 of the mode that is being permuted.
19255 * optabs-query.h (selector_fits_mode_p): Declare.
19256 * optabs-query.c: Include vec-perm-indices.h.
19257 (selector_fits_mode_p): New function.
19258 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19259 is defined, instead of checking whether the vec_perm_const_optab
19260 exists. Use targetm.vectorize.vec_perm_const instead of
19261 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19262 fit in the vector mode before using a variable permute.
19263 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19264 vec_perm_indices instead of an rtx.
19265 (expand_vec_perm): Replace with...
19266 (expand_vec_perm_const): ...this new function. Take the selector
19267 as a vec_perm_indices rather than an rtx. Also take the mode of
19268 the selector. Update call to shift_amt_for_vec_perm_mask.
19269 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19270 Use vec_perm_indices::new_expanded_vector to expand the original
19271 selector into bytes. Check whether the indices fit in the vector
19272 mode before using a variable permute.
19273 (expand_vec_perm_var): Make global.
19274 (expand_mult_highpart): Use expand_vec_perm_const.
19275 * fold-const.c: Includes vec-perm-indices.h.
19276 * tree-ssa-forwprop.c: Likewise.
19277 * tree-vect-data-refs.c: Likewise.
19278 * tree-vect-generic.c: Likewise.
19279 * tree-vect-loop.c: Likewise.
19280 * tree-vect-slp.c: Likewise.
19281 * tree-vect-stmts.c: Likewise.
19282 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19284 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19285 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19286 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19287 (aarch64_vectorize_vec_perm_const): ...this new function.
19288 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19289 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19290 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19291 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19292 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19293 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19294 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19296 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19297 check for NEON modes.
19298 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19299 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19300 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19301 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19303 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19304 the old VEC_PERM_CONST conditions.
19305 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19306 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19307 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19308 (ia64_vectorize_vec_perm_const_ok): Merge into...
19309 (ia64_vectorize_vec_perm_const): ...this new function.
19310 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19311 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19312 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19313 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19314 * config/mips/mips.c (mips_expand_vec_perm_const)
19315 (mips_vectorize_vec_perm_const_ok): Merge into...
19316 (mips_vectorize_vec_perm_const): ...this new function.
19317 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19318 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19319 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19320 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19321 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19322 (rs6000_expand_vec_perm_const): Delete.
19323 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19325 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19326 (altivec_expand_vec_perm_const_le): Take each operand individually.
19327 Operate on constant selectors rather than rtxes.
19328 (altivec_expand_vec_perm_const): Likewise. Update call to
19329 altivec_expand_vec_perm_const_le.
19330 (rs6000_expand_vec_perm_const): Delete.
19331 (rs6000_vectorize_vec_perm_const_ok): Delete.
19332 (rs6000_vectorize_vec_perm_const): New function.
19333 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19334 an element count and rtx array.
19335 (rs6000_expand_extract_even): Update call accordingly.
19336 (rs6000_expand_interleave): Likewise.
19337 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19338 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19339 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19340 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19341 (rs6000_expand_vec_perm_const): Delete.
19342 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19343 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19344 (altivec_expand_vec_perm_const_le): Take each operand individually.
19345 Operate on constant selectors rather than rtxes.
19346 (altivec_expand_vec_perm_const): Likewise. Update call to
19347 altivec_expand_vec_perm_const_le.
19348 (rs6000_expand_vec_perm_const): Delete.
19349 (rs6000_vectorize_vec_perm_const_ok): Delete.
19350 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19351 reference to the SPE evmerge intructions.
19352 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19353 an element count and rtx array.
19354 (rs6000_expand_extract_even): Update call accordingly.
19355 (rs6000_expand_interleave): Likewise.
19356 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19357 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19359 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19361 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19363 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19364 vector mode and that that mode matches the mode of the data
19366 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19367 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19368 directly using expand_vec_perm_1 when forcing selectors into
19370 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19374 * optabs-query.h (can_vec_perm_p): Delete.
19375 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19376 * optabs-query.c (can_vec_perm_p): Split into...
19377 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19378 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19379 particular selector is valid.
19380 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19381 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19382 (vect_grouped_load_supported): Likewise.
19383 (vect_shift_permute_load_chain): Likewise.
19384 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19385 (vect_transform_slp_perm_load): Likewise.
19386 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19387 (vectorizable_bswap): Likewise.
19388 (vect_gen_perm_mask_checked): Likewise.
19389 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19390 implementations of variable permutation vectors into account
19391 when deciding which selector to use.
19392 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19393 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19394 with a false third argument.
19395 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19396 to test whether the constant selector is valid and can_vec_perm_var_p
19397 to test whether a variable selector is valid.
19399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19401 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19402 * optabs-query.c (can_vec_perm_p): Likewise.
19403 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19404 instead of vec_perm_indices.
19405 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19406 (vect_gen_perm_mask_checked): Likewise,
19407 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19408 (vect_gen_perm_mask_checked): Likewise,
19410 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19412 * optabs-query.h (qimode_for_vec_perm): Declare.
19413 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19414 (qimode_for_vec_perm): ...this new function.
19415 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19417 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19419 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19420 does not have a conditional at the top.
19422 2018-01-02 Richard Biener <rguenther@suse.de>
19424 * ipa-inline.c (big_speedup_p): Fix expression.
19426 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19429 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19432 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19436 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19437 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19438 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19439 cond_taken_branch_cost 3->4.
19441 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19443 PR tree-optimization/83581
19444 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19445 TODO_cleanup_cfg if any changes have been made.
19447 PR middle-end/83608
19448 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19449 convert_modes if target mode has the right side, but different mode
19452 PR middle-end/83609
19453 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19454 last argument when extracting from CONCAT. If either from_real or
19455 from_imag is NULL, use expansion through memory. If result is not
19456 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19457 the parts directly to inner mode, if even that fails, use expansion
19460 PR middle-end/83623
19461 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19462 check for bswap in mode rather than HImode and use that in expand_unop
19465 Copyright (C) 2018 Free Software Foundation, Inc.
19467 Copying and distribution of this file, with or without modification,
19468 are permitted in any medium without royalty provided the copyright
19469 notice and this notice are preserved.