1 2018-07-09 Martin Sebor <msebor@redhat.com>
5 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6 accessing implicitly initialized array elements.
7 * expr.c (string_constant): Handle string initializers of
8 character arrays within aggregates.
9 * gimple-fold.c (fold_array_ctor_reference): Add argument.
10 Store element offset. As a special case, handle zero size.
11 (fold_nonarray_ctor_reference): Same.
12 (fold_ctor_reference): Add argument. Store subobject offset.
13 * gimple-fold.h (fold_ctor_reference): Add argument.
15 2018-07-09 Paul Koning <ni1d@arrl.net>
17 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
18 (pdp11_insn_cost): New function.
19 (pdp11_md_asm_adjust): New function.
20 (TARGET_INVALID_WITHIN_DOLOOP): Define.
21 (pdp11_rtx_costs): Update to match machine better.
22 (output_addr_const_pdp11): Correct format mismatch warnings.
23 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
24 * config/pdp11/pdp11.md: General change to add base_cost and/or
25 length attributes for use by new pdp11_insn_cost function.
26 (MIN_BRANCH): Correct definition.
28 (doloop_end): Use standard pattern name for looping pattern.
29 (doloop_end_nocc): New.
30 (movsf): Add another constraint alternative.
31 (zero_extendqihi2): Add constraint alternatives for not in place
33 (zero_extendhisi2): Remove.
34 (shift patterns): Add CC handling variants.
38 (define_peephole2): New peephole to recognize mov that sets CC for
41 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
43 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
44 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
47 2018-07-09 Jakub Jelinek <jakub@redhat.com>
50 * real.c (real_nextafter): Return true if result is denormal.
52 2018-07-09 Martin Liska <mliska@suse.cz>
54 * common.opt: Add back wrongly removed attribute.
56 2018-07-09 Richard Biener <rguenther@suse.de>
59 * dwarf2out.c (gen_block_die): For an early generated DIE
60 always output high/low PC attributes.
62 2018-07-09 Tom de Vries <tdevries@suse.de>
64 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
65 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
66 onto VAR_DECL with abstract origin.
68 2018-07-07 Jim Wilson <jimw@sifive.com>
70 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
72 2018-07-07 Tom de Vries <tdevries@suse.de>
74 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
76 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
78 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
80 (combine_bound): Use wide-int overflow calculation instead of
82 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
84 * fold-const.c (int_const_binop_2): Same.
85 (extract_muldiv_1): Same.
86 (fold_div_compare): Same.
87 (fold_abs_const): Same.
89 * poly-int.h (add): Same.
93 * predict.c (predict_iv_comparison): Same.
94 * profile-count.c (slow_safe_scale_64bit): Same.
95 * simplify-rtx.c (simplify_const_binary_operation): Same.
96 * tree-chrec.c (tree_fold_binomial): Same.
97 * tree-data-ref.c (split_constant_offset_1): Same.
98 * tree-if-conv.c (idx_within_array_bound): Same.
99 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
100 * tree-ssa-phiopt.c (minmax_replacement): Same.
101 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
102 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
103 * vr-values.c (vr_values::adjust_range_with_scev): Same.
104 * wide-int.cc (wi::add_large): Same.
105 (wi::mul_internal): Same.
106 (wi::sub_large): Same.
107 (wi::divmod_internal): Same.
108 * wide-int.h: Change overflow type to overflow_type for neg, add,
109 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
110 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
111 mul_internal, divmod_internal.
112 (overflow_type): New enum.
113 (accumulate_overflow): New.
115 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
117 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
118 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
120 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
122 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
123 argument is checked for zero before entering loop, avoid checking again.
125 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
127 * gimplify.h (generic_expr_could_trap_p): Set as global function.
128 * gimplify.h (generic_expr_could_trap_p): Likwise.
129 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
131 2018-07-06 Jakub Jelinek <jakub@redhat.com>
133 PR tree-optimization/86401
134 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
135 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
136 (fold_bit_and_mask): ... here. New helper function for match.pd.
137 * fold-const.h (fold_bit_and_mask): Declare.
138 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
140 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
143 * target.def (translate_mode_attribute): New hook.
144 * targhooks.h (default_translate_mode_attribute): Declare.
145 * targhooks.c (default_translate_mode_attribute): New function.
146 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
147 * doc/tm.texi: Regenerate.
148 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
149 (rs6000_translate_mode_attribute): New function.
151 2018-07-06 Paul Koning <ni1d@arrl.net>
153 * doc/md.texi (define_split): Document DONE and FAIL.
154 (define_peephole2): Ditto.
156 2018-07-05 Jeff Law <law@redhat.com>
158 PR tree-optimization/86010
159 * tree-ssa-dse.c (compute_trims): More aggressively trim at
160 both the head and tail of mem* and str* calls.
162 2018-07-05 Jim Wilson <jimw@sifive.com>
164 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
166 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
168 * config/aarch64/aarch64-simd.md: correct flags text for
169 MIN_EXPR replacement.
171 2018-07-05 James Clarke <jrtc27@jrtc27.com>
173 * configure: Regenerated.
175 2018-07-05 Carl Love <cel@us.ibm.com>
177 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
178 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
179 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
180 VSX_BUILTIN_DOUBLEL_V4SF.
182 2018-07-05 Martin Sebor <msebor@redhat.com>
185 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
186 than its domain to compute its the upper bound of a char array.
188 2018-07-05 Nathan Sidwell <nathan@acm.org>
190 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
191 * doc/cpp.texi: Update comment.
192 * doc/tm.texi: Rebuilt.
193 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
194 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
195 * doc/extend.texi (Backwards Compatibility): Clarify it is system
196 headers affected by extern "C".
197 * system.h: Poison NO_IMPLICIT_EXTERN_C.
198 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
199 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
200 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
201 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
202 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
203 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
204 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
205 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
206 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
207 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
208 config/sparc/sp64-elf.h, config/spu/spu.h,
209 config/stormy16/stormy16.h, config/v850/v850.h,
210 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
211 define NO_IMPLICIT_EXTERN_C.
212 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
214 2018-07-05 Tamar Christina <tamar.christina@arm.com>
217 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
218 instead of GET_MODE_SIZE when comparing Units.
220 2018-07-05 Tamar Christina <tamar.christina@arm.com>
223 * rtlanal.c (set_noop_p): Constrain on mode change,
224 include hard-reg-set.h
226 2018-07-05 Tamar Christina <tamar.christina@arm.com>
228 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
230 2018-07-05 Jakub Jelinek <jakub@redhat.com>
233 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
236 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
238 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
240 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
243 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
245 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
247 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
250 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
252 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
254 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
257 2018-07-04 Martin Liska <mliska@suse.cz>
259 * tree-switch-conversion.c: Define
260 max_ratio_for_speed and max_ratio_for_size constants.
262 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
263 Martin Liska <mliska@suse.cz>
268 * common.opt: Rename align options with 'str_' prefix.
269 * common/config/i386/i386-common.c (set_malign_value): New
271 (ix86_handle_option): Use it to set -falign-* options/
272 * config/aarch64/aarch64-protos.h (struct tune_params): Change
273 type from int to string.
274 * config/aarch64/aarch64.c: Update default values from int
276 * config/alpha/alpha.c (alpha_override_options_after_change):
278 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
279 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
280 max skip conditionally.
281 * config/i386/freebsd.h (SUBALIGN_LOG): New.
282 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
283 max skip conditionally.
284 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
285 max skip conditionally.
286 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
287 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
288 max skip conditionally.
289 * config/i386/i386.c (struct ptt): Change type from int to
291 (ix86_default_align): Set default values.
292 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
293 max skip conditionally.
294 * config/i386/iamcu.h (SUBALIGN_LOG): New.
295 (ASM_OUTPUT_MAX_SKIP_ALIGN):
296 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
297 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
298 max skip conditionally.
299 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
300 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
301 * config/i386/x86-64.h (SUBALIGN_LOG): New.
302 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
303 max skip conditionally.
304 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
305 * config/ia64/ia64.c (ia64_option_override): Set default values
306 for alignment options.
307 * config/m68k/m68k.c: Handle new str_align_* options.
308 * config/mips/mips.c (mips_set_compression_mode): Change
310 (mips_option_override): Set default values for options.
311 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
313 * config/rs6000/rs6000.c (rs6000_option_override_internal):
315 * config/rx/rx.c (rx_option_override): Likewise.
316 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
317 (LABEL_ALIGN): Use align_labels_log.
318 (LOOP_ALIGN): Use align_loops_align.
319 * config/s390/s390.c (s390_asm_output_function_label): Use new
321 * config/sh/sh.c (sh_override_options_after_change):
322 Change type of constants.
323 * config/spu/spu.c (spu_sched_init): Likewise.
324 * config/sparc/sparc.c (sparc_option_override): Set default
326 * config/visium/visium.c (visium_option_override): Likewise.
327 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
328 emit p2align format with last argument if it's not needed.
329 * doc/invoke.texi: Document extended format of -falign-*.
330 * final.c: Use align_labels alignment.
331 * flags.h (struct target_flag_state): Change type to use
333 (struct align_flags_tuple): New.
334 (struct align_flags): Likewise.
335 (align_loops_log): Redefine macro to use new types.
336 (align_loops_max_skip): Redefine macro to use new types.
337 (align_jumps_log): Redefine macro to use new types.
338 (align_jumps_max_skip): Redefine macro to use new types.
339 (align_labels_log): Redefine macro to use new types.
340 (align_labels_max_skip): Redefine macro to use new types.
341 (align_functions_log): Redefine macro to use new types.
342 (align_loops): Redefine macro to use new types.
343 (align_jumps): Redefine macro to use new types.
344 (align_labels): Redefine macro to use new types.
345 (align_functions): Redefine macro to use new types.
346 (align_functions_max_skip): Redefine macro to use new types.
347 (align_loops_value): New macro.
348 (align_jumps_value): New macro.
349 (align_labels_value): New macro.
350 (align_functions_value): New macro.
351 * function.c (invoke_set_current_function_hook): Propagate
352 alignment values from flags to global variables default in
354 * ipa-icf.c (sem_function::equals_wpa): Use
355 cl_optimization_option_eq instead of memcmp.
356 * lto-streamer.h (cl_optimization_stream_out): Support streaming
358 (cl_optimization_stream_in): Likewise.
359 * optc-save-gen.awk: Support strings in cl_optimization.
360 * opth-gen.awk: Likewise.
361 * opts.c (finish_options): Remove error checking of invalid
363 (MAX_CODE_ALIGN): Remove.
364 (MAX_CODE_ALIGN_VALUE): Likewise.
365 (parse_and_check_align_values): New function.
366 (check_alignment_argument): Likewise.
367 (common_handle_option): Use check_alignment_argument.
368 * opts.h (parse_and_check_align_values): Declare.
369 * toplev.c (init_alignments): Remove.
370 (read_log_maxskip): New.
371 (parse_N_M): Likewise.
372 (parse_alignment_opts): Likewise.
373 (backend_init_target): Remove usage of init_alignments.
374 * toplev.h (parse_alignment_opts): Declare.
375 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
377 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
378 * tree.c (cl_option_hasher::equal): New.
379 * varasm.c: Use new global macros.
381 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
383 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
384 Use a simpler align directive also if MAXSKIP = ALIGN-1.
385 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
386 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
387 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
388 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
389 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
390 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
391 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
392 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
393 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
394 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
395 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
397 2018-07-04 Martin Liska <mliska@suse.cz>
398 Jonathan Wakely <jwakely@redhat.com>
400 * coverage.c: Use correct type.
401 * doc/invoke.texi: Language correction.
403 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
406 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
407 ENDBRANCH for non-tail call which may return via indirect branch.
408 * doc/extend.texi: Document indirect_return attribute.
410 2018-07-03 Martin Sebor <msebor@redhat.com>
412 PR tree-optimization/86274
413 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
415 (format_floating): Correct handling of infinities and NaNs.
417 2018-07-03 Martin Sebor <msebor@redhat.com>
419 * print-tree.c (print_real_cst): New function.
420 (print_node_brief): Call it.
423 2018-07-03 Jeff Law <law@redhat.com>
425 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
426 into a single pattern.
428 * config/h8300/h8300.md (ors code_iterator): New.
429 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
430 a single pattern and single splitter.
431 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
432 (iorqi3_1, xorqi3_1): Likewise.
433 (iorqi3, xorqi3 expanders): Similarly.
435 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
436 (movmd_internal) into a single pattern using the P mode iterator.
437 (movmd splitters): Similarly.
438 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
439 (movsd splitters): Similarly.
441 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
442 ADDB, ADDW and ADDL into a single ADD attribute which selects the
443 right table based on the size of the operand.
444 * config/h8300/h8300.md (length_table): Corresponding changes. All
445 references to "addb", "addw" and "addl" changed to "add".
446 (btst patterns): Merge two variants into a single pattern.
447 (tstqi, tsthi): Likewise.
448 (addhi3_incdec, addsi3_incdec): Likewise.
449 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
450 (mulhi3, mulsi3): Likewise.
451 (udivhi3, udivsi3): Likewise.
452 (divhi3, divsi3): Likewise.
453 (andorqi3, andorhi3, andorsi3): Likewise.
455 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
458 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
459 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
461 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
463 PR tree-optimization/85694
464 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
465 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
467 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
468 (<u>avg<mode>3_ceil): New patterns.
470 2018-07-03 David Malcolm <dmalcolm@redhat.com>
472 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
473 scan-tree-dump directive.
474 * gcc.dg/vect/slp-perm-2.c: Likewise.
475 * gcc.dg/vect/slp-perm-3.c: Likewise.
476 * gcc.dg/vect/slp-perm-5.c: Likewise.
477 * gcc.dg/vect/slp-perm-6.c: Likewise.
478 * gcc.dg/vect/slp-perm-7.c: Likewise.
479 * gcc.dg/vect/slp-perm-8.c: Likewise.
481 2018-07-03 Marek Polacek <polacek@redhat.com>
484 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
486 2018-07-03 Richard Biener <rguenther@suse.de>
489 * tree-ssa-structalias.c (find_func_clobbers): Properly
490 handle indirect calls.
492 2018-07-03 Jeff Law <law@redhat.com>
494 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
495 (shifts): New code iterator.
496 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
497 expander. Fix HImode handling on H8/SX.
498 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
499 (subqi3, subhi3, subsi3 expanders): Likewise.
500 (andqi3, andhi3, andsi3 expanders): Likewise.
501 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
502 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
503 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
504 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
505 (zero_extendqihi2, zero_extendqisi2): Likewise.
506 (extendqihi2, extendqisi2): Likewise.
507 (rotlqi3, rotlhi3, rotlsi3): Likewise.
508 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
509 (rotlqi3_1, rotlhi3_1): Likewise.
510 (logicalhi3_sn, logicalsi3_sn): Likewise.
511 (logicalhi3, logicalsi3): Likewise.
513 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
515 * tree-vect-patterns.c (vect_recog_rotate_pattern)
516 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
517 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
518 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
519 type to append_pattern_def_seq instead of creating a stmt_vec_info
521 (build_mask_conversion): Likewise. Remove vinfo argument.
522 (vect_add_conversion_to_patterm): Likewise, renaming to...
523 (vect_add_conversion_to_pattern): ...this.
524 (vect_recog_mask_conversion_pattern): Update call to
525 build_mask_conversion. Pass the vector type to
526 append_pattern_def_seq here too.
527 (vect_recog_gather_scatter_pattern): Update call to
528 vect_add_conversion_to_pattern.
530 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
532 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
533 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
534 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
535 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
536 STMT_VINFO_PATTERN_DEF_SEQ to null here.
537 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
538 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
539 append_pattern_def_seq instead of new_pattern_def_seq.
540 (vect_recog_divmod_pattern): Do both of the above.
541 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
544 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
546 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
547 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
548 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
549 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
550 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
551 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
552 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
553 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
554 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
555 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
556 parameter with a single stmt_vec_info.
557 (vect_recog_func_ptr): Likewise.
558 (vect_recog_gather_scatter_pattern): Likewise, folding in...
559 (vect_try_gather_scatter_pattern): ...this.
560 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
561 the stmt_vec_info of the statement to be matched. Don't clear
562 STMT_VINFO_RELATED_STMT.
563 (vect_pattern_recog): Update call accordingly.
565 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
567 PR tree-optimization/85694
568 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
569 (uavgM3_ceil): Document new optabs.
570 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
571 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
573 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
574 (savg_ceil_optab): New optabs.
575 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
576 (vect_vect_recog_func_ptrs): Add it.
577 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
578 constant directly from the associated lhs.
580 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
582 * tree-vect-patterns.c (vect_split_statement): New function.
583 (vect_convert_input): Use it to try to split an existing cast.
585 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
587 * poly-int.h (print_hex): New function.
588 * dumpfile.h (dump_dec, dump_hex): Declare.
589 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
590 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
591 min_input_precision, operation_precision and operation_sign.
592 * tree-vect-patterns.c (vect_get_range_info): New function.
593 (vect_same_loop_or_bb_p, vect_single_imm_use)
594 (vect_operation_fits_smaller_type): Delete.
595 (vect_look_through_possible_promotion): Add an optional
596 single_use_p parameter.
597 (vect_recog_over_widening_pattern): Rewrite to use new
598 stmt_vec_info infomration. Handle one operation at a time.
599 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
600 (vect_truncatable_operation_p, vect_set_operation_type)
601 (vect_set_min_input_precision): New functions.
602 (vect_determine_min_output_precision_1): Likewise.
603 (vect_determine_min_output_precision): Likewise.
604 (vect_determine_precisions_from_range): Likewise.
605 (vect_determine_precisions_from_users): Likewise.
606 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
607 (vect_vect_recog_func_ptrs): Put over_widening first.
609 (vect_pattern_recog): Call vect_determine_precisions.
611 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
613 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
614 statements that have been replaced by further pattern statements.
615 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
617 2018-07-03 Richard Biener <rguenther@suse.de>
619 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
620 always set *dt. Dump vectype in vectype overload.
621 * dumpfile.h (dump_gimple_expr): New function.
622 (dump_gimple_expr_loc): Likewise.
623 * dumpfile.c (dump_gimple_expr): New function.
624 (dump_gimple_expr_loc): Likewise.
626 2018-07-02 Jeff Law <law@redhat.com>
628 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
629 the H8/300, H8/300H and H8/S variants into a single pattern.
630 (movhi_h8300, movqi_h8300hs): Similarly.
631 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
632 (QHI mode iterator): New.
634 * config/h8300/h8300.md: Remove trailing whitespace.
636 2018-07-02 Jim Wilson <jimw@sifive.com>
638 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
639 instead of emit_insn for interrupt returns.
640 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
641 (riscv_sret, riscv_uret): Likewise.
643 2018-07-02 David Malcolm <dmalcolm@redhat.com>
645 * pretty-print.c (selftest::test_pp_format): Move save and restore
646 of quotes to class auto_fix_quotes, and add an instance.
647 * selftest.c: Include "intl.h".
648 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
649 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
650 * selftest.h (selftest::auto_fix_quotes): New class.
652 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
654 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
655 (aarch64_sve_prepare_conditional_op): Remove.
656 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
657 Allow aarch64_simd_reg_or_zero as select operand; remove
658 the aarch64_sve_prepare_conditional_op call.
659 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
660 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
661 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
662 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
663 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
664 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
665 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
666 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
667 and a splitters to match all of the *_any patterns.
668 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
670 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
671 (SVE_COND_FP_BINARY_REV): Remove.
672 (sve_int_op_rev, sve_fp_op_rev): New.
673 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
674 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
675 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
676 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
677 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
678 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
679 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
680 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
681 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
683 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
684 Remove match_dup 1 from the inner unspec.
685 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
687 * config/aarch64/aarch64.md (movprfx): New attr.
688 (length): Default movprfx to 8.
689 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
690 (*madd<SVE_I>, *msub<SVE_I): Likewise.
691 (*<su>mul<SVE_I>3_highpart): Likewise.
692 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
693 (*v<ASHIFT><SVE_I>3): Likewise.
694 (*<su><MAXMIN><SVE_I>3): Likewise.
695 (*<su><MAXMIN><SVE_F>3): Likewise.
696 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
697 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
698 (*div<SVE_F>4): Likewise.
700 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
702 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
705 2018-07-02 Richard Biener <rguenther@suse.de>
707 PR tree-optimization/86363
708 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
709 memset argument refers to a non-variable address.
711 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
713 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
714 {PLUS,MINUS}_EXPR code to...
715 (adjust_symbolic_bound): ...here,
716 (combine_bound): ...here,
717 (set_value_range_with_overflow): ...and here.
719 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
721 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
723 (extract_range_from_abs_expr): ...here.
725 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
727 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
728 -fno-omit-frame-pointer when not optimizing.
730 2018-07-02 Martin Liska <mliska@suse.cz>
733 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
734 (propagate_nothrow): Likewise.
736 2018-07-02 Martin Liska <mliska@suse.cz>
739 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
741 2018-07-02 David Malcolm <dmalcolm@redhat.com>
743 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
744 function in r262149, changing "loc" param from source_location to
745 const dump_location_t &.
746 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
747 declaration, as above.
749 2018-07-01 Paul Koning <ni1d@arrl.net>
751 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
752 -munit-asm, -mgnu-asm, -mdec-asm.
753 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
754 (pdp11_output_labelref): New.
755 (pdp11_output_def): New.
756 (pdp11_output_addr_vec_elt): New.
757 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
758 %# and %@ format codes.
759 (pdp11_option_override): New.
760 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
761 (pdp11_output_ident): New.
762 (pdp11_asm_named_section): New.
763 (pdp11_asm_init_sections): New.
764 (pdp11_file_start): New.
765 (pdp11_file_end): New.
766 (output_ascii): Use .ascii/.asciz for -mdec-asm.
767 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
768 %o, like %c but octal.
769 (pdp11_option_override): New.
770 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
772 (DATA_SECTION_ASM_OP): Ditto.
773 (READONLY_DATA_SECTION_ASM_OP): New.
774 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
775 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
776 (ASM_OUTPUT_LABELREF): Ditto.
777 (ASM_OUTPUT_DEF): Ditto.
778 (ASM_OUTPUT_EXTERNAL): New.
779 (ASM_OUTPUT_SOURCE_FILENAME): New.
780 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
781 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
782 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
783 %# and %@ format codes.
784 * config/pdp11/pdp11.opt (mgnu-asm): New.
785 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
786 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
787 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
789 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
791 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
792 dereferencing path[] beyond its length.
794 (debug_all_paths): New.
795 (rewire_first_differing_edge): New.
796 (adjust_paths_after_duplication): New.
797 (duplicate_thread_path): Call adjust_paths_after_duplication.
799 (thread_through_all_blocks): Add new argument to
800 duplicate_thread_path.
802 2018-06-30 Jim Wilson <jimw@sifive.com>
804 * config/riscv/predicates.md (p2m1_shift_operand): New.
805 (high_mask_shift_operand): New.
806 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
807 pattern using p2m1_shift_operand.
808 (lshsi3_zero_extend_3+2): New combiner pattern using
809 high_mask_shift_operand.
811 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
813 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
815 (vect_recog_rotate_pattern): ...here.
816 (vect_convert_input): Try to insert casts of invariants in the
818 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
819 preheader to be empty.
821 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
823 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
824 vector type. If given, install it in the new statement's
826 (vect_element_precision): New function.
827 (vect_unpromoted_value): New struct.
828 (vect_unpromoted_value::vect_unpromoted_value): New function.
829 (vect_unpromoted_value::set_op): Likewise.
830 (vect_look_through_possible_promotion): Likewise.
831 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
832 (vect_widened_op_tree, vect_convert_input): Likewise.
833 (vect_convert_inputs, vect_convert_output): Likewise.
834 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
835 to handle the optional cast of the multiplication result and
836 vect_widened_op_tree to detect the widened multiplication itself.
837 Do not require the input and output of promotion casts to have
838 the same sign, but base the signedness of the operation on the
839 input rather than the result. If the pattern includes two
840 promotions, check that those promotions have the same sign.
841 Do not restrict the MULT_EXPR handling to a double-width result;
842 handle quadruple-width results and wider. Use vect_convert_inputs
843 to convert the inputs to the common type.
844 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
845 to handle the optional cast of the ABS result. Also allow a sign
846 change or a sign extension between the ABS and MINUS.
847 Use vect_widened_op_tree to detect the widened subtraction and use
848 vect_convert_inputs to convert the inputs to the common type.
849 (vect_handle_widen_op_by_const): Delete.
850 (vect_recog_widen_op_pattern): New function.
851 (vect_recog_widen_mult_pattern): Use it.
852 (vect_recog_widen_shift_pattern): Likewise.
853 (vect_recog_widen_sum_pattern): Use
854 vect_look_through_possible_promotion to handle the promoted
857 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
859 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
860 the containing gimple_seq *.
861 * gimple-iterator.h (gsi_for_stmt): Declare it.
862 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
863 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
864 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
865 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
866 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
868 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
870 (vect_mark_pattern_stmts): ...here. Handle cases in which the
871 statement being replaced is part of an existing pattern
872 definition sequence, inserting the new pattern statements before
874 (vect_pattern_recog_1): Don't return a bool. If the statement
875 is already part of a pattern, instead apply pattern matching
876 to the pattern definition statements. Don't clear the
877 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
878 (vect_pattern_recog): Don't break after the first match;
879 continue processing the pattern definition statements instead.
880 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
882 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
884 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
885 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
886 (vect_recog_widen_sum_pattern): Use it.
888 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
890 * tree-vect-loop.c (vectorizable_reduction): Assert that the
891 phi is not a pattern statement and has not been replaced by
893 * tree-vect-patterns.c (type_conversion_p): Don't check
894 STMT_VINFO_IN_PATTERN_P.
895 (vect_recog_vector_vector_shift_pattern): Likewise.
896 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
897 the pattern statement rather than the original statement; check
898 directly for a WIDEN_MULT_EXPR here.
899 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
900 vect_is_simple_use to return the pattern statement rather
901 than the original statement; use is_pattern_stmt_p to check
902 for such a pattern statement.
903 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
904 to return the pattern statement rather than the original statement;
905 don't do the same transformation here.
906 (vect_is_simple_use): If the defining statement has been replaced
907 by a pattern statement, return the pattern statement instead.
908 Remove the corresponding (local) transformation from the vectype
911 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
913 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
914 end and default to null.
915 * tree-vect-loop.c (vect_create_epilog_for_reduction)
916 (vectorizable_reduction): Update calls accordingly, dropping the
917 gimple ** argument if the passed-back statement isn't needed.
918 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
919 (vect_recog_rotate_pattern): Likewise.
920 (vect_recog_mask_conversion_pattern): Likewise.
921 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
922 (vect_mask_constant_operand_p): Likewise.
923 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
924 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
925 (get_group_load_store_type, get_load_store_type): Likewise.
926 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
927 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
928 (vectorizable_conversion, vectorizable_assignment): Likewise.
929 (vectorizable_shift, vectorizable_operation): Likewise.
930 (vectorizable_store, vect_is_simple_cond): Likewise.
931 (vectorizable_condition, vectorizable_comparison): Likewise.
932 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
933 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
934 and move it to the end. Cope with null def_stmt_outs.
936 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
938 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
940 2018-06-29 Jeff Law <law@redhat.com>
942 * config/v850/v850.c (v850_legitimate_address_p): Handle large
943 displacements for TARGET_V850E2V3 and newer.
944 (TARGET_LRA_P): Remove. Defaults to LRA now.
945 * config/v850/v850.md (sign23byte_load): Remove.
946 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
947 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
949 2018-06-29 Martin Liska <mliska@suse.cz>
952 * coverage.c (coverage_init): Mangle full path name.
953 * doc/invoke.texi: Document the change.
954 * gcov-io.c (mangle_path): New.
955 * gcov-io.h (mangle_path): Likewise.
956 * gcov.c (mangle_name): Use mangle_path for path mangling.
958 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
960 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
961 if starting source register is not even.
963 2018-06-29 Martin Liska <mliska@suse.cz>
965 PR tree-optimization/86263
966 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
967 Make edge redirection.
969 2018-06-29 David Malcolm <dmalcolm@redhat.com>
971 * dumpfile.c (dump_loc): Add indentation based on scope depth.
972 (dump_scope_depth): New variable.
973 (get_dump_scope_depth): New function.
974 (dump_begin_scope): New function.
975 (dump_end_scope): New function.
976 * dumpfile.h (get_dump_scope_depth): New declaration.
977 (dump_begin_scope): New declaration.
978 (dump_end_scope): New declaration.
979 (class auto_dump_scope): New class.
980 (AUTO_DUMP_SCOPE): New macro.
981 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
984 2018-06-29 Richard Biener <rguenther@suse.de>
986 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
987 compute_all_dependences succeeds.
988 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
989 exceed --param loop-max-datarefs-for-datadeps.
991 2018-06-29 Jakub Jelinek <jakub@redhat.com>
993 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
995 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
998 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
999 alternative 0 in preferred_for_speed attribute.
1001 2018-06-28 Paul Koning <ni1d@arrl.net>
1003 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1004 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1005 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1006 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1007 for shift insn patterns.
1009 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1011 * cgraph.c (cgraph_node::get_body): Replace assignments to
1012 "dump_file" with calls to set_dump_file.
1013 * dumpfile.c (alt_dump_file): Make static, and group with...
1014 (alt_flags): ...this definition.
1015 (dumps_are_enabled): New variable.
1016 (refresh_dumps_are_enabled): New function.
1017 (set_dump_file): New function.
1018 (set_alt_dump_file): New function.
1019 (gcc::dump_manager::dump_start): Replace assignments to
1020 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1022 (gcc::dump_manager::dump_finish): Likewise.
1023 * dumpfile.h (alt_dump_file): Delete decl.
1024 (dumps_are_enabled): New variable decl.
1025 (set_dump_file): New function decl.
1026 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1028 * tree-nested.c (lower_nested_functions): Replace assignments to
1029 "dump_file" with calls to set_dump_file.
1031 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1033 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1034 goto_locus of each outgoing edge of each basic block.
1036 2018-06-28 Richard Biener <rguenther@suse.de>
1038 * dwarf2out.c (decl_scope_table): Remove.
1039 (push_decl_scope): Likewise.
1040 (pop_decl_scope): Likewise.
1041 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1042 (gen_struct_or_union_type_die): Likewise.
1043 (gen_tagged_type_die): Likewise.
1044 (dwarf2out_init): Do not initialize decl_scope_table.
1045 (dwarf2out_c_finalize): Do not free it.
1047 2018-06-28 Richard Biener <rguenther@suse.de>
1049 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1050 deciding whether to not re-use a DIE.
1052 2018-06-28 Richard Biener <rguenther@suse.de>
1054 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1055 DW_AT_abstract_origin attribute.
1057 2018-06-28 Martin Liska <mliska@suse.cz>
1059 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1060 Use newly introduced constants.
1061 * tree-switch-conversion.h (struct jump_table_cluster):
1062 Define max_ratio_for_size and max_ratio_for_speed.
1064 2018-06-28 Martin Liska <mliska@suse.cz>
1066 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1067 Add new checking assert to catch invalid state.
1068 (jump_table_cluster::can_be_handled): Handle single case
1070 (jump_table_cluster::is_beneficial): Bail out for such case.
1071 (bit_test_cluster::find_bit_tests):
1072 Add new checking assert to catch invalid state.
1073 (bit_test_cluster::can_be_handled): Handle single case
1075 (bit_test_cluster::is_beneficial): Bail out for such case.
1076 (switch_decision_tree::analyze_switch_statement):
1079 2018-06-28 Martin Liska <mliska@suse.cz>
1081 * common.opt: Introduce -completion option.
1082 * gcc.c (driver_handle_option): Handle it.
1083 (driver::main): Print completions if completion
1085 * opt-suggestions.c (option_proposer::get_completions):
1087 (option_proposer::suggest_completion): Likewise.
1088 (option_proposer::find_param_completions): Likewise.
1089 (verify_autocompletions): Likewise.
1090 (test_completion_valid_options): Likewise.
1091 (test_completion_valid_params): Likewise.
1092 (in_completion_p): Likewise.
1093 (empty_completion_p): Likewise.
1094 (test_completion_partial_match): Likewise.
1095 (test_completion_garbage): Likewise.
1096 (opt_proposer_c_tests): Likewise.
1097 * opt-suggestions.h: Declare new functions.
1098 * opts.c (common_handle_option): Handle OPT__completion_.
1099 * selftest-run-tests.c (selftest::run_tests): Add
1100 opt_proposer_c_tests.
1101 * selftest.c (assert_str_startswith): New.
1102 * selftest.h (assert_str_startswith): Likewise.
1103 (opt_proposer_c_tests): New.
1104 (ASSERT_STR_STARTSWITH): Likewise.
1106 2018-06-28 Martin Liska <mliska@suse.cz>
1108 * Makefile.in: Add opt-suggestions.o.
1109 * gcc-main.c: Include opt-suggestions.h.
1110 * gcc.c (driver::driver): Likewise.
1111 (driver::~driver): Remove m_option_suggestions.
1112 (driver::build_option_suggestions): Moved to option_proposer.
1113 (driver::suggest_option): Likewise.
1114 (driver::handle_unrecognized_options): Use option_proposer.
1115 * gcc.h (class driver): Add new memver m_option_proposer.
1116 * opt-suggestions.c: New file.
1117 * opt-suggestions.h: New file.
1119 2018-06-28 Martin Liska <mliska@suse.cz>
1121 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1122 (auto_string_vec::~auto_string_vec): Likewise.
1124 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1126 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1127 prevent_decl_creation_for_types fields up and add reset_location field.
1128 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1129 statement if id->reset_location is true.
1130 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1131 id->reset_location is true.
1132 (copy_phis_for_bb): Force input_location on the arguments if
1133 id->reset_location is true.
1134 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1135 is set on the function to be inlined.
1137 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1139 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1141 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1143 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1144 registers for Pmode.
1145 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1146 hard registers for the clobbered pseudo.
1148 2018-06-27 Paul Koning <ni1d@arrl.net>
1150 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1151 mutually exclusive options.
1152 * config/pdp11/constraints.md (h): New constraint.
1153 (O): Update definition to match shift code generation.
1154 (D): New constraint.
1155 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1157 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1159 (output_jump): Change arguments.
1160 (pdp11_fixed_cc_regs): New function.
1161 (pdp11_cc_mode): Ditto.
1162 (pdp11_expand_shift): Ditto.
1163 (pdp11_assemble_shift): Ditto.
1164 (pdp11_small_shift): Ditto.
1165 (pdp11_branch_cost): Remove.
1166 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1168 (pdp11_register_move_cost): Update for CC registers.
1169 (pdp11_rtx_costs): Add case for LSHIFTRT.
1170 (pdp11_output_jump): Add CCNZ mode conditional branches.
1171 (notice_update_cc_on_set): Remove.
1172 (pdp11_cc_mode): New function.
1173 (simple_memory_operand): Correct pre/post decrement case.
1174 (no_side_effect_operand): New function.
1175 (pdp11_regno_reg_class): Add CC_REGS class.
1176 (pdp11_fixed_cc_regs): New function.
1177 (pdp11_small_shift): New function.
1178 (pdp11_expand_shift): New function to expand shift insns.
1179 (pdp11_assemble_shift): New function to output shifts.
1180 (pdp11_branch_cost): Remove.
1181 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1182 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1183 (WCHAR_TYPE): Ditto.
1184 (PTRDIFF_TYPE): Ditto.
1185 (ADJUST_INSN_LENGTH): New macro.
1186 (FIXED_REGISTERS): Add CC registers.
1187 (CALL_USED_REGISTERS): Ditto.
1189 (REG_CLASS_NAMES): Ditto.
1190 (REG_CLASS_CONTENTS): Ditto.
1191 (SELECT_CC_MODE): Use new function.
1192 (TARGET_FLAGS_REGNUM): New macro.
1193 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1194 (cc0_reg_rtx): Remove.
1195 (CC_STATUS_MDEP): Remove.
1196 (CC_STATUS_MDEFP_INIT): Remove.
1197 (CC_IN_FPU): Remove.
1198 (NOTICE_UPDATE_CC): Remove.
1199 (REGISTER_NAMES): Add CC registers.
1200 (BRANCH_COST): Change to constant 1.
1201 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1203 * config/pdp11/pdp11.opt (mbcopy): Remove.
1204 (mbcopy-builtin): Remove.
1205 (mbranch-cheap): Remove.
1206 (mbranch-expensive): Remove.
1207 * config/pdp11/predicates.md (expand_shift_operand): Update to
1208 match shift code generation.
1209 (ccnz_operator): New predicate.
1210 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1211 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1212 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1214 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1215 description of O constraint.
1217 2018-06-27 Jeff Law <law@redhat.com>
1218 Austin Law <austinklaw@gmail.com>
1220 * config/v850/v850.md (addsi3_set_flags): New pattern.
1221 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1222 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1223 (zero_extendhisi2_v850_set_flags): Likewise.
1224 (zero_extendqisi2_v850_set_flags): Likewise.
1225 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1226 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1227 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1229 * config/v850/v850-protos.h (notice_update_cc): Remove.
1230 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1231 (v850_print_operand): Handle 'D' and "d".
1232 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1233 Add handling of arithmetic/logical operations compared against zero.
1234 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1235 Do not look at v850_compare_op, instead get mode from last argument.
1236 (v850_gen_compare): Remove
1237 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1238 after reload for prologue insns.
1239 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1241 (construct_save_jarl): Likewise.
1242 (TARGET_FLAGS_REGNUM): Define.
1243 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1244 (NOTICE_UPDATE_CC): Remove.
1245 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1246 than cc0. Conditionalize on reload_completed.
1247 (cmpsi_insn, setfcc_insn): Likewise.
1248 (tst1 splitter): Turn into define_and_split which sets the flags
1250 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1251 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1252 (cstoresf4, cstoredf4): Clobber the flags.
1253 (cmpsi, cmpsf, cmpdf): Remove expanders.
1254 (setf_insn): Remove pattern.
1255 (addsi3): Turn into define_and_split which clobbers the flags after
1256 reload and a suitable pattern (addsi3_clobber_flags) for use after
1258 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1259 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1260 (ashrsi3, ashrsi3_v850e2): Likewise.
1261 (bins): Clobber the flags.
1262 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1263 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1264 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1265 (call_value_internal_short, call_value_internal_long): Likewise.
1266 (callt_save_interrupt, callt_return_interrupt): Likewise.
1267 (save_interrupt, return_interrupt): Likewise.
1268 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1269 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1270 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1271 (All FP comparisons): Only allow after reload has completed.
1273 (divh, divhu): Tweak output template.
1274 (branch_z_normal, branch_z_invert): Remove
1275 (branch_nz_normal, branch_nz_invert): Likewise.
1276 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1278 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1279 * config/v850/v850.c (notice_update_cc): Remove.
1280 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1281 (CC_NO_CARRY): Likewise.
1282 (NOTICE_UPDATE_CC): Define to nothing.
1283 * config/v850/v850.md: Remove block comment on cc0 handling
1284 Remove "cc" attribute from all patterns. Remove cc_status handling
1285 from all patterns. Minor formatting fixes.
1287 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1289 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1290 (cortex-a76.cortex-a55): Likewise.
1291 * config/aarch64/aarch64-tune.md: Regenerate.
1292 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1293 cortex-a76.cortex-a55.
1295 2018-06-27 Jeff Law <law@redhat.com>
1297 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1298 (MULTILIB_DIRNAMES): Similarly.
1300 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1302 * gimple.h (gimple_return_retbnd): Delete.
1303 (gimple_return_set_retbnd): Likewise.
1304 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1305 gimple_return_set_retbnd.
1306 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1307 gimple_return_retbnd and adjust.
1308 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1309 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1310 Explicitly return NULL in a couple more cases. Move assertion
1311 on debug statements and remove unreachable code.
1312 (reset_debug_binding): Do not test id->retbnd.
1313 (expand_call_inline): Do not set it.
1315 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1317 * configure.ac: Add --disable-gcov option.
1318 * configure: Regenerate.
1319 * Makefile.in: Honour @enable_gcov@.
1320 * doc/install.texi: Document --disable-gcov.
1322 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1324 * config/arm/arm-cpus.in (cortex-a76): New entry.
1325 (cortex-a76.cortex-a55): Likewise.
1326 * config/arm/arm-tables.opt: Regenerate.
1327 * config/arm/arm-tune.md: Likewise.
1328 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1329 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1330 cortex-a76.cortex-a55.
1332 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1335 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1337 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1339 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1341 (EPILOGUE_USES): Likewise.
1343 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1345 * tree-inline.c (remap_location): New function extracted from...
1346 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1347 (copy_phis_for_bb): ...here. Call remap_location.
1348 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1350 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1352 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1353 unaligned vsx for 16B memset.
1355 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1358 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1359 ieee128_float_type_node to long_double_type_node unless
1360 TARGET_LONG_DOUBLE_128 is set.
1362 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1364 * cfgloop.c (get_loop_location): Convert return type from
1365 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1366 by implicit construction from rtx_insn *, and using
1367 dump_user_location_t::from_function_decl for the fallback case.
1368 * cfgloop.h (get_loop_location): Convert return type from
1369 location_t to dump_user_location_t.
1370 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1371 dump_printf_loc to pass in a dump_location_t rather than a
1372 location_t, via the gimple stmt.
1373 * coverage.c (get_coverage_counts): Update calls to
1374 dump_printf_loc to pass in dump_location_t rather than a
1376 * doc/optinfo.texi (Dump types): Convert example of
1377 dump_printf_loc from taking "locus" to taking "insn". Update
1378 description of the "_loc" calls to cover dump_location_t.
1379 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1381 (dump_user_location_t::dump_user_location_t): New constructors,
1382 from gimple *stmt and rtx_insn *.
1383 (dump_user_location_t::from_function_decl): New function.
1384 (dump_loc): Make static.
1385 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1386 const dump_location_t &.
1387 (dump_generic_expr_loc): Delete.
1388 (dump_printf_loc): Convert param "loc" from location_t to
1389 const dump_location_t &.
1390 (selftest::test_impl_location): New function.
1391 (selftest::dumpfile_c_tests): New function.
1392 * dumpfile.h: Include "profile-count.h".
1393 (class dump_user_location_t): New class.
1394 (struct dump_impl_location_t): New struct.
1395 (class dump_location_t): New class.
1396 (dump_printf_loc): Convert 2nd param from source_location to
1397 const dump_location_t &.
1398 (dump_generic_expr_loc): Delete.
1399 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1400 const dump_location_t &.
1401 * gimple-fold.c (fold_gimple_assign): Update call to
1402 dump_printf_loc to pass in a dump_location_t rather than a
1403 location_t, via the gimple stmt.
1404 (gimple_fold_call): Likewise.
1405 * gimple-loop-interchange.cc
1406 (loop_cand::analyze_iloop_reduction_var): Update for change to
1407 check_reduction_path.
1408 (tree_loop_interchange::interchange): Update for change to
1410 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1411 change in return-type of find_loop_location.
1412 (graphite_regenerate_ast_isl): Likewise.
1413 * graphite-optimize-isl.c (optimize_isl): Likewise.
1414 * graphite.c (graphite_transform_loops): Likewise.
1415 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1416 pass in a dump_location_t rather than a location_t, via the
1418 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1419 * ipa.c (walk_polymorphic_call_targets): Likewise.
1420 * loop-unroll.c (report_unroll): Convert "locus" param from
1421 location_t to dump_location_t.
1422 (decide_unrolling): Update for change to get_loop_location's
1424 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1425 location_t to dump_user_location_t.
1426 (grid_find_single_omp_among_assignments_1): Updates calls to
1427 dump_printf_loc to pass in a dump_location_t rather than a
1428 location_t, via the gimple stmt.
1429 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1430 location_t to dump_location_t. Updates calls to dump_printf_loc
1431 to pass in a dump_location_t rather than a location_t, via the
1433 (grid_inner_loop_gridifiable_p): Likewise.
1434 (grid_dist_follows_simple_pattern): Likewise.
1435 (grid_gfor_follows_tiling_pattern): Likewise.
1436 (grid_target_follows_gridifiable_pattern): Likewise.
1437 (grid_attempt_target_gridification): Convert initialization
1438 of local "grid" from memset to zero-initialization; FIXME: does
1439 this require C++11? Update call to dump_printf_loc to pass in a
1440 optinfo_location rather than a location_t, via the gimple stmt.
1441 * profile.c (read_profile_edge_counts): Updates call to
1442 dump_printf_loc to pass in a dump_location_t rather than a
1444 (compute_branch_probabilities): Likewise.
1445 * selftest-run-tests.c (selftest::run_tests): Call
1447 * selftest.h (dumpfile_c_tests): New decl.
1448 * tree-loop-distribution.c (pass_loop_distribution::execute):
1449 Update for change in return type of find_loop_location.
1450 * tree-parloops.c (parallelize_loops): Likewise.
1451 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1452 "locus" from location_t to dump_user_location_t.
1453 (canonicalize_loop_induction_variables): Likewise.
1454 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1455 for change in return type of find_loop_location.
1456 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1457 to dump_printf_loc to pass in a dump_location_t rather than a
1458 location_t, via the stmt.
1459 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1461 * tree-vect-loop-manip.c (find_loop_location): Convert return
1462 type from source_location to dump_user_location_t.
1463 (vect_do_peeling): Update for above change.
1464 (vect_loop_versioning): Update for change in type of
1466 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1467 from location_t to dump_user_location_t.
1468 (vect_estimate_min_profitable_iters): Update for change in type
1470 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1471 location_t to dump_location_t.
1472 (vect_slp_bb): Update for change in type of vect_location.
1473 * tree-vectorizer.c (vect_location): Convert from source_location
1474 to dump_user_location_t.
1475 (try_vectorize_loop_1): Update for change in vect_location's type.
1476 (vectorize_loops): Likewise.
1477 (increase_alignment): Likewise.
1478 * tree-vectorizer.h (vect_location): Convert from source_location
1479 to dump_user_location_t.
1480 (find_loop_location): Convert return type from source_location to
1481 dump_user_location_t.
1482 (check_reduction_path): Convert 1st param from location_t to
1483 dump_user_location_t.
1484 * value-prof.c (check_counter): Update call to dump_printf_loc to
1485 pass in a dump_user_location_t rather than a location_t; update
1486 call to error_at for change in type of "locus".
1487 (check_ic_target): Update call to dump_printf_loc to
1488 pass in a dump_user_location_t rather than a location_t, via the
1491 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1493 * config/s390/s390.h (enum processor_flags): Do not use
1494 default tune parameter when -march was specified.
1496 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1499 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1500 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1503 2018-06-26 Richard Biener <rguenther@suse.de>
1505 PR tree-optimization/86287
1507 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1508 (vect_analyze_loop): Initialize n_stmts.
1510 2018-06-26 Richard Biener <rguenther@suse.de>
1513 * fold-const.c (fold_convertible_p): Pointer extension
1516 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1519 * dwarf2out.c (loc_list_has_views): Adjust comments.
1520 (dw_loc_list): Split single cross-partition range with
1523 2018-06-25 Jeff Law <law@redhat.com>
1525 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1526 on -mbig-switch by default.
1528 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1530 (const_float_0_operand): Remove unused predicate.
1531 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1532 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1533 (recipsf2): New expander. Original pattern now called
1535 (recipdf2, recipdf2_insn): Similarly.
1536 (rsqrtsf2, rsqrtsf2_insn): Similarly
1537 (rsqrtdf2, rsqrtdf2_insn): Similarly
1539 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1541 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1542 Simplify logic for FreeBSD (twice).
1544 2018-06-25 Martin Sebor <msebor@redhat.com>
1546 PR tree-optimization/86204
1547 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1548 a strnlen result if it's less than the length of the string.
1550 2018-06-25 Martin Sebor <msebor@redhat.com>
1552 PR tree-optimization/85700
1553 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1554 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1555 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1557 2018-06-25 Martin Sebor <msebor@redhat.com>
1559 * doc/extend.texi (Zero-length arrays): Update and clarify.
1561 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1563 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1564 added IEEE/IBM long double multilib support on PowerPC little
1565 endian Linux systems.
1566 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1567 (MULTILIB_DEFAULTS): Likewise.
1568 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1570 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1571 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1572 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1574 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1577 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1578 (REORDER_45): Likewise.
1580 2018-06-25 Jeff Law <law@redhat.com>
1582 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1583 dividend to 32 bits. Adjust length.
1584 (udivmodhi4): Cleanup output template. Fix length.
1586 2018-06-25 Carl Love <cel@us.ibm.com>
1588 * config/rs6000/vsx.md: Change word selector to prefered location.
1590 2018-06-25 Richard Biener <rguenther@suse.de>
1592 PR tree-optimization/86304
1593 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1594 epilogue-if-converted loops as well.
1596 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1598 * lto-section-out.c (lto_begin_section): Do not print section
1599 name for noaddr and unnumbered dumps.
1601 2018-06-25 Richard Biener <rguenther@suse.de>
1603 * tree-vectorizer.h (struct vec_info_shared): New structure
1604 with parts split out from struct vec_info and loop_nest from
1605 struct _loop_vec_info.
1606 (struct vec_info): Adjust accordingly.
1607 (struct _loop_vec_info): Likewise.
1608 (LOOP_VINFO_LOOP_NEST): Adjust.
1609 (LOOP_VINFO_DATAREFS): Likewise.
1610 (LOOP_VINFO_DDRS): Likewise.
1611 (struct _bb_vec_info): Likewise.
1612 (BB_VINFO_DATAREFS): Likewise.
1613 (BB_VINFO_DDRS): Likewise.
1614 (struct _stmt_vec_info): Add dr_aux member.
1615 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1616 (DR_MISALIGNMENT_UNINITIALIZED): New.
1617 (set_dr_misalignment): Adjust.
1618 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1619 (vect_analyze_loop): Adjust prototype.
1620 (vect_analyze_loop_form): Likewise.
1621 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1622 Compute dependences lazily.
1623 (vect_record_base_alignments): Use shared datarefs/ddrs.
1624 (vect_verify_datarefs_alignment): Likewise.
1625 (vect_analyze_data_refs_alignment): Likewise.
1626 (vect_analyze_data_ref_accesses): Likewise.
1627 (vect_analyze_data_refs): Likewise.
1628 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1629 constructor parameter for shared part.
1630 (vect_analyze_loop_form): Pass in shared part and adjust.
1631 (vect_analyze_loop_2): Pass in storage for the number of
1632 stmts. Move loop nest finding to the caller. Compute
1634 (vect_analyze_loop): Pass in shared part.
1635 (vect_transform_loop): Verify shared datarefs are unchanged.
1636 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1637 constructor parameter for shared part.
1638 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1639 (vect_slp_bb): Verify shared datarefs are unchanged before
1641 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1643 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1644 DR_MISALIGNMENT_UNINITIALIZED.
1645 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1646 parameter for shared part.
1647 (vec_info::~vec_info): Adjust.
1648 (vec_info_shared::vec_info_shared): New.
1649 (vec_info_shared::~vec_info_shared): Likewise.
1650 (vec_info_shared::save_datarefs): Likewise.
1651 (vec_info_shared::check_datarefs): Likewise.
1652 (try_vectorize_loop_1): Construct shared part live for analyses
1653 of a single loop for multiple vector sizes.
1654 * tree-parloops.c (gather_scalar_reductions): Adjust.
1656 2018-06-25 Richard Biener <rguenther@suse.de>
1658 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1659 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1660 (vect_analyze_data_refs): Remove similar code from here and
1661 simplify accordingly.
1663 2018-06-25 Richard Biener <rguenther@suse.de>
1665 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1666 for reverse storage order accesses rather than asserting
1667 they cannot happen here.
1669 2018-06-25 Tom de Vries <tdevries@suse.de>
1672 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1673 Use data16 instead of .byte for insn prefix.
1675 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1678 * parser.c (make_char_string_pack): Pass this literal chars
1679 through cpp_interpret_string.
1680 (cp_parser_userdef_numeric_literal): Check the result of
1681 make_char_string_pack.
1683 2018-06-24 Maya Rashish <coypu@sdf.org>
1685 * ginclude/stddef.h: Simplify conditions around avoiding
1686 re-definition of __size_t.
1688 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1690 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1691 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1693 2018-06-22 Maya Rashish <coypu@sdf.org>
1695 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1696 made to OpenBSD and NetBSD through the years.
1698 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1700 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1701 behavior of vec_pack (vector double, vector double) to match
1702 behavior of vec_float2 (vector double, vector double).
1704 2018-06-22 Olivier Hainque <hainque@adacore.com>
1706 * gimplify.c (gimplify_function_tree): Prevent creation
1707 of a trampoline for the address of the current function
1708 passed to entry/exit instrumentation hooks.
1710 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1713 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1716 2018-06-22 Martin Liska <mliska@suse.cz>
1718 PR tree-optimization/86263
1719 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1720 Bail out if is_enabled is false.
1721 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1723 (jump_table_cluster::is_enabled): New function.
1725 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1727 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1728 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1729 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1730 (lto_input_ts_binfo_tree_pointers): Likewise.
1731 * tree-streamer-out.c (streamer_write_tree_bitfields,
1732 write_ts_binfo_tree_pointers): Likewise.
1733 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1735 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1737 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1739 2018-06-22 Martin Liska <mliska@suse.cz>
1741 * symbol-summary.h (get): Make it pure and inline move
1742 functionality from ::get function.
1743 (get): Remove and inline into ::get and ::get_create.
1744 (get_create): Move code from ::get function.
1746 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1749 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1750 -x assembler-with-cpp.
1752 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1754 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1755 _FILE_OFFSET_BITS=64 for C++.
1757 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1759 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1760 conversion insn that shows up when pr85657-3.c is compiled using
1761 IEEE 128-bit long double.
1762 (neg<mode>2_internal): Use the correct mode to check whether the
1763 mode is IBM extended.
1764 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1765 multiply and divide external functions from being created more
1768 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1770 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1772 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1773 the edge can be forwarded.
1774 (cfg_layout_merge_blocks): Likewise.
1776 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1778 * except.c (finish_eh_generation): Commit edge insertions only after
1779 the EH edges have been redirected from post-landing to landing pads.
1781 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1783 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1784 create_tmp_var_for to create the FRAME decl.
1785 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1787 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1789 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1790 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1791 debug statement when resetting its value.
1792 (expand_call_inline): Copy the locus of the call onto the assignment
1793 of the return value, if any. Use local variable in more cases.
1795 2018-06-21 Martin Liska <mliska@suse.cz>
1797 * ipa-pure-const.c (propagate_nothrow): Use
1798 funct_state_summaries->get.
1799 (dump_malloc_lattice): Likewise.
1800 (propagate_malloc): Likewise.
1802 2018-06-21 Richard Biener <rguenther@suse.de>
1804 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1805 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1806 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1808 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1809 BLOCK_ABSTRACT_ORIGIN unconditionally.
1811 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1813 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1815 * ipa-reference.c (ipa_reference_c_finalize): Delete
1816 ipa_ref_opt_sum_summaries and set it to NULL.
1818 2018-06-21 Tom de Vries <tdevries@suse.de>
1820 PR tree-optimization/85859
1821 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1822 test with comment from bb_no_side_effects_p.
1824 2018-06-21 Richard Biener <rguenther@suse.de>
1826 PR tree-optimization/86232
1827 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1828 max for constant niter.
1830 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1832 * config/aarch64/aarch64-simd.md
1833 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1835 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1837 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1838 Make opernads of the unspec commutative.
1840 2018-06-21 Richard Biener <rguenther@suse.de>
1842 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1843 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1844 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1845 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1847 (vect_record_base_alignments): Do not record base alignment
1848 for gathers or scatters.
1849 (vect_compute_data_ref_alignment): Drop return value that is always
1850 true. Bail out early for gathers or scatters.
1851 (vect_enhance_data_refs_alignment): Bail out early for gathers
1853 (vect_find_same_alignment_drs): Likewise.
1854 (vect_analyze_data_refs_alignment): Remove dead code.
1855 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1856 (vect_analyze_data_refs): For possible gathers or scatters do
1857 not create an alternate DR, just check their possible validity
1858 and mark them. Adjust DECL_NONALIASED handling to not rely
1860 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1861 update inits of gathers or scatters.
1862 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1863 Also copy gather/scatter flag to pattern vinfo.
1865 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1867 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1868 behavior of vec_packsu (vector unsigned long long, vector unsigned
1869 long long) to match behavior of vec_packs with same signature.
1871 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1872 Thomas Schwinge <thomas@codesourcery.com>
1873 Cesar Philippidis <cesar@codesourcery.com>
1875 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1876 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1877 (gimplify_adjust_omp_clauses): Likewise.
1878 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1879 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1880 (gimplify_omp_target_update): Update handling of acc update and
1882 * omp-low.c (install_var_field): Remove unused parameter
1883 base_pointers_restrict.
1884 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1885 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1887 (omp_target_base_pointers_restrict_p): Delete.
1888 (scan_omp_target): Update call to scan_sharing_clauses.
1889 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1891 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1892 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1893 (convert_local_omp_clauses): Likewise.
1894 * tree-pretty-print.c (dump_omp_clause): Likewise.
1895 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1897 (omp_clause_code_name): Likewise.
1899 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1902 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1905 PR tree-optimization/86231
1906 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1907 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1909 2018-06-20 Tom de Vries <tdevries@suse.de>
1911 PR tree-optimization/86097
1912 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1913 iv type if signedness of iv type is not the same as that of *nit.
1915 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1917 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1918 EDGE_EH edges, verify they are all EDGE_EH.
1920 2018-06-20 Maya Rashish <coypu@sdf.org>
1922 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1924 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1926 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1927 * config/aarch64/aarch64.c (xgene1_tunings): Add
1928 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1929 (aarch64_mode_valid_for_sched_fusion_p):
1930 Allow 16-byte modes.
1931 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1932 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1934 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1936 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1937 * config/aarch64/iterators.md (VQ2): New mode iterator.
1939 2018-06-20 Martin Liska <mliska@suse.cz>
1941 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1942 Change default ratio from 10 to 8.
1944 2018-06-20 Martin Liska <mliska@suse.cz>
1946 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1948 (bit_test_cluster::find_bit_tests): Likewise.
1949 (switch_decision_tree::analyze_switch_statement): Find clusters.
1950 * tree-switch-conversion.h (struct jump_table_cluster): Document
1953 2018-06-20 Martin Liska <mliska@suse.cz>
1955 * tree-switch-conversion.c (switch_conversion::collect):
1956 Record m_uniq property.
1957 (switch_conversion::expand): Bail out for special conditions.
1958 (group_cluster::~group_cluster): New.
1959 (group_cluster::group_cluster): Likewise.
1960 (group_cluster::dump): Likewise.
1961 (jump_table_cluster::emit): New.
1962 (switch_decision_tree::fix_phi_operands_for_edges): New.
1963 (struct case_node): Remove struct.
1964 (jump_table_cluster::can_be_handled): New.
1965 (case_values_threshold): Moved to header.
1966 (reset_out_edges_aux): Likewise.
1967 (jump_table_cluster::is_beneficial): New.
1968 (bit_test_cluster::can_be_handled): Likewise.
1969 (add_case_node): Remove.
1970 (bit_test_cluster::is_beneficial): New.
1971 (case_bit_test::cmp): New.
1972 (bit_test_cluster::emit): New.
1973 (expand_switch_as_decision_tree_p): Remove.
1974 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1975 (fix_phi_operands_for_edge): Likewise.
1976 (switch_decision_tree::analyze_switch_statement): New.
1977 (compute_cases_per_edge): Move ...
1978 (switch_decision_tree::compute_cases_per_edge): ... here.
1979 (try_switch_expansion): Likewise.
1980 (switch_decision_tree::try_switch_expansion): Likewise.
1981 (record_phi_operand_mapping): Likewise.
1982 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1983 (emit_case_decision_tree): Likewise.
1984 (switch_decision_tree::emit): Likewise.
1985 (balance_case_nodes): Likewise.
1986 (switch_decision_tree::balance_case_nodes): Likewise.
1987 (dump_case_nodes): Likewise.
1988 (switch_decision_tree::dump_case_nodes): Likewise.
1989 (emit_jump): Likewise.
1990 (switch_decision_tree::emit_jump): Likewise.
1991 (emit_cmp_and_jump_insns): Likewise.
1992 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1993 (emit_case_nodes): Likewise.
1994 (switch_decision_tree::emit_case_nodes): Likewise.
1995 (conditional_probability): Remove.
1996 * tree-switch-conversion.h (enum cluster_type): New.
1998 (struct cluster): Likewise.
1999 (cluster::cluster): Likewise.
2000 (struct simple_cluster): Likewise.
2001 (simple_cluster::simple_cluster): Likewise.
2002 (struct group_cluster): Likewise.
2003 (struct jump_table_cluster): Likewise.
2004 (struct bit_test_cluster): Likewise.
2005 (struct min_cluster_item): Likewise.
2006 (struct case_tree_node): Likewise.
2007 (case_tree_node::case_tree_node): Likewise.
2008 (jump_table_cluster::case_values_threshold): Likewise.
2009 (struct case_bit_test): Likewise.
2010 (struct switch_decision_tree): Likewise.
2011 (struct switch_conversion): Likewise.
2012 (switch_decision_tree::reset_out_edges_aux): Likewise.
2014 2018-06-20 Martin Liska <mliska@suse.cz>
2016 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2017 (hoist_edge_and_branch_if_true): Likewise.
2018 (expand_switch_using_bit_tests_p): Likewise.
2019 (struct case_bit_test): Likewise.
2020 (case_bit_test_cmp): Likewise.
2021 (emit_case_bit_tests): Likewise.
2022 (switch_conversion::switch_conversion): New class.
2023 (struct switch_conv_info): Remove old struct.
2024 (collect_switch_conv_info): More to ...
2025 (switch_conversion::collect): ... this.
2026 (check_range): Likewise.
2027 (switch_conversion::check_range): Likewise.
2028 (check_all_empty_except_final): Likewise.
2029 (switch_conversion::check_all_empty_except_final): Likewise.
2030 (check_final_bb): Likewise.
2031 (switch_conversion::check_final_bb): Likewise.
2032 (create_temp_arrays): Likewise.
2033 (switch_conversion::create_temp_arrays): Likewise.
2034 (free_temp_arrays): Likewise.
2035 (gather_default_values): Likewise.
2036 (switch_conversion::gather_default_values): Likewise.
2037 (build_constructors): Likewise.
2038 (switch_conversion::build_constructors): Likewise.
2039 (constructor_contains_same_values_p): Likewise.
2040 (switch_conversion::contains_same_values_p): Likewise.
2041 (array_value_type): Likewise.
2042 (switch_conversion::array_value_type): Likewise.
2043 (build_one_array): Likewise.
2044 (switch_conversion::build_one_array): Likewise.
2045 (build_arrays): Likewise.
2046 (switch_conversion::build_arrays): Likewise.
2047 (gen_def_assigns): Likewise.
2048 (switch_conversion::gen_def_assigns): Likewise.
2049 (prune_bbs): Likewise.
2050 (switch_conversion::prune_bbs): Likewise.
2051 (fix_phi_nodes): Likewise.
2052 (switch_conversion::fix_phi_nodes): Likewise.
2053 (gen_inbound_check): Likewise.
2054 (switch_conversion::gen_inbound_check): Likewise.
2055 (process_switch): Use the newly created class.
2056 (switch_conversion::expand): New.
2057 (switch_conversion::~switch_conversion): New.
2058 * tree-switch-conversion.h: New file.
2060 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2062 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2063 tree-vect-patterns.c.
2064 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2065 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2066 (vect_recog_sad_pattern): Likewise.
2067 (vect_recog_widen_sum_pattern): Likewise.
2068 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2069 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2070 (vect_recog_rotate_pattern): Likewise.
2071 (vect_recog_mult_pattern): Likewise.
2072 (vect_recog_vector_vector_shift_pattern): Likewise.
2073 (vect_recog_divmod_pattern): Likewise.
2074 (vect_recog_mixed_size_cond_pattern): Likewise.
2075 (vect_recog_bool_pattern): Likewise.
2076 (vect_recog_mask_conversion_pattern): Likewise.
2077 (vect_try_gather_scatter_pattern): Likewise.
2078 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2079 (vect_recog_over_widening_pattern): Likewise.
2080 (vect_recog_gather_scatter_pattern): Likewise.
2081 (vect_recog_func_ptr): Move from tree-vectorizer.h
2082 (vect_vect_recog_func_ptrs): Move further down the file.
2083 (vect_recog_func): Likewise. Remove the third argument.
2084 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2085 (vect_pattern_recog_1): Expect the pattern function to do any
2086 necessary target tests. Also expect it to provide a vector type.
2087 Remove the type_in handling.
2089 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2091 * tree-vect-patterns.c (vect_pattern_detected): New function.
2092 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2093 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2094 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2095 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2096 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2097 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2098 (vect_recog_mask_conversion_pattern)
2099 (vect_try_gather_scatter_pattern): Likewise.
2101 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2103 * tree-vect-patterns.c (vect_get_internal_def): New function.
2104 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2105 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2106 (search_type_for_mask_1): Use it.
2108 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2110 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2111 redundant WIDEN_SUM_EXPR handling.
2112 (vect_recog_sad_pattern): Likewise.
2114 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2116 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2117 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2118 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2119 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2120 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2122 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2124 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2125 use the stmt_vec_info of the original bb statement for the
2126 new zero assignment, even if the call is part of a pattern.
2128 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2130 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2131 that the sequence is attached to the original statement rather
2132 than the pattern statement.
2133 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2134 PATTERN_DEF_SEQ from the original statement rather than
2135 the main pattern statement.
2136 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2137 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2138 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2140 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2142 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2143 definition statements before the early exit for statements that aren't
2145 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2147 (vect_transform_loop): ...here. Process pattern definition
2148 statements without first checking whether the main pattern
2149 statement is live or relevant.
2151 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2153 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2154 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2156 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2158 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2159 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2160 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2161 (expand_block_compare): Change select_block_compare_mode call.
2162 (expand_strncmp_align_check): Use new functions, fix comment.
2163 (emit_final_str_compare_gpr): New function.
2164 (expand_strn_compare): Refactor and clean up code.
2165 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2167 2018-06-19 Tony Reix <tony.reix@atos.com>
2168 Damien Bergamini <damien.bergamini@atos.com>
2169 David Edelsohn <dje.gcc@gmail.com>
2171 * collect2.c (static_obj): New variable.
2172 (static_libs): New variable.
2173 (is_in_list): Uncomment declaration.
2174 (main): Track AIX libraries linked statically.
2175 (is_in_list): Uncomment definition.
2176 (scan_prog_file): Don't add AIX shared libraries initializer
2177 to constructor list if linking statically.
2179 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2181 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2183 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2185 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2187 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2190 2018-06-19 Martin Liska <mliska@suse.cz>
2192 * config/i386/i386.c (ix86_can_inline_p): Do not use
2193 ipa_fn_summaries::get_create.
2194 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2196 (devirtualization_time_bonus): Likewise.
2197 (ipcp_propagate_stage): Likewise.
2198 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2199 (edge_set_predicate): Likewise.
2200 (evaluate_conditions_for_known_args): Likewise.
2201 (evaluate_properties_for_edge): Likewise.
2202 (ipa_call_summary::reset): Tranform to ...
2203 (ipa_call_summary::~ipa_call_summary): ... this.
2204 (ipa_fn_summary::reset): Transform to ...
2205 (ipa_fn_summary::~ipa_fn_summary): ... this.
2206 (ipa_fn_summary_t::remove): Rename to ...
2207 (ipa_fn_summary_t::remove_callees): ... this.
2208 (ipa_fn_summary_t::duplicate): Use placement new
2209 instead of memory copy.
2210 (ipa_call_summary_t::duplicate): Likewise.
2211 (ipa_call_summary_t::remove): Remove.
2212 (dump_ipa_call_summary): Change get_create to get.
2213 (ipa_dump_fn_summary): Dump only when summary exists.
2214 (analyze_function_body): Use symbol_summary::get instead
2216 (compute_fn_summary): Likewise.
2217 (estimate_edge_devirt_benefit): Likewise.
2218 (estimate_edge_size_and_time): Likewise.
2219 (inline_update_callee_summaries): Likewise.
2220 (remap_edge_change_prob): Likewise.
2221 (remap_edge_summaries): Likewise.
2222 (ipa_merge_fn_summary_after_inlining): Likewise.
2223 (write_ipa_call_summary): Likewise.
2224 (ipa_fn_summary_write): Likewise.
2225 (ipa_free_fn_summary): Likewise.
2226 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2227 (struct ipa_call_summary): Likewise.
2228 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2230 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2231 (estimate_size_after_inlining): Likewise.
2232 (estimate_growth): Likewise.
2233 (growth_likely_positive): Likewise.
2234 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2235 (inline_call): Likewise.
2236 * ipa-inline.c (caller_growth_limits): Likewise.
2237 (can_inline_edge_p): Likewise.
2238 (can_inline_edge_by_limits_p): Likewise.
2239 (compute_uninlined_call_time): Likewise.
2240 (compute_inlined_call_time): Likewise.
2241 (want_inline_small_function_p): Likewise.
2242 (edge_badness): Likewise.
2243 (update_caller_keys): Likewise.
2244 (update_callee_keys): Likewise.
2245 (inline_small_functions): Likewise.
2246 (inline_to_all_callers_1): Likewise.
2247 (dump_overall_stats): Likewise.
2248 (early_inline_small_functions): Likewise.
2249 (early_inliner): Likewise.
2250 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2251 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2252 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2253 * ipa-split.c (execute_split_functions): Likewise.
2254 * symbol-summary.h: Likewise.
2255 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2257 2018-06-19 Richard Biener <rguenther@suse.de>
2259 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2260 (vectorize_loops): ... here. Fix dbgcnt handling.
2261 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2263 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2266 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2267 ieee128 argument takes up only one (vector) register, not two (floating
2270 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2272 * gimplify.c (gimplify_init_constructor): Really never clear for an
2273 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2275 2018-06-19 Richard Biener <rguenther@suse.de>
2277 PR tree-optimization/86179
2278 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2279 after failed recognition.
2281 2018-06-18 Martin Sebor <msebor@redhat.com>
2284 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2285 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2286 Handle integer subtraction.
2287 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2288 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2290 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2292 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2293 param from rtx to rtx_insn *.
2294 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2296 (frv_ifcvt_modify_insn): Likwise.
2297 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2298 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2299 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2300 as_a <rtx_insn *> cast to local "unprotected_region" once
2301 it's been established that it's not NULL or pc_rtx.
2302 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2303 param "sethi" from rtx to rtx_insn *.
2304 (nds32_group_float_insns): Likewise for param "insn".
2305 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2307 (vax_output_int_subtract): Likewise.
2308 * config/vax/vax.c (vax_output_int_add): Likewise for param
2310 (vax_output_int_subtract): Likewise.
2311 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2312 (emit_pattern_after): Likewise for param "after".
2313 (emit_insn_after): Likewise.
2314 (emit_jump_insn_after): Likewise.
2315 (emit_call_insn_after): Likewise.
2316 (emit_debug_insn_after): Likewise.
2317 (emit_pattern_before): Likewise for param "before".
2318 (emit_insn_before): Likewise.
2319 (emit_jump_insn_before): Likewise.
2320 * final.c (get_insn_template): Likewise for param "insn", removing
2322 * output.h (get_insn_template): Likewise for 2nd param.
2323 * rtl.h (emit_insn_before): Likewise.
2324 (emit_jump_insn_before): Likewise.
2325 (emit_debug_insn_before_noloc): Likewise.
2326 (emit_insn_after): Likewise.
2327 (emit_jump_insn_after): Likewise.
2328 (emit_call_insn_after): Likewise.
2329 (emit_debug_insn_after): Likewise.
2330 (set_insn_deleted): Likewise for param.
2332 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2335 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2336 floating point modes, so that IFmode is numerically greater than
2337 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2338 to declare the ordering. This prevents IFmode from being
2339 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2340 machine. Include rs6000-modes.h to share the fractional values
2341 between genmodes* and the rest of the compiler.
2345 * config/rs6000/rs6000-modes.h: New file.
2346 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2347 meaning of rs6000_long_double_size so that 126..128 selects an
2348 appropriate 128-bit floating point type.
2349 (rs6000_option_override_internal): Likewise.
2350 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2351 (TARGET_LONG_DOUBLE_128): Change the meaning of
2352 rs6000_long_double_size so that 126..128 selects an appropriate
2353 128-bit floating point type.
2354 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2355 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2356 source and destination to match the standard usage.
2357 (truncifkf2): Likewise.
2358 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2359 ISA 2.07 to use an explicit clobber, instead of passing in a
2361 (copysign<mode>3_soft): Likewise.
2363 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2365 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2366 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2367 (vect_slp_analyze_instance_dependence): Likewise.
2368 (vect_enhance_data_refs_alignment): Likewise.
2369 (vect_analyze_data_refs_alignment): Likewise.
2370 (vect_slp_analyze_and_verify_instance_alignment
2371 (vect_analyze_data_ref_accesses): Likewise.
2372 (vect_prune_runtime_alias_test_list): Likewise.
2373 (vect_analyze_data_refs): Likewise.
2374 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2375 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2376 (vect_analyze_scalar_cycles_1): Likewise.
2377 (vect_get_loop_niters): Likewise.
2378 (vect_analyze_loop_form_1): Likewise.
2379 (vect_update_vf_for_slp): Likewise.
2380 (vect_analyze_loop_operations): Likewise.
2381 (vect_analyze_loop): Likewise.
2382 (vectorizable_induction): Likewise.
2383 (vect_transform_loop): Likewise.
2384 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2385 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2386 (vect_make_slp_decision): Likewise.
2387 (vect_detect_hybrid_slp): Likewise.
2388 (vect_slp_analyze_operations): Likewise.
2389 (vect_slp_bb): Likewise.
2390 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2391 (vectorizable_bswap): Likewise.
2392 (vectorizable_call): Likewise.
2393 (vectorizable_simd_clone_call): Likewise.
2394 (vectorizable_conversion): Likewise.
2395 (vectorizable_assignment): Likewise.
2396 (vectorizable_shift): Likewise.
2397 (vectorizable_operation): Likewise.
2398 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2400 2018-06-18 Martin Sebor <msebor@redhat.com>
2402 PR tree-optimization/81384
2403 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2404 * builtins.c (expand_builtin_strnlen): New function.
2405 (expand_builtin): Call it.
2406 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2407 * builtins.def (BUILT_IN_STRNLEN): New.
2408 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2409 Warn for bounds in excess of maximum object size.
2410 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2411 single-value ranges. Handle strnlen.
2412 (handle_builtin_strlen): Handle strnlen.
2413 (strlen_check_and_optimize_stmt): Same.
2414 * doc/extend.texi (Other Builtins): Document strnlen.
2416 2018-06-18 Maya Rashish <coypu@sdf.org>
2418 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2419 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2420 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2422 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2424 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2426 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2428 * tree.c (escaped_string::escape): Replace cast to char * by
2429 const_cast<char *> (unescaped).
2431 2018-06-18 Nick Clifton <nickc@redhat.com>
2434 * tree.c (escaped_string): New class. Converts an unescaped
2435 string into its escaped equivalent.
2436 (warn_deprecated_use): Use the new class to convert the
2437 deprecation message, if present.
2438 (test_escaped_strings): New self test.
2439 (test_c_tests): Add test_escaped_strings.
2440 * doc/extend.texi (deprecated): Add a note that the
2441 deprecation message is affected by the -fmessage-length
2442 option, and that control characters will be escaped.
2443 (#pragma GCC error): Document this pragma.
2444 (#pragma GCC warning): Likewise.
2445 * doc/invoke.texi (-fmessage-length): Document this option's
2446 effect on the #warning and #error preprocessor directives and
2447 the deprecated attribute.
2449 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2451 * tree.c (decl_value_expr_lookup): Revert latest change.
2452 (decl_value_expr_insert): Likewise.
2454 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2456 * gimplify.c (nonlocal_vlas): Delete.
2457 (nonlocal_vla_vars): Likewise.
2458 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2460 (gimplify_body): Do not create and destroy nonlocal_vlas.
2461 * tree-nested.c: Include diagnostic.h.
2462 (use_pointer_in_frame): Tweak.
2463 (lookup_field_for_decl): Add assertion and declare the transformation.
2464 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2465 internal error when the reference is in a wrong context. Do not
2466 create a debug decl by default.
2467 (note_nonlocal_block_vlas): Delete.
2468 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2469 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2470 create a debug decl by default.
2471 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2472 call is in a wrong context.
2473 (fixup_vla_decls): New function.
2474 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2475 debug variables were created.
2476 * tree.c (decl_value_expr_lookup): Add checking assertion.
2477 (decl_value_expr_insert): Likewise.
2479 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2482 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2483 * tree-scalar-evolution.c (interpret_expr): Likewise.
2484 (expression_expensive_p): Likewise.
2485 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2486 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2487 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2488 (ssa_defined_by_minus_one_stmt_p): New.
2490 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2493 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2494 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2495 * dojump.c (do_jump): Likewise.
2496 * expr.c (expand_expr_real_2): Check operand type's sign.
2497 * fold-const.c (const_unop): Handle ABSU_EXPR.
2498 (fold_abs_const): Likewise.
2499 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2500 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2501 (strip_sign_op_1): Likesise.
2502 * match.pd: Add new pattern to generate ABSU_EXPR.
2503 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2504 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2505 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2506 * tree-inline.c (estimate_operator_cost): Likewise.
2507 * tree-pretty-print.c (dump_generic_node): Likewise.
2508 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2509 * tree.def (ABSU_EXPR): New.
2511 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2514 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2515 documented as preserved for backward compatibility only.
2516 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2518 PR rtl-optimization/86108
2519 * bb-reorder.c (create_forwarder_block): Renamed to ...
2520 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2521 jump from new landing pad to the second part.
2522 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2525 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2528 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2529 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2530 Only call store_expr for halves if the mode is the same.
2533 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2534 Fix up comment formatting.
2536 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2538 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2539 (typed_splay_tree::closure,
2540 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2541 (typed_splay_tree::typed_splay_tree,
2542 typed_splay_tree::operator =): Declared private.
2543 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2544 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2545 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2546 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2547 typed_splay_tree::splay_tree_splay,
2548 typed_splay_tree::splay_tree_foreach_helper,
2549 typed_splay_tree::splay_tree_insert,
2550 typed_splay_tree::splay_tree_remove,
2551 typed_splay_tree::splay_tree_lookup,
2552 typed_splay_tree::splay_tree_predecessor,
2553 typed_splay_tree::splay_tree_successor,
2554 typed_splay_tree::splay_tree_min,
2555 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2556 (typed_splay_tree::root, typed_splay_tree::comp,
2557 typed_splay_tree::delete_key,
2558 typed_splay_tree::delete_value): New data members.
2559 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2560 typed_splay_tree::remove.
2562 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2564 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2565 -mginv and -mno-ginv to the assembler.
2566 * config/mips/mips.opt (-mcrc): New option.
2568 * doc/invoke.text (-mcrc): Document.
2571 2018-06-15 Nick Clifton <nickc@redhat.com>
2574 * tree.c (escaped_string): New class. Converts an unescaped
2575 string into its escaped equivalent.
2576 (warn_deprecated_use): Use the new class to convert the
2577 deprecation message, if present.
2578 (test_escaped_strings): New self test.
2579 (test_c_tests): Add test_escaped_strings.
2580 * doc/extend.texi (deprecated): Add a note that the
2581 deprecation message is affected by the -fmessage-length
2582 option, and that control characters will be escaped.
2583 (#pragma GCC error): Document this pragma.
2584 (#pragma GCC warning): Likewise.
2585 * doc/invoke.texi (-fmessage-length): Document this option's
2586 effect on the #warning and #error preprocessor directives and
2587 the deprecated attribute.
2589 2018-06-15 Richard Biener <rguenther@suse.de>
2591 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2592 here, also noting vector size used.
2593 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2594 size used in MSG_OPTIMIZED_LOCATIONS dump.
2595 (pass_slp_vectorize::execute): Adjust.
2597 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2600 * config/arc/arc.c (arc_return_address_register): Fix
2603 2018-06-15 Richard Biener <rguenther@suse.de>
2606 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2607 leave useless conversion stripping to force_gimple_operand_gsi.
2608 (gimplify_build2): Likewise.
2609 (gimplify_build1): Likewise.
2611 2018-06-15 Richard Biener <rguenther@suse.de>
2614 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2615 before adjusting their block.
2617 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2619 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2620 multilibs for *-*-rtems*.
2621 * config/riscv/t-rtems: New file.
2623 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2626 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2627 unsigned_type_for returns NULL.
2630 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2631 subregs of multi-word pseudos unless the float mode has word size.
2633 2018-06-14 Richard Biener <rguenther@suse.de>
2636 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2637 duplicate and harmful type_hash_canon.
2638 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2640 2018-06-14 Richard Biener <rguenther@suse.de>
2643 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2646 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2648 * config/rtems.h (STDINT_LONG32): Define.
2650 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2651 Prachi Godbole <prachi.godbole@imgtec.com>
2653 * config/mips/mips-cpus.def: Define P6600.
2654 * config/mips/mips-tables.opt: Regenerate.
2655 * config/mips/mips.c (mips_ucbranch_type): New enum.
2656 (mips_rtx_cost_data): Add support for P6600.
2657 (mips_issue_rate): Likewise.
2658 (mips_multipass_dfa_lookahead): Likewise.
2659 (mips_avoid_hazard): Likewise.
2660 (mips_reorg_process_insns): Likewise.
2661 (mips_classify_branch_p6600): New function.
2662 * config/mips/mips.h (TUNE_P6600): New define.
2663 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2664 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2665 * config/mips/mips.md: Include p6600.md.
2666 (processor): Add p6600.
2667 * config/mips/p6600.md: New file.
2668 * doc/invoke.texi: Add p6600 to supported architectures.
2670 2018-06-13 Martin Sebor <msebor@redhat.com>
2672 PR tree-optimization/86114
2673 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2675 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2677 2018-06-13 Richard Biener <rguenther@suse.de>
2679 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2680 Properly set vector type of the intermediate stmt.
2681 * tree-vect-stmts.c (vectorizable_operation): The destination
2682 var always has vectype_out type.
2684 2018-06-13 Jeff Law <law@redhat.com>
2686 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2687 integer 0 for argument to print_rtl_with_bb.
2688 (rl78_reorg): Likewise.
2690 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2692 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2693 from rtx to rtx_insn *.
2694 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2696 (add_sched_insns_for_speculation): Likewise for local "target",
2697 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2698 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2699 from rtx_insn ** to rtx_code_label **.
2700 (reorg_emit_nops): Likewise.
2701 (c6x_reorg): Likewise for local "call_labels".
2702 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2704 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2705 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2706 the loops over LABEL_REFS.
2707 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2709 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2710 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2711 (split_branches): Strengthen local "olabel" from rtx to
2712 rtx_insn *, adding a safe_as_a cast.
2713 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2715 (add_insn_after): Likewise for first two params.
2716 (add_insn_before): Likewise.
2717 (remove_insn): Likewise for param.
2718 (emit_pattern_before_noloc): Likewise for second and third params.
2719 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2720 (emit_call_insn_before_noloc): Likewise.
2721 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2723 (emit_barrier_before): Likewise.
2724 (emit_label_before): Strengthen "label" param from "rtx" to
2725 "rtx_code_label *". Strengthen "before" param from "rtx" to
2727 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2729 (emit_pattern_after_noloc): Likewise.
2730 (emit_insn_after_noloc): Likewise.
2731 (emit_jump_insn_after_noloc): Likewise.
2732 (emit_call_insn_after_noloc): Likewise.
2733 (emit_debug_insn_after_noloc): Likewise.
2734 (emit_barrier_after): Likewise.
2735 (emit_label_after): Likewise for both params.
2736 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2737 "loc" param from "int" to "location_t".
2738 (emit_insn_after_setloc): Likewise.
2739 (emit_jump_insn_after_setloc): Likewise.
2740 (emit_call_insn_after_setloc): Likewise.
2741 (emit_debug_insn_after_setloc): Likewise.
2742 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2743 "loc" param from "int" to "location_t".
2744 (emit_pattern_before): Convert NULL_RTX to NULL.
2745 (emit_insn_before_setloc): Convert "loc" param from "int" to
2747 (emit_jump_insn_before_setloc): Likewise.
2748 (emit_call_insn_before_setloc): Likewise.
2749 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2750 rtx_insn *. Convert "loc" param from "int" to "location_t".
2751 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2752 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2753 Convert 3rd param from "int" to "location_t".
2754 (emit_barrier_before, emit_barrier_after, next_real_insn):
2755 Strengthen param from rtx to rtx_insn *.
2756 (emit_label_before): Strengthen 1st param from "rtx" to
2757 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2759 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2760 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2761 Strengthen 2nd param from "rtx" to "rtx_insn *".
2762 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2763 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2764 Likewise. Convert 3rd param from "int" to "location_t".
2765 (emit_label_after): Strengthen 1st param from "rtx" to
2767 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2769 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2770 from "rtx" to "rtx_insn *".
2772 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2774 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2775 bodies streamed in with -Q.
2776 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2777 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2778 * gimple-streamer-out.c: Include gimple-pretty-print.h
2779 (output_bb): Dump stmts streamed.
2780 * lto-section-out.c: Include print-tree.h
2781 (lto_begin_section): Dump sections created.
2782 (lto_output_decl_index): Dump decl encoded.
2783 * lto-streamer-out.c: Include print-tree.h
2784 (create_output_block): Dump output block created.
2785 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2786 (output_function): Dump function output.
2787 (output_constructor): Dump constructor streamed.
2788 (write_global_stream): Output indexes encoded.
2789 (produce_asm_for_decls): Dump streams encoded.
2790 * lto-streamer.c (streamer_dump_file): New global var.
2791 * lto-streamer.h (streamer_dump_file): Declare.
2792 * passes.c (ipa_write_summaries): Initialize streamer dump.
2793 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2796 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2799 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2800 offsets for register save directives. Emit a second batch of save
2801 directives, if need be, when the function accesses prior frames.
2803 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2805 * config/arc/fpu.md (fmasf4): Force operand to register.
2806 (fnmasf4): Likewise.
2808 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2810 * config/arc/arc-protos.h (arc_pad_return): Remove.
2811 * config/arc/arc.c (machine_function): Remove force_short_suffix
2813 (arc_print_operand): Adjust printing of '&'.
2814 (arc_verify_short): Remove conditional printing of short suffix.
2815 (arc_final_prescan_insn): Remove reference to size_reason.
2816 (pad_return): New function.
2817 (arc_reorg): Call pad_return.
2818 (arc_pad_return): Remove.
2819 (arc_init_machine_status): Remove reference to force_short_suffix.
2820 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2821 (attr length): When attribute iscompact is true force to 2
2822 regardless; in the case of maybe check if we want to force the
2823 instruction to have 4 bytes length.
2824 (nopv): Change it to generate 4 byte long nop as well.
2825 (blockage): New pattern.
2826 (simple_return): Remove call to arc_pad_return.
2827 (p_return_i): Likewise.
2829 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2831 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2833 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2835 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2838 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2840 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2841 for ARC700 and ARCv2.
2843 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2846 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2847 operands[2] instead of operands[1].
2850 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2852 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2853 case, check whether the outer register overlaps an unallocatable
2854 register, not just whether it fits the required class.
2856 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2858 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2860 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2861 (memrefs_conflict_p): Likewise.
2862 (init_alias_analysis): Likewise.
2863 * cfgexpand.c (expand_debug_expr): Likewise.
2864 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2865 * cse.c (fold_rtx): Likewise.
2866 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2867 * expr.c (emit_block_move_hints): Likewise.
2868 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2869 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2870 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2871 (emit_group_store): Likewise.
2872 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2873 to read the PRE/POST_MODIFY increment.
2874 * calls.c (store_one_arg): Use strip_offset.
2875 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2877 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2879 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2880 (simplify_binary_operation_1): Extend CONST_INT handling to
2882 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2883 than a HOST_WIDE_INT.
2884 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2886 (adjust_mems, add_stores): Update accodingly.
2887 (vt_canonicalize_addr): Track polynomial offsets.
2888 (emit_note_insn_var_location): Likewise.
2889 (vt_add_function_parameter): Likewise.
2890 (vt_initialize): Likewise.
2892 2018-06-12 Jeff Law <law@redhat.com>
2894 * config.gcc (alpha*-*-freebsd*): Remove.
2895 * config/alpha/freebsd.h: Remove.
2897 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2900 * spellcheck-tree.c (levenshtein_distance): Rename to...
2901 (get_edit_distance): ...this, and update for underlying renaming.
2902 * spellcheck-tree.h (levenshtein_distance): Rename to...
2903 (get_edit_distance): ...this.
2904 * spellcheck.c (levenshtein_distance): Rename to...
2905 (get_edit_distance): ...this. Convert from Levenshtein distance
2906 to Damerau-Levenshtein distance by supporting transpositions of
2907 adjacent characters. Rename "v1" to "v_next" and "v0" to
2909 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2910 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2911 update for underlying renaming.
2912 (selftest::levenshtein_distance_unit_test): Rename to...
2913 (selftest::test_get_edit_distance_unit): ...this, and update for
2914 underlying renaming.
2915 (selftest::test_find_closest_string): Add example from PR 69968
2916 where transposition helps
2917 (selftest::test_metric_conditions): Update for renaming.
2918 (selftest::test_metric_conditions): Likewise.
2919 (selftest::spellcheck_c_tests): Likewise.
2920 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2921 (get_edit_distance): ...this.
2922 (best_match::consider): Update for renaming.
2924 2018-06-12 Martin Sebor <msebor@redhat.com>
2926 PR tree-optimization/85259
2927 * builtins.c (compute_objsize): Handle constant offsets.
2928 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2929 true iff a warning has been issued.
2930 * gimple.h (gimple_nonartificial_location): New function.
2931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2932 gimple_nonartificial_location and handle -Wno-system-headers.
2933 (handle_builtin_stxncpy): Same.
2935 2018-06-12 Martin Sebor <msebor@redhat.com>
2938 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2940 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2942 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2943 BUILTIN_VEC_XST entries for pointer to double and long long.
2945 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2948 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2950 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2953 2018-06-12 Martin Liska <mliska@suse.cz>
2955 * doc/options.texi: Document IntegerRange.
2957 2018-06-12 Martin Liska <mliska@suse.cz>
2959 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2960 * opt-functions.awk: Handle Deprecated flag.
2961 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2963 (read_cmdline_option): Report warning for a deprecated option.
2964 * opts.h (struct cl_option): Add new field cl_deprecated.
2965 (CL_ERR_DEPRECATED): New.
2967 2018-06-12 Martin Liska <mliska@suse.cz>
2969 * doc/options.texi: Document Deprecated option flag.
2971 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2973 * config/arc/arc-arch.h (arc_extras): New enum.
2974 (arc_cpu_t):Add field extra.
2975 (arc_cpu_types): Consider the extras.
2976 * config/arc/arc-cpus.def: Add extras info.
2977 * config/arc/arc-opts.h (processor_type): Consider extra field.
2978 * config/arc/arc.c (arc_override_options): Handle extra field.
2980 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2982 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2983 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2985 * config/arc/arc.c (arc_init): Update pic warning.
2986 (irq_range): Update irq range parsing warnings.
2987 (arc_override_options): Update various warning messages.
2988 (arc_handle_aux_attribute): Likewise.
2990 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2992 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2994 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2996 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2997 compiler messages directives.
2999 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3001 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3002 * config/mips/mips-tables.opt: Regenerate.
3003 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3005 * doc/invoke.texi: Document -march=i6500.
3007 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3009 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3010 (i6400_gpmul): Add cpu_unit.
3011 (i6400_gpdiv): Likewise.
3012 (i6400_msa_add_d): Update reservations.
3013 (i6400_msa_int_add) Likewise.
3014 (i6400_msa_short_logic3) Likewise.
3015 (i6400_msa_short_logic2) Likewise.
3016 (i6400_msa_short_logic) Likewise.
3017 (i6400_msa_move) Likewise.
3018 (i6400_msa_cmp) Likewise.
3019 (i6400_msa_short_float2) Likewise.
3020 (i6400_msa_div_d) Likewise.
3021 (i6400_msa_long_logic1) Likewise.
3022 (i6400_msa_long_logic2) Likewise.
3023 (i6400_msa_mult) Likewise.
3024 (i6400_msa_long_float2) Likewise.
3025 (i6400_msa_long_float4) Likewise.
3026 (i6400_msa_long_float5) Likewise.
3027 (i6400_msa_long_float8) Likewise.
3028 (i6400_fpu_fadd): Include frint type.
3029 (i6400_fpu_store): New define_insn_reservation.
3030 (i6400_fpu_load): Likewise.
3031 (i6400_fpu_move): Likewise.
3032 (i6400_fpu_fcmp): Likewise.
3033 (i6400_fpu_fmadd): Likewise.
3034 (i6400_int_mult): Include imul3nc type and update reservation.
3035 (i6400_int_div): Include idiv3 type and update reservation.
3036 (i6400_int_load): Update to check type not move_type.
3037 (i6400_int_store): Likewise.
3038 (i6400_int_prefetch): Set zero latency.
3040 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3042 * gcc.c: Document new %@{...} sequence.
3043 (LINK_COMMAND_SPEC): Use it for the -L switches.
3044 (cpp_unique_options): Use it for the -I switches.
3045 (at_file_argbuf): New global variable.
3046 (in_at_file): Likewise.
3047 (alloc_args): Create at_file_argbuf.
3048 (clear_args): Truncate at_file_argbuf.
3049 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3050 (open_at_file): New function.
3051 (close_at_file): Likewise.
3052 (create_at_file): Delete.
3053 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3056 (validate_switches_from_spec): Deal with %@{...} sequence.
3057 (validate_switches): Likewise.
3058 (driver::finalize): Call clear_args.
3060 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3062 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3064 2018-06-11 Martin Sebor <msebor@redhat.com>
3066 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3067 (Wno-class-memaccess): Add @opindex.
3068 (Wno-templates, Wno-multiple-inheritance): Same.
3069 (Wno-virtual-inheritance, Wno-namespaces): Same.
3070 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3071 (Wno-format-overflow, Wno-format-truncation): Same.
3072 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3073 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3074 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3075 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3076 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3077 (Wno-misspelled-isr): Same.
3079 2018-06-11 Martin Sebor <msebor@redhat.com>
3081 * PR tree-optimization/86083
3082 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3084 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3086 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3088 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3091 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3092 on the correct operand.
3093 (*movdi_internal64): Ditto.
3095 2018-06-11 Martin Liska <mliska@suse.cz>
3097 PR tree-optimization/86089
3098 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3100 2018-06-11 Julia Koval <julia.koval@intel.com>
3102 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3103 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3104 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3106 2018-06-11 Olivier Hainque <hainque@adacore.com>
3108 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3109 for Ada with strict dwarf2.
3111 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3114 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3117 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3119 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3121 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3123 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3126 2018-06-08 Martin Liska <mliska@suse.cz>
3128 * tree-cfg.h (debug_function): Fix argument type to match
3131 2018-06-08 Martin Liska <mliska@suse.cz>
3133 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3134 Remove usage of MPX-related (and removed) fields.
3135 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3137 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3139 * cfg.c (debug): Use TDF_NONE rather than 0.
3140 * cfghooks.c (debug): Likewise.
3141 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3142 (struct dump_option_value_info): Convert to...
3143 (struct kv_pair): ...this template type.
3144 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3146 (optinfo_verbosity_options): Likewise.
3147 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3149 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3150 than int for "optgroup_flags" param.
3151 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3153 (dump_dec): Likewise.
3154 (dump_finish): Use TDF_NONE rather than 0.
3155 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3156 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3157 than 0. Update for change to option_ptr.
3158 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3159 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3160 0. Update for changes to optinfo_verbosity_options and
3162 (opt_info_switch_p): Convert optgroup_flags from int to
3164 (dump_basic_block): Use dump_flags_t rather than int
3165 for "dump_kind" param.
3166 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3167 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3168 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3169 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3170 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3171 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3172 TDF_NONE): Convert from macros to...
3173 (enum dump_flag): ...this new enum.
3174 (dump_flags_t): Update to use enum.
3175 (operator|, operator&, operator~, operator|=, operator&=):
3176 Implement for dump_flags_t.
3177 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3178 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3179 Convert from macros to...
3180 (enum optgroup_flag): ...this new enum.
3181 (optgroup_flags_t): New typedef.
3182 (operator|, operator|=): Implement for optgroup_flags_t.
3183 (struct dump_file_info): Convert field "alt_flags" to
3184 dump_flags_t. Convert field "optgroup_flags" to
3186 (dump_basic_block): Use dump_flags_t rather than int for param.
3187 (dump_generic_expr_loc): Likewise.
3188 (dump_dec): Likewise.
3189 (dump_register): Convert param "optgroup_flags" to
3191 (opt_info_enable_passes): Likewise.
3192 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3194 * gimple-pretty-print.c (debug): Likewise.
3195 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3196 (merged_store_group::apply_stores): Likewise.
3197 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3198 * gimple.c (verify_gimple_pp): Likewise.
3199 * graphite-poly.c (print_pbb_body): Likewise.
3200 * passes.c (pass_manager::register_one_dump_file): Convert
3201 local "optgroup_flags" to optgroup_flags_t.
3202 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3204 (debug_body): Likewise.
3205 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3206 to optgroup_flags_t.
3207 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3209 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3210 (convert_mult_to_fma): Likewise.
3211 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3212 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3213 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3214 "dump_kind" to dump_flags_t.
3216 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3218 * config/rs6000/rs6000.c (min, max): Delete.
3220 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3222 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3223 -mabi=spe and -mabi=no-spe.
3225 2018-06-08 Martin Liska <mliska@suse.cz>
3227 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3228 where we expect an existing summary.
3230 2018-06-08 Martin Liska <mliska@suse.cz>
3232 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3233 * ipa-inline.h (estimate_edge_growth): Likewise.
3235 2018-06-08 Martin Liska <mliska@suse.cz>
3237 * cgraph.c (function_version_hasher::hash): Use
3238 cgraph_node::get_uid ().
3239 (function_version_hasher::equal):
3240 * cgraph.h (cgraph_node::get_uid): New method.
3241 * ipa-inline.c (update_caller_keys): Use
3242 cgraph_node::get_uid ().
3243 (update_callee_keys): Likewise.
3244 * ipa-utils.c (searchc): Likewise.
3245 (ipa_reduced_postorder): Likewise.
3246 * lto-cgraph.c (input_node): Likewise.
3247 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3248 * symbol-summary.h (symtab_insertion): Likewise.
3249 (symtab_removal): Likewise.
3250 (symtab_duplication): Likewise.
3251 * tree-pretty-print.c (dump_function_header): Likewise.
3252 * tree-sra.c (convert_callers_for_node): Likewise.
3254 2018-06-08 Martin Liska <mliska@suse.cz>
3256 * cgraph.c (symbol_table::create_edge): Always assign a new
3258 (symbol_table::free_edge): Do not recycle numbers.
3259 * cgraph.h (cgraph_edge::get): New method.
3260 * symbol-summary.h (symtab_removal): Use it.
3261 (symtab_duplication): Likewise.
3262 (call_summary::hashable_uid): Remove.
3264 2018-06-08 Martin Liska <mliska@suse.cz>
3266 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3267 (initialize_growth_caches): Remove.
3268 (free_growth_caches): Likewise.
3269 (do_estimate_edge_time): Use edge_growth_cache.
3270 (do_estimate_edge_size): Likewise.
3271 (do_estimate_edge_hints): Likewise.
3272 * ipa-inline.c (reset_edge_caches): Likewise.
3273 (recursive_inlining): Likewise.
3274 (inline_small_functions): Likewise.
3275 * ipa-inline.h (initialize_growth_caches): Remove.
3276 (estimate_edge_size): Likewise.
3277 (estimate_edge_time): Likewise.
3278 (estimate_edge_hints): Likewise.
3279 (reset_edge_growth_cache): Likewise.
3280 * symbol-summary.h (call_summary::remove): New method.
3282 2018-06-08 Martin Liska <mliska@suse.cz>
3284 * ipa-cp.c (class edge_clone_summary): New summary.
3285 (grow_edge_clone_vectors): Remove.
3286 (ipcp_edge_duplication_hook): Remove.
3287 (class edge_clone_summary_t): New call_summary class.
3288 (ipcp_edge_removal_hook): Remove.
3289 (edge_clone_summary_t::duplicate): New function.
3290 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3291 (create_specialized_node): Likewise.
3292 (ipcp_driver): Initialize edge_clone_summaries and do not
3295 2018-06-08 Martin Liska <mliska@suse.cz>
3297 * symbol-summary.h (get): New function.
3298 (call_summary::m_initialize_when_cloning): New class member.
3300 2018-06-08 Martin Liska <mliska@suse.cz>
3302 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3303 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3304 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3305 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3306 (remove_cgraph_node_from_order): Use the removed_nodes set.
3307 (do_per_function_toporder): Likwise.
3308 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3309 instead of summary_uid.
3310 (symtab_removal): Likewise.
3311 (symtab_duplication): Likewise.
3313 2018-06-08 Martin Liska <mliska@suse.cz>
3315 * ipa-cp.c (ipcp_store_bits_results): Use
3316 ipcp_transformation_sum.
3317 (ipcp_store_vr_results): Likewise.
3318 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3320 (ipcp_transformation_initialize): ... this.
3321 (ipa_set_node_agg_value_chain):
3322 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3323 (write_ipcp_transformation_info): Likewise.
3324 (read_ipcp_transformation_info): Likewise.
3325 (ipcp_update_bits): Likewise.
3326 (ipcp_update_vr): Likewise.
3327 (ipcp_transform_function): Likewise.
3328 * ipa-prop.h: Rename ipcp_transformation_summary to
3329 ipcp_transformation.
3330 (class ipcp_transformation_t): New function summary.
3331 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3332 (ipa_get_agg_replacements_for_node): Likewise.
3334 2018-06-08 Martin Liska <mliska@suse.cz>
3336 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3338 (class funct_state_summary_t): New function_summary class.
3339 (has_function_state): Remove.
3340 (get_function_state): Likewise.
3341 (set_function_state): Likewise.
3342 (add_new_function): Likewise.
3343 (funct_state_summary_t::insert): New function.
3344 (duplicate_node_data): Remove.
3345 (remove_node_data): Remove.
3346 (funct_state_summary_t::duplicate): New function.
3347 (register_hooks): Create new funct_state_summaries.
3348 (pure_const_generate_summary): Use it.
3349 (pure_const_write_summary): Likewise.
3350 (pure_const_read_summary): Likewise.
3351 (propagate_pure_const): Likewise.
3352 (propagate_nothrow): Likewise.
3353 (dump_malloc_lattice): Likewise.
3354 (propagate_malloc): Likewise.
3355 (execute): Do not register hooks, just remove summary
3357 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3360 2018-06-08 Martin Liska <mliska@suse.cz>
3362 * ipa-reference.c (remove_node_data): Remove.
3363 (duplicate_node_data): Likewise.
3364 (class ipa_ref_var_info_summary_t): New class.
3365 (class ipa_ref_opt_summary_t): Likewise.
3366 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3367 (get_reference_optimization_summary): Use
3368 ipa_ref_opt_sum_summaries.
3369 (set_reference_vars_info): Remove.
3370 (set_reference_optimization_summary): Likewise.
3371 (ipa_init): Create summaries.
3372 (init_function_info): Use function summary.
3373 (ipa_ref_opt_summary_t::duplicate): New function.
3374 (ipa_ref_opt_summary_t::remove): New function.
3375 (get_read_write_all_from_node): Fix GNU coding style.
3376 (propagate): Use function summary.
3377 (write_node_summary_p): Fix GNU coding style.
3378 (stream_out_bitmap): Likewise.
3379 (ipa_reference_read_optimization_summary): Use function summary.
3380 (ipa_reference_c_finalize): Do not release hooks.
3382 2018-06-08 Martin Liska <mliska@suse.cz>
3384 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3385 (analyze_function_body): Extract multiple calls of get_create.
3386 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3387 * ipa-inline.c (recursive_inlining): Use ::get method.
3388 * ipa-inline.h (estimate_edge_growth): Likewise.
3390 2018-06-08 Martin Liska <mliska@suse.cz>
3392 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3394 (hsa_function_summary::hsa_function_summary): Use the new enum
3396 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3397 * hsa-gen.c (hsa_get_host_function): Likewise.
3398 (get_brig_function_name): Likewise.
3399 * ipa-hsa.c (process_hsa_functions): Likewise.
3400 (ipa_hsa_write_summary): Likewise.
3401 * symbol-summary.h (symtab_duplication): Use ::get function/
3402 (get): New function.
3404 2018-06-08 Martin Liska <mliska@suse.cz>
3406 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3408 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3409 (hsa_register_kernel): Likewise.
3410 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3411 * hsa-gen.c (hsa_get_host_function): Likewise.
3412 (get_brig_function_name): Likewise.
3413 (generate_hsa): Likewise.
3414 (pass_gen_hsail::execute): Likewise.
3415 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3416 (devirtualization_time_bonus): Likewise.
3417 (ipcp_propagate_stage): Likewise.
3418 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3419 (edge_set_predicate): Likewise.
3420 (evaluate_conditions_for_known_args): Likewise.
3421 (evaluate_properties_for_edge): Likewise.
3422 (ipa_fn_summary::reset): Likewise.
3423 (ipa_fn_summary_t::duplicate): Likewise.
3424 (dump_ipa_call_summary): Likewise.
3425 (ipa_dump_fn_summary): Likewise.
3426 (analyze_function_body): Likewise.
3427 (compute_fn_summary): Likewise.
3428 (estimate_edge_devirt_benefit): Likewise.
3429 (estimate_edge_size_and_time): Likewise.
3430 (estimate_calls_size_and_time): Likewise.
3431 (estimate_node_size_and_time): Likewise.
3432 (inline_update_callee_summaries): Likewise.
3433 (remap_edge_change_prob): Likewise.
3434 (remap_edge_summaries): Likewise.
3435 (ipa_merge_fn_summary_after_inlining): Likewise.
3436 (ipa_update_overall_fn_summary): Likewise.
3437 (read_ipa_call_summary): Likewise.
3438 (inline_read_section): Likewise.
3439 (write_ipa_call_summary): Likewise.
3440 (ipa_fn_summary_write): Likewise.
3441 (ipa_free_fn_summary): Likewise.
3442 * ipa-hsa.c (process_hsa_functions): Likewise.
3443 (ipa_hsa_write_summary): Likewise.
3444 (ipa_hsa_read_section): Likewise.
3445 * ipa-icf.c (sem_function::merge): Likewise.
3446 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3447 (do_estimate_edge_time): Likewise.
3448 (estimate_size_after_inlining): Likewise.
3449 (estimate_growth): Likewise.
3450 (growth_likely_positive): Likewise.
3451 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3452 (inline_call): Likewise.
3453 * ipa-inline.c (caller_growth_limits): Likewise.
3454 (can_inline_edge_p): Likewise.
3455 (can_inline_edge_by_limits_p): Likewise.
3456 (compute_uninlined_call_time): Likewise.
3457 (compute_inlined_call_time): Likewise.
3458 (want_inline_small_function_p): Likewise.
3459 (edge_badness): Likewise.
3460 (update_caller_keys): Likewise.
3461 (update_callee_keys): Likewise.
3462 (recursive_inlining): Likewise.
3463 (inline_small_functions): Likewise.
3464 (inline_to_all_callers_1): Likewise.
3465 (dump_overall_stats): Likewise.
3466 (early_inline_small_functions): Likewise.
3467 (early_inliner): Likewise.
3468 * ipa-inline.h (estimate_edge_growth): Likewise.
3469 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3470 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3471 * ipa-prop.h (IPA_NODE_REF): Likewise.
3472 (IPA_EDGE_REF): Likewise.
3473 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3474 (propagate_malloc): Likewise.
3475 * ipa-split.c (execute_split_functions): Likewise.
3476 * symbol-summary.h: Rename get to get_create.
3478 (get_create): Likewise.
3479 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3481 2018-06-08 Martin Liska <mliska@suse.cz>
3483 * symbol-summary.h (release): Move definition out of class
3485 (symtab_removal): Likewise.
3486 (symtab_duplication): Likewise.
3488 2018-06-08 Martin Liska <mliska@suse.cz>
3490 * symbol-summary.h (function_summary): Move constructor
3491 implementation out of class declaration.
3492 (release): Likewise.
3493 (symtab_insertion): Likewise.
3494 (symtab_removal): Likewise.
3495 (symtab_duplication): Likewise.
3498 2018-06-08 Martin Liska <mliska@suse.cz>
3500 * Makefile.in: Remove support for MPX (macros, related functions,
3501 fields in cgraph_node, ...).
3502 * builtin-types.def (BT_BND): Likewise.
3503 (BT_FN_BND_CONST_PTR): Likewise.
3504 (BT_FN_CONST_PTR_BND): Likewise.
3505 (BT_FN_VOID_PTR_BND): Likewise.
3506 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3507 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3508 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3509 (expand_builtin_mempcpy_with_bounds): Likewise.
3510 (expand_builtin_memset_with_bounds): Likewise.
3511 (expand_builtin_memset_args): Likewise.
3512 (std_expand_builtin_va_start): Likewise.
3513 (expand_builtin): Likewise.
3514 (expand_builtin_with_bounds): Likewise.
3515 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3516 (DEF_LIB_BUILTIN_CHKP): Likewise.
3517 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3518 (DEF_CHKP_BUILTIN): Likewise.
3519 (BUILT_IN_MEMCPY): Likewise.
3520 (BUILT_IN_MEMMOVE): Likewise.
3521 (BUILT_IN_MEMPCPY): Likewise.
3522 (BUILT_IN_MEMSET): Likewise.
3523 (BUILT_IN_STPCPY): Likewise.
3524 (BUILT_IN_STRCAT): Likewise.
3525 (BUILT_IN_STRCHR): Likewise.
3526 (BUILT_IN_STRCPY): Likewise.
3527 (BUILT_IN_STRLEN): Likewise.
3528 (BUILT_IN_MEMCPY_CHK): Likewise.
3529 (BUILT_IN_MEMMOVE_CHK): Likewise.
3530 (BUILT_IN_MEMPCPY_CHK): Likewise.
3531 (BUILT_IN_MEMSET_CHK): Likewise.
3532 (BUILT_IN_STPCPY_CHK): Likewise.
3533 (BUILT_IN_STRCAT_CHK): Likewise.
3534 (BUILT_IN_STRCPY_CHK): Likewise.
3535 * calls.c (store_bounds): Likewise.
3536 (emit_call_1): Likewise.
3537 (special_function_p): Likewise.
3538 (maybe_warn_nonstring_arg): Likewise.
3539 (initialize_argument_information): Likewise.
3540 (finalize_must_preallocate): Likewise.
3541 (compute_argument_addresses): Likewise.
3542 (expand_call): Likewise.
3543 * cfgexpand.c (expand_call_stmt): Likewise.
3544 (expand_return): Likewise.
3545 (expand_gimple_stmt_1): Likewise.
3546 (pass_expand::execute): Likewise.
3547 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3548 (cgraph_node::remove): Likewise.
3549 (cgraph_node::dump): Likewise.
3550 (cgraph_node::verify_node): Likewise.
3551 * cgraph.h (chkp_function_instrumented_p): Likewise.
3552 (symtab_node::get_alias_target): Likewise.
3553 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3554 (cgraph_local_p): Likewise.
3555 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3556 (cgraph_edge::rebuild_references): Likewise.
3557 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3558 (walk_polymorphic_call_targets): Likewise.
3559 (cgraph_node::expand_thunk): Likewise.
3560 (symbol_table::output_weakrefs): Likewise.
3561 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3562 (ix86_handle_option): Likewise.
3563 * config/i386/constraints.md: Likewise.
3564 * config/i386/i386-builtin-types.def (BND): Likewise.
3568 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3569 (BDESC_FIRST): Likewise.
3571 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3572 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3573 * config/i386/i386.c (enum reg_class): Likewise.
3574 (ix86_target_string): Likewise.
3575 (ix86_option_override_internal): Likewise.
3576 (ix86_conditional_register_usage): Likewise.
3577 (ix86_valid_target_attribute_inner_p): Likewise.
3578 (ix86_set_indirect_branch_type): Likewise.
3579 (ix86_set_current_function): Likewise.
3580 (ix86_function_arg_regno_p): Likewise.
3581 (init_cumulative_args): Likewise.
3582 (ix86_function_arg_advance): Likewise.
3583 (ix86_function_arg): Likewise.
3584 (ix86_pass_by_reference): Likewise.
3585 (ix86_function_value_regno_p): Likewise.
3586 (ix86_function_value_1): Likewise.
3587 (ix86_function_value_bounds): Likewise.
3588 (ix86_return_in_memory): Likewise.
3589 (ix86_setup_incoming_vararg_bounds): Likewise.
3590 (ix86_va_start): Likewise.
3591 (indirect_thunk_need_prefix): Likewise.
3592 (print_reg): Likewise.
3593 (ix86_print_operand): Likewise.
3594 (ix86_expand_call): Likewise.
3595 (ix86_output_function_return): Likewise.
3596 (reg_encoded_number): Likewise.
3597 (BDESC_VERIFYS): Likewise.
3598 (ix86_init_mpx_builtins): Likewise.
3599 (ix86_init_builtins): Likewise.
3600 (ix86_emit_cmove): Likewise.
3601 (ix86_emit_move_max): Likewise.
3602 (ix86_expand_builtin): Likewise.
3603 (ix86_builtin_mpx_function): Likewise.
3604 (ix86_get_arg_address_for_bt): Likewise.
3605 (ix86_load_bounds): Likewise.
3606 (ix86_store_bounds): Likewise.
3607 (ix86_load_returned_bounds): Likewise.
3608 (ix86_store_returned_bounds): Likewise.
3609 (ix86_class_likely_spilled_p): Likewise.
3610 (ix86_hard_regno_mode_ok): Likewise.
3611 (x86_order_regs_for_local_alloc): Likewise.
3612 (ix86_mitigate_rop): Likewise.
3613 (ix86_bnd_prefixed_insn_p): Likewise.
3614 (ix86_mpx_bound_mode): Likewise.
3615 (ix86_make_bounds_constant): Likewise.
3616 (ix86_initialize_bounds): Likewise.
3617 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3618 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3619 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3620 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3621 (TARGET_CHKP_BOUND_MODE): Likewise.
3622 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3623 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3624 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3625 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3626 * config/i386/i386.h (TARGET_MPX): Likewise.
3627 (TARGET_MPX_P): Likewise.
3628 (VALID_BND_REG_MODE): Likewise.
3629 (FIRST_BND_REG): Likewise.
3630 (LAST_BND_REG): Likewise.
3631 (enum reg_class): Likewise.
3632 (BND_REG_P): Likewise.
3633 (BND_REGNO_P): Likewise.
3634 (BNDmode): Likewise.
3635 (ADJUST_INSN_LENGTH): Likewise.
3636 * config/i386/i386.md: Likewise.
3637 * config/i386/i386.opt: Likewise.
3638 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3639 (defined): Likewise.
3640 (LINK_MPX): Likewise.
3641 (MPX_SPEC): Likewise.
3642 (LIBMPX_SPEC): Likewise.
3643 (LIBMPXWRAPPERS_SPEC): Likewise.
3644 (CHKP_SPEC): Likewise.
3645 * config/i386/predicates.md: Likewise.
3646 * dbxout.c (dbxout_type): Likewise.
3647 * doc/extend.texi: Likewise.
3648 * doc/invoke.texi: Likewise.
3649 * doc/md.texi: Likewise.
3650 * doc/tm.texi: Likewise.
3651 * doc/tm.texi.in: Likewise.
3652 * dwarf2out.c (is_base_type): Likewise.
3653 (gen_formal_types_die): Likewise.
3654 (gen_subprogram_die): Likewise.
3655 (gen_type_die_with_usage): Likewise.
3656 (gen_decl_die): Likewise.
3657 (dwarf2out_late_global_decl): Likewise.
3658 * expr.c (expand_assignment): Likewise.
3659 (emit_storent_insn): Likewise.
3660 (store_expr_with_bounds): Likewise.
3661 (store_expr): Likewise.
3662 (expand_expr_real_1): Likewise.
3663 * expr.h (store_expr_with_bounds): Likewise.
3664 * function.c (use_register_for_decl): Likewise.
3665 (struct bounds_parm_data): Likewise.
3666 (assign_parms_augmented_arg_list): Likewise.
3667 (assign_parm_find_entry_rtl): Likewise.
3668 (assign_parm_is_stack_parm): Likewise.
3669 (assign_parm_load_bounds): Likewise.
3670 (assign_bounds): Likewise.
3671 (assign_parms): Likewise.
3672 (expand_function_start): Likewise.
3673 * gcc.c (CHKP_SPEC): Likewise.
3674 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3675 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3676 (wrestrict_dom_walker::check_call): Likewise.
3677 * gimple.c (gimple_build_call_from_tree): Likewise.
3678 * gimple.h (enum gf_mask): Likewise.
3679 (gimple_call_with_bounds_p): Likewise.
3680 (gimple_call_set_with_bounds): Likewise.
3681 * gimplify.c (gimplify_init_constructor): Likewise.
3682 * ipa-cp.c (initialize_node_lattices): Likewise.
3683 (propagate_constants_across_call): Likewise.
3684 (find_more_scalar_values_for_callers_subset): Likewise.
3685 * ipa-hsa.c (process_hsa_functions): Likewise.
3686 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3687 * ipa-icf.c (sem_function::merge): Likewise.
3688 * ipa-inline.c (early_inliner): Likewise.
3689 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3690 (warn_function_cold): Likewise.
3691 (propagate_pure_const): Likewise.
3692 * ipa-ref.h (enum GTY): Likewise.
3693 * ipa-split.c (find_retbnd): Likewise.
3694 (consider_split): Likewise.
3695 (split_function): Likewise.
3696 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3697 * ipa.c (walk_polymorphic_call_targets): Likewise.
3698 (symbol_table::remove_unreachable_nodes): Likewise.
3699 (process_references): Likewise.
3700 (cgraph_build_static_cdtor_1): Likewise.
3701 * lto-cgraph.c (lto_output_node): Likewise.
3702 (output_refs): Likewise.
3703 (compute_ltrans_boundary): Likewise.
3704 (input_overwrite_node): Likewise.
3705 (input_node): Likewise.
3706 (input_cgraph_1): Likewise.
3707 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3708 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3709 (class pass_chkp_instrumentation_passes): Likewise.
3710 (make_pass_chkp_instrumentation_passes): Likewise.
3711 * passes.def: Likewise.
3712 * rtl.h (struct GTY): Likewise.
3713 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3714 * stor-layout.c (layout_type): Likewise.
3715 * symtab.c: Likewise.
3716 * target.def: Likewise.
3717 * targhooks.c (default_chkp_bound_type): Likewise.
3718 (default_chkp_bound_mode): Likewise.
3719 (default_builtin_chkp_function): Likewise.
3720 (default_chkp_function_value_bounds): Likewise.
3721 (default_chkp_make_bounds_constant): Likewise.
3722 (default_chkp_initialize_bounds): Likewise.
3723 * targhooks.h (default_chkp_bound_type): Likewise.
3724 (default_chkp_bound_mode): Likewise.
3725 (default_builtin_chkp_function): Likewise.
3726 (default_chkp_function_value_bounds): Likewise.
3727 (default_chkp_make_bounds_constant): Likewise.
3728 (default_chkp_initialize_bounds): Likewise.
3729 * toplev.c (compile_file): Likewise.
3730 (process_options): Likewise.
3731 * tree-core.h (DEF_BUILTIN): Likewise.
3732 (DEF_BUILTIN_CHKP): Likewise.
3733 * tree-inline.c (declare_return_variable): Likewise.
3734 (remap_gimple_stmt): Likewise.
3735 (copy_bb): Likewise.
3736 (initialize_inlined_parameters): Likewise.
3737 (expand_call_inline): Likewise.
3738 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3739 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3740 (make_pass_ipa_chkp_produce_thunks): Likewise.
3741 (make_pass_chkp): Likewise.
3742 (make_pass_chkp_opt): Likewise.
3743 (make_pass_chkp_instrumentation_passes): Likewise.
3744 * tree-pretty-print.c (dump_generic_node): Likewise.
3745 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3746 * tree-ssa-dce.c (propagate_necessity): Likewise.
3747 (eliminate_unnecessary_stmts): Likewise.
3748 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3749 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3750 * tree-ssa-sccvn.h: Likewise.
3751 * tree-ssa-strlen.c (get_string_length): Likewise.
3752 (valid_builtin_call): Likewise.
3753 (adjust_last_stmt): Likewise.
3754 (handle_builtin_strchr): Likewise.
3755 (handle_builtin_strcpy): Likewise.
3756 (handle_builtin_stxncpy): Likewise.
3757 (handle_builtin_memcpy): Likewise.
3758 (handle_builtin_strcat): Likewise.
3759 (strlen_check_and_optimize_stmt): Likewise.
3760 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3761 * tree-streamer-in.c: Likewise.
3762 * tree-streamer.c (record_common_node): Likewise.
3763 * tree.c (tree_code_size): Likewise.
3764 (wide_int_to_tree_1): Likewise.
3765 (type_contains_placeholder_1): Likewise.
3766 (build_common_tree_nodes): Likewise.
3767 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3768 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3769 (POINTER_BOUNDS_P): Likewise.
3770 (BOUNDED_TYPE_P): Likewise.
3771 (BOUNDED_P): Likewise.
3772 (CALL_WITH_BOUNDS_P): Likewise.
3773 (pointer_bounds_type_node): Likewise.
3774 * value-prof.c (gimple_ic): Likewise.
3775 * var-tracking.c (vt_add_function_parameters): Likewise.
3776 * varasm.c (make_decl_rtl): Likewise.
3777 (assemble_start_function): Likewise.
3778 (output_constant): Likewise.
3779 (maybe_assemble_visibility): Likewise.
3780 * varpool.c (ctor_for_folding): Likewise.
3781 * chkp-builtins.def: Remove.
3782 * ipa-chkp.c: Remove.
3783 * ipa-chkp.h: Remove.
3784 * rtl-chkp.c: Remove.
3785 * rtl-chkp.h: Remove.
3786 * tree-chkp-opt.c: Remove.
3787 * tree-chkp.c: Remove.
3788 * tree-chkp.h: Remove.
3790 2018-06-07 Carl Love <cel@us.ibm.com>
3792 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3793 vextract_fp_from_shortl): Add BE support.
3795 2018-06-07 Paul Koning <ni1d@arrl.net>
3797 * compare-elim.c (try_merge_compare): Don't merge compare if
3798 address contains a side effect.
3799 (try_eliminate_compare): Likewise.
3801 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3803 * config.gcc: Support "tremont".
3804 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3805 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3807 * config/i386/i386.c (m_TREMONT): Define.
3808 (processor_target_table): Add "tremont".
3809 (PTA_TREMONT): Define.
3810 (ix86_lea_outperforms): Add TARGET_TREMONT.
3811 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3812 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3813 and M_INTEL_GOLDMONT_PLUS.
3814 (fold_builtin_cpu): Add "tremont".
3815 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3816 (ix86_option_override_internal): Add "tremont".
3817 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3818 (processor_type): Add PROCESSOR_TREMONT.
3819 * config/i386/x86-tune.def: Add m_TREMONT.
3820 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3822 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3824 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3825 symbol defined for msp430i* devices to be lower case.
3827 gcc/testsuite/gcc.target/msp430/
3828 * msp430i-device-symbol.c: New test.
3829 * msp430f-device-symbol.c: New test.
3830 * msp430.h: New test header file.
3832 2018-06-07 Richard Biener <rguenther@suse.de>
3834 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3835 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3836 Properly wrap signed arithmetic if overflow wraps.
3838 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3840 PR tree-optimization/69615
3841 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3842 of a cast from a same precision integral SSA_NAME in a bb dominated
3843 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3844 cast to utype if rhs2 has already a compatible type.
3846 2018-06-07 Richard Biener <rguenther@suse.de>
3848 PR tree-optimization/85935
3849 * graphite-scop-detection.c (find_params_in_bb): Analyze
3850 condition operands with respect to the correct loop. Assert
3851 the analysis doesn't fail.
3853 2018-06-04 Carl Love <cel@us.ibm.com>
3855 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3856 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3857 as it is slightly cheaper.
3858 (first_match_or_eos_index_<mode>):
3859 Calculate index using natural element order.
3860 (first_match_index_<mode>):
3861 Calculate index using natural element order.
3862 (first_match_or_eos_index_<mode>):
3863 Calculate index using natural order.
3864 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3865 for BE and LE modes.
3866 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3867 P9V_BUILTIN_VCLZLSBB_V16QI.
3868 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3871 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3873 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3874 indentation and line wrap for many prototypes. Add missing
3875 @smallexample directives around block of prototypes for vec_xl and
3878 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3880 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3881 track if we pass or return IEEE 128-bit floating point.
3882 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3883 C++ mangling that is compatible with GCC 8.1.
3884 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3885 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3886 floating point types.
3887 (rs6000_function_arg_advance_1): Likewise.
3888 (rs6000_mangle_type): Optionally generate mangled names that match
3889 what GCC 8.1 generated for IEEE 128-bit floating point types.
3890 (rs6000_globalize_decl_name): If we have an external function that
3891 passes or returns IEEE 128-bit types, generate a weak reference
3892 from the mangled name used in GCC 8.1 to the current mangled
3894 (rs6000_init_builtins): Make __ibm128 use the long double type if
3895 long double is IBM extended double. Make __float128 use the long
3896 double type if long double is IEEE 128-bit.
3899 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3900 macro for __ibm128 built-in functions.
3901 (PACK_IF): Add __ibm128 pack/unpack functions.
3902 (UNPACK_IF): Likewise.
3903 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3904 enable long double built-in functions if long double is IEEE
3905 128-bit floating point.
3906 (rs6000_invalid_builtin): Update long double built-in function
3908 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3909 functions, adjust the built-in function to use the long double
3910 built-in function if __ibm128 and long double are the same type.
3911 * doc/extend.texi (PowerPC builtins): Update documention for
3912 __builtin_{,un}pack_longdouble. Add documentation for
3913 __builtin_{,un}pack_ibm128.
3915 2018-06-06 Jim Wilson <jimw@sifive.com>
3917 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3918 (struct machine_function): New field interrupt_mode.
3919 (riscv_handle_type_attribute): New function. Add forward declaration.
3920 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3921 (riscv_expand_epilogue): Check interrupt_mode field.
3922 (riscv_set_current_function): Check interrupt attribute args and
3923 set interrupt_mode field.
3924 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3925 (riscv_sret, riscv_uret): New.
3926 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3927 new arguments to interrupt attribute.
3929 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3932 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3933 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3935 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3937 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3938 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3939 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3940 several redundant entries.
3942 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3944 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3945 type from "rtx" to "rtx_insn *".
3946 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3947 for local "call_insn", removing cast.
3948 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3950 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3952 PR tree-optimization/86066
3953 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3954 for BIT_INSERT_EXPR stores.
3956 2018-06-06 Richard Biener <rguenther@suse.de>
3958 PR tree-optimization/86062
3959 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3960 component refs ontop
3961 of to be offsetted base.
3963 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3965 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3966 to be static and remove check on interrupt attribute name.
3968 gcc/testsuite/gcc.target/msp430/
3969 * function-attributes-4.c: New test.
3970 * static-interrupts.c: New test.
3972 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3974 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3975 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3977 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3980 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3982 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3983 Remove second argument, change how error is called.
3984 (aarch64_layout_arg): Remove second argument from
3985 aarch64_err_no_fpadvsimd call.
3986 (aarch64_init_cumulative_args): Ditto.
3987 (aarch64_gimplify_va_arg_expr): Ditto.
3988 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3990 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3992 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3993 (*simple_return_indirect_internal<mode>): Rename from
3994 simple_return_indirect_internal. Use W mode iterator.
3995 (rstorssp): New expander.
3996 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3997 (clrssbsy): New expander.
3998 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4000 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4002 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4004 (cmse_check_pointed_object): Likewise.
4006 2018-06-05 Martin Liska <mliska@suse.cz>
4008 PR gcov-profile/47618
4009 * doc/invoke.texi: Document how -fprofile-dir format
4012 2018-06-05 Richard Biener <rguenther@suse.de>
4014 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4015 removal pretend DOM info isn't available so we do not update
4016 it and only remove edges, not dominated blocks. Actually free
4017 DOM info in case we removed something. Remove unreachable blocks.
4018 (mfb_keep_latches): Work with either DOM info or marked backedges.
4019 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4020 first. Mark backedges if DOM info isn't available.
4021 (Re-)compute DOM info after cleanup_control_flow_pre.
4023 2018-06-05 Richard Biener <rguenther@suse.de>
4025 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4026 (locus_discrim_hasher::hash): Adjust.
4027 (locus_discrim_hasher::equal): Likewise.
4028 (next_discriminator_for_locus): Work on line directly.
4029 (same_line_p): Pass in expanded locus1 as well.
4030 (assign_discriminators): Avoid redundant location expansions.
4032 2018-06-05 Richard Biener <rguenther@suse.de>
4034 PR tree-optimization/86046
4035 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4036 if required after clearing TREE_ADDRESSABLE.
4038 2018-06-05 Richard Biener <rguenther@suse.de>
4040 PR tree-optimization/86047
4041 * tree-ssa-loop.c (for_each_index): Glob handling of all
4042 decls and constants and really handle all of them.
4044 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4047 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4048 qualifier_void_pointer and qualifier_const_void_pointer.
4049 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4050 (arm_init_builtins): Handle the above.
4051 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4052 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4053 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4056 2018-06-05 Martin Liska <mliska@suse.cz>
4058 * auto-profile.c (read_autofdo_file): Do not use
4059 gcov_ctr_summary struct.
4060 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4061 * coverage.c (struct counts_entry): Likewise.
4062 (read_counts_file): Read just single summary entry.
4063 (get_coverage_counts): Use gcov_summary struct.
4064 * coverage.h (get_coverage_counts): Likewise.
4065 * gcov-dump.c (dump_working_sets): Likewise.
4066 (tag_summary): Dump just single summary.
4067 * gcov-io.c (gcov_write_summary): Write just histogram
4069 (gcov_read_summary): Read just single summary.
4070 (compute_working_sets): Use gcov_summary struct.
4071 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4072 of GCOV_COUNTERS_SUMMABLE.
4073 (GCOV_COUNTERS_SUMMABLE): Remove.
4074 (GCOV_FIRST_VALUE_COUNTER): Replace with
4075 GCOV_COUNTER_V_INTERVAL.
4076 (struct gcov_ctr_summary): Remove.
4077 (struct gcov_summary): Directly use fields of former
4079 (compute_working_sets): Use gcov_summary struct.
4080 * gcov.c (read_count_file): Do not use ctrs fields.
4081 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4083 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4085 * profile.c: Likewise.
4086 * profile.h: Likewise.
4088 2018-06-05 Martin Liska <mliska@suse.cz>
4090 PR gcov-profile/84846
4091 * gcov.c (output_lines): Print working directory only
4092 in intermediate format.
4094 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4096 * config/s390/s390-builtin-types.def: Add void function type.
4097 * config/s390/s390-builtins.def: Use the function type for the
4100 2018-06-04 Jim Wilson <jimw@sifive.com>
4102 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4104 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4105 and maybe_eh_return. Change regno to unsigned int. Use new args to
4106 handle EH_RETURN_DATA_REGNO registers properly.
4107 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4108 (riscv_expand_epilogue): Update comment. Change argument name and
4109 type. Update code to use new name and type. Pass new args to
4110 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4112 * config/riscv/riscv.md (NORMAL_RETURN): New.
4113 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4114 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4115 (eh_return): Call gen_eh_return_internal and emit barrier.
4116 (eh_return_internal): Call riscv_expand_epilogue.
4118 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4120 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4121 bit_insertion field and declare can_be_merged_into method.
4122 (merged_store_group::can_be_merged_into): New method.
4123 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4124 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4125 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4127 2018-06-04 Richard Biener <rguenther@suse.de>
4129 PR tree-optimization/85955
4130 * builtins.c (fold_builtin_sincos): Convert pointers to
4131 destination to appropriate type before dereferencing.
4133 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4135 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4137 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4139 * expr.c (expand_expr_real_1): Force the operand into memory if
4140 its TYPE_MODE is BLKmode and if there is no integer mode for
4141 the number of bits being extracted.
4143 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4147 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4148 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4150 2018-06-04 Richard Biener <rguenther@suse.de>
4152 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4153 (cleanup_tree_cfg_noloop): ... single caller. Do
4154 start_recording_case_labels later.
4156 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4158 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4159 to _IMMINTRIN_H_INCLUDED.
4160 * config/i386/pconfigintrin.h: Ditto.
4161 * config/i386/waitpkgintrin.h: Ditto.
4162 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4163 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4164 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4165 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4166 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4167 waitpkgintrin.h and cldemoteintrin.h.
4169 2018-06-04 Richard Biener <rguenther@suse.de>
4171 PR tree-optimization/86038
4172 * tracer.c (find_best_successor): Check probability for
4173 being initialized, bail out if not.
4175 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4178 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4179 of bits to ignore when comparing architectures.
4181 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4183 PR tree-optimization/69615
4184 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4185 maximum or minimum of the type, try to merge it also as if
4186 range1 is + [-, x - 1] or + [x + 1, -].
4189 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4191 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4193 PR tree-optimization/86034
4194 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4195 the unsigned bitfield type in a bit insertion sequence if it does not
4196 have a larger precision than the bitfield size.
4197 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4199 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4201 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4203 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4205 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4206 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4207 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4208 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4210 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4212 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4213 Disable -fdelete-null-pointer-checks for ELF toolchain.
4215 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4216 Kito Cheng <kito.cheng@gmail.com>
4218 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4219 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4220 (nds32*-*-*): Add glibc and uclibc conditions.
4221 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4222 (TARGET_EXCEPT_UNWIND_INFO): Define.
4223 * config/nds32/elf.h: New file.
4224 * config/nds32/linux.h: New file.
4225 * config/nds32/nds32-elf.opt: New file.
4226 * config/nds32/nds32-linux.opt: New file.
4227 * config/nds32/nds32-fp-as-gp.c
4228 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4229 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4231 (nds32_asm_file_end): Ditto.
4232 (nds32_print_operand): Ditto.
4233 (nds32_insert_attributes): Ditto.
4234 (nds32_init_libfuncs): New function.
4235 (TARGET_HAVE_TLS): Define.
4236 (TARGET_INIT_LIBFUNCS): Define.
4237 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4239 (TARGET_ELF): Apply different mcmodel setting.
4240 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4241 been migrated into elf.h and linux.h files.
4242 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4243 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4244 (mcmodel): The content has been migrated into nds32-elf.opt and
4245 nds32-linux.opt files.
4246 * config/nds32/t-elf: New file.
4247 * config/nds32/t-linux: New file.
4249 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4250 Shiva Chen <shiva0217@gmail.com>
4252 * config/nds32/constants.md (unspec_volatile_element): Add
4253 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4254 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4256 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4257 (make_pass_nds32_fp_as_gp): Declare.
4258 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4260 (nds32_asm_function_end_prologue): Remove unused asm output.
4261 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4262 (nds32_asm_file_start): Output necessary fp_as_gp information.
4263 (nds32_option_override): Adjust register usage.
4264 (nds32_expand_prologue): Consider fp_as_gp situation.
4265 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4266 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4269 (simple_return): Ditto.
4270 (omit_fp_begin): Output special directive for fp_as_gp.
4271 (omit_fp_end): Output special directive for fp_as_gp.
4272 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4273 mforbid-fp-as-gp): New options.
4275 2018-06-01 Mark Wielaard <mark@klomp.org>
4277 * dwarf2out.c (dwarf2out_finish): Remove generation of
4278 DW_AT_loclists_base.
4280 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4282 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4283 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4284 (struct merged_store_group): Add bit_insertion field.
4285 (dump_char_array): Use standard hexadecimal format.
4286 (merged_store_group::merged_store_group): Set bit_insertion to false.
4287 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4288 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4289 also print the mask in the dump file.
4290 (pass_store_merging::gate): Minor tweak.
4291 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4292 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4293 stores with INTEGER_CST stores.
4294 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4295 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4296 and use it throughout. Generate bit insertion sequences if need be.
4297 (pass_store_merging::process_store): Remove redundant condition.
4298 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4300 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4302 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4303 the 128-bit floating point types. Fix function comment.
4305 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4307 * config/aarch64/aarch64-simd.md
4308 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4310 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4313 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4315 PR tree-optimization/85989
4316 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4318 (backprop::intersect_uses): Check it when deciding whether this
4319 is a backedge reference.
4320 (backprop::process_block): Add each phi to m_visited_phis
4321 after visiting it, then clear it at the end.
4323 2018-06-01 Richard Biener <rguenther@suse.de>
4325 * tree-vectorizer.h (vect_dr_stmt): New function.
4326 (vect_get_load_cost): Adjust.
4327 (vect_get_store_cost): Likewise.
4328 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4329 Use vect_dr_stmt instead of DR_SMTT.
4330 (vect_record_base_alignments): Likewise.
4331 (vect_calculate_target_alignment): Likewise.
4332 (vect_compute_data_ref_alignment): Likewise and make static.
4333 (vect_update_misalignment_for_peel): Likewise.
4334 (vect_verify_datarefs_alignment): Likewise.
4335 (vector_alignment_reachable_p): Likewise.
4336 (vect_get_data_access_cost): Likewise. Pass down
4337 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4338 (vect_get_peeling_costs_all_drs): Likewise.
4339 (vect_peeling_hash_get_lowest_cost): Likewise.
4340 (vect_enhance_data_refs_alignment): Likewise.
4341 (vect_find_same_alignment_drs): Likewise.
4342 (vect_analyze_data_refs_alignment): Likewise.
4343 (vect_analyze_group_access_1): Likewise.
4344 (vect_analyze_group_access): Likewise.
4345 (vect_analyze_data_ref_access): Likewise.
4346 (vect_analyze_data_ref_accesses): Likewise.
4347 (vect_vfa_segment_size): Likewise.
4348 (vect_small_gap_p): Likewise.
4349 (vectorizable_with_step_bound_p): Likewise.
4350 (vect_prune_runtime_alias_test_list): Likewise.
4351 (vect_analyze_data_refs): Likewise.
4352 (vect_supportable_dr_alignment): Likewise.
4353 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4354 (vect_gen_prolog_loop_niters): Likewise.
4355 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4356 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4358 (vect_recog_mask_conversion_pattern): Likewise.
4359 (vect_try_gather_scatter_pattern): Likewise.
4360 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4361 to vect_get_store_cost.
4362 (vect_get_store_cost): Get stmt_info instead of DR.
4363 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4364 (vect_get_load_cost): Get stmt_info instead of DR.
4366 2018-06-01 Richard Biener <rguenther@suse.de>
4369 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4370 vars wrapped in MEM_REFs.
4372 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4374 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4375 Fix subreg tests so that we only return a choice between
4376 GENERAL_REGS and FP_REGS if the original classes included both.
4378 2018-06-01 Richard Biener <rguenther@suse.de>
4381 * tree-ssa-structalias.c (get_function_part_constraint):
4382 Handle NULL fi->decl.
4383 (find_func_aliases_for_call): Properly handle indirect
4384 fi from direct call.
4385 (find_func_clobbers): Likewise.
4386 (ipa_pta_execute): Likewise.
4387 (create_variable_info_for): For functions that are ifunc_resolver
4388 resolve to a varinfo that contains the result of the resolver call.
4389 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4392 2018-05-31 Michael Collison <michael.collison@arm.com>
4394 * config/aarch64/aarch64.md:
4395 (*fix_to_zero_extenddfdi2): New pattern.
4396 * gcc.target/aarch64/fix_extend1.c: New testcase.
4398 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4402 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4403 and BUILT_IN_STRNCMP_EQ.
4404 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4405 BUILT_IN_STRNCMP_EQ.
4406 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4407 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4408 (gimple_fold_builtin): Likewise.
4409 * tree-ssa-strlen.c (compute_string_length): New function.
4410 (determine_min_obsize): New function.
4411 (handle_builtin_string_cmp): New function to handle calls to
4412 string compare functions.
4413 (strlen_optimize_stmt): Add handling to builtin string compare
4415 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4416 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4417 * tree.c (build_common_builtin_nodes): Add new defines of
4418 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4420 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4423 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4424 functions with naked attribute.
4426 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4428 * config/i386/sse.md (avx_vec_concat<mode>):
4429 Substitute concat_tg_mode mode attribute with xtg_mode.
4430 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4431 (concat_tg_mode): Remove mode attribute.
4433 2018-05-31 Martin Sebor <msebor@redhat.com>
4436 * calls.c (alloc_max_size): Correct a logic error/typo.
4437 Treat excessive arguments as infinite. Warn for invalid arguments.
4438 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4440 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4443 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4444 and movx for Haswell.
4446 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4447 Cesar Philippidis <cesar@codesourcery.com>
4450 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4451 when emitting error on private/firstprivate reductions.
4452 * omp-low.c (lower_omp_target): Avoid reference-type processing
4453 on pointers for firstprivate clause.
4455 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4457 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4460 * config/aarch64/aarch64-simd.md
4461 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4462 (aarch64_ld1_x3_<mode>): Likewise
4463 (aarch64_st1x2<VALLDIF:mode>): Likewise
4464 (aarch64_st1_x2_<mode>): Likewise
4465 (aarch64_st1x3<VALLDIF:mode>): Likewise
4466 (aarch64_st1_x3_<mode>): Likewise
4467 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4468 (vld1_s8_x3): Likewise.
4469 (vld1_u16_x3): Likewise.
4470 (vld1_s16_x3): Likewise.
4471 (vld1_u32_x3): Likewise.
4472 (vld1_s32_x3): Likewise.
4473 (vld1_u64_x3): Likewise.
4474 (vld1_s64_x3): Likewise.
4475 (vld1_f16_x3): Likewise.
4476 (vld1_f32_x3): Likewise.
4477 (vld1_f64_x3): Likewise.
4478 (vld1_p8_x3): Likewise.
4479 (vld1_p16_x3): Likewise.
4480 (vld1_p64_x3): Likewise.
4481 (vld1q_u8_x3): Likewise.
4482 (vld1q_s8_x3): Likewise.
4483 (vld1q_u16_x3): Likewise.
4484 (vld1q_s16_x3): Likewise.
4485 (vld1q_u32_x3): Likewise.
4486 (vld1q_s32_x3): Likewise.
4487 (vld1q_u64_x3): Likewise.
4488 (vld1q_s64_x3): Likewise.
4489 (vld1q_f16_x3): Likewise.
4490 (vld1q_f32_x3): Likewise.
4491 (vld1q_f64_x3): Likewise.
4492 (vld1q_p8_x3): Likewise.
4493 (vld1q_p16_x3): Likewise.
4494 (vld1q_p64_x3): Likewise.
4495 (vst1_s64_x2): Likewise.
4496 (vst1_u64_x2): Likewise.
4497 (vst1_f64_x2): Likewise.
4498 (vst1_s8_x2): Likewise.
4499 (vst1_p8_x2): Likewise.
4500 (vst1_s16_x2): Likewise.
4501 (vst1_p16_x2): Likewise.
4502 (vst1_s32_x2): Likewise.
4503 (vst1_u8_x2): Likewise.
4504 (vst1_u16_x2): Likewise.
4505 (vst1_u32_x2): Likewise.
4506 (vst1_f16_x2): Likewise.
4507 (vst1_f32_x2): Likewise.
4508 (vst1_p64_x2): Likewise.
4509 (vst1q_s8_x2): Likewise.
4510 (vst1q_p8_x2): Likewise.
4511 (vst1q_s16_x2): Likewise.
4512 (vst1q_p16_x2): Likewise.
4513 (vst1q_s32_x2): Likewise.
4514 (vst1q_s64_x2): Likewise.
4515 (vst1q_u8_x2): Likewise.
4516 (vst1q_u16_x2): Likewise.
4517 (vst1q_u32_x2): Likewise.
4518 (vst1q_u64_x2): Likewise.
4519 (vst1q_f16_x2): Likewise.
4520 (vst1q_f32_x2): Likewise.
4521 (vst1q_f64_x2): Likewise.
4522 (vst1q_p64_x2): Likewise.
4523 (vst1_s64_x3): Likewise.
4524 (vst1_u64_x3): Likewise.
4525 (vst1_f64_x3): Likewise.
4526 (vst1_s8_x3): Likewise.
4527 (vst1_p8_x3): Likewise.
4528 (vst1_s16_x3): Likewise.
4529 (vst1_p16_x3): Likewise.
4530 (vst1_s32_x3): Likewise.
4531 (vst1_u8_x3): Likewise.
4532 (vst1_u16_x3): Likewise.
4533 (vst1_u32_x3): Likewise.
4534 (vst1_f16_x3): Likewise.
4535 (vst1_f32_x3): Likewise.
4536 (vst1_p64_x3): Likewise.
4537 (vst1q_s8_x3): Likewise.
4538 (vst1q_p8_x3): Likewise.
4539 (vst1q_s16_x3): Likewise.
4540 (vst1q_p16_x3): Likewise.
4541 (vst1q_s32_x3): Likewise.
4542 (vst1q_s64_x3): Likewise.
4543 (vst1q_u8_x3): Likewise.
4544 (vst1q_u16_x3): Likewise.
4545 (vst1q_u32_x3): Likewise.
4546 (vst1q_u64_x3): Likewise.
4547 (vst1q_f16_x3): Likewise.
4548 (vst1q_f32_x3): Likewise.
4549 (vst1q_f64_x3): Likewise.
4550 (vst1q_p64_x3): Likewise.
4552 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4554 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4555 user_label_prefix to name.
4557 * tree-core.h: Update comment about the format of NAME string
4558 passed to handler in attribute_spec.
4560 * config/msp430/msp430.md: Remove erroneous subreg expression from
4561 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4562 zero_extend{q,h}isi2.
4564 2018-05-30 Borislav Petkov <bp@suse.de>
4566 * doc/extend.texi: Document some architecture specific
4567 constraints and sort entries.
4569 2018-05-30 Martin Sebor <msebor@redhat.com>
4572 * builtins.c (expand_builtin_stpcpy_1): New function.
4573 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4574 only if the former succeeds.
4576 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4578 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4581 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4583 * doc/invoke.texi (-flinker-output): Document
4585 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4587 * passes.c (ipa_write_summaries): Only modify statements if body
4589 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4590 incrementally linking.
4591 (ipa_passes): Likewise.
4592 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4593 pass down resolution info.
4594 * common.opt (flag_incremental_link): Update info.
4595 * gcc.c (plugin specs): Turn flinker-output=* to
4596 -plugin-opt=-linker-output-known
4597 * toplev.c (compile_file): Also cut compilation when doing incremental
4599 * flag-types. (enum lto_partition_model): Add
4600 LTO_LINKER_OUTPUT_NOLTOREL.
4601 (invoke.texi): Add -flinker-output docs.
4602 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4603 link same way as WPA; do not stream in dead initializers.
4605 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4606 darf2out_register_external_decl): Support incremental link.
4608 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4610 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4612 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4614 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4615 it down to simple_object_copy_lto_debug_sections.
4616 (run_gcc): Determine incremental LTO link time and configure
4617 lto1 into non-wpa mode, disable renaming of debug sections.
4619 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4621 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4622 descriptions of various incorrectly documented functions.
4624 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4627 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4628 address check not strict.
4630 2018-05-30 Richard Biener <rguenther@suse.de>
4632 PR tree-optimization/85964
4633 * tracer.c (better_p): Drop initialized count check, we only
4634 call the function with initialized counts now.
4635 (find_best_successor): Do find a best edge if one
4636 has uninitialized count.
4637 (find_best_predecessor): Likewise. Do BB frequency check only
4638 if count is initialized.
4640 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4642 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4643 (aarch64_ldrstr_offset_compare): New.
4644 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4645 load/store orderings.
4646 (aarch64_gen_adjusted_ldpstp): Likewise.
4648 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4650 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4651 Check for subset of GENERAL_REGS and FP_REGS.
4652 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4655 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4657 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4658 and wi::to_poly_offset. Add the current offset and then check
4659 whether the sum fits, rather than using an unchecked addition of
4660 a checked term. Check for a shwi rather than a uhwi.
4661 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4662 (store_constructor): Use poly_int_tree_p.
4663 (expand_expr_real_1): Likewise.
4664 * function.c (assign_temp): Likewise.
4665 * fold-const.c (const_binop): Use poly_int_tree_p and
4667 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4669 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4670 to_poly_offset for MEM offsets.
4671 * ipa-icf.c (sem_variable::equals): Likewise.
4672 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4673 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4674 wi::to_poly_offset for BIT_FIELD_REF offsets.
4675 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4677 * var-tracking.c (emit_note_insn_var_location): Use
4678 tree_to_poly_uint64.
4680 2018-05-29 Jim Wilson <jimw@sifive.com>
4682 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4684 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4687 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4688 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4690 (sse4_1_round<mode>2): Use nonimmediate_operand
4691 for operand 1 predicate.
4693 2018-05-29 Martin Sebor <msebor@redhat.com>
4694 Richard Biener <rguenther@suse.de>
4697 * calls.c (get_size_range): Call determine_value_range instead
4698 of get_value_range..
4699 * tree-vrp.h (determine_value_range): Declared new function.
4700 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4702 2018-05-29 Richard Biener <rguenther@suse.de>
4704 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4705 sure to use non-pattern stmts for get_earlier_stmt arguments.
4706 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4707 called on pattern stmts.
4708 (get_later_stmt): Likewise.
4710 2018-05-29 Martin Liska <mliska@suse.cz>
4712 PR gcov-profile/85759
4713 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4716 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4718 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4720 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4724 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4725 VEC_PACK_FLOAT_EXPR): New tree codes.
4726 * tree-pretty-print.c (op_code_prio): Handle
4727 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4728 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4729 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4730 * tree-inline.c (estimate_operator_cost): Likewise.
4731 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4732 * fold-const.c (const_binop): Likewise.
4733 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4734 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4735 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4736 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4737 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4738 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4739 * expr.c (expand_expr_real_2): Likewise.
4740 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4741 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4742 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4744 * optabs.c (expand_widen_pattern_expr): For
4745 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4746 sign from result type rather than operand's type.
4747 (expand_binop_directly): For vec_packu_float_optab and
4748 vec_packs_float_optab allow result type to be different from operand's
4750 * optabs-tree.c (optab_for_tree_code): Handle
4751 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4752 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4753 * tree-vect-generic.c (expand_vector_operations_1): Handle
4754 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4755 VEC_PACK_FLOAT_EXPR.
4756 * tree-vect-stmts.c (supportable_widening_operation): Handle
4758 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4759 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4760 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4761 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4762 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4764 (vec_pack<floatprefix>_float_<mode>): New expander.
4765 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4767 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4768 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4769 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4770 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4771 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4773 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4774 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4775 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4776 VEC_PACK_FLOAT_EXPR): Document.
4778 2018-05-29 Richard Biener <rguenther@suse.de>
4780 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4782 (stmt_vec_info_vec): Make pointer.
4783 (init_stmt_vec_info_vec): Remove.
4784 (free_stmt_vec_info_vec): Likewise.
4785 (set_stmt_vec_info_vec): New function.
4786 (free_stmt_vec_infos): Likewise.
4787 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4788 (set_vinfo_for_stmt): Likewise.
4789 (get_earlier_stmt): Likewise.
4790 (get_later_stmt): Likewise.
4791 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4792 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4793 (vec_info::~vec_info): Free stmt_vec_infos.
4794 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4795 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4796 (pass_slp_vectorize::execute): Likewise.
4797 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4798 (free_stmt_vec_info_vec): Likewise.
4799 (set_stmt_vec_info_vec): New function.
4800 (free_stmt_vec_infos): Likewise.
4801 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4802 the global stmt_vec_info_vec.
4803 * tree-parloops.c (gather_scalar_reductions): Use
4804 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4807 2018-05-29 Richard Biener <rguenther@suse.de>
4809 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4811 2018-05-29 Martin Liska <mliska@suse.cz>
4812 David Malcolm <dmalcolm@redhat.com>
4814 * vec.c (test_reverse): New.
4815 (vec_c_tests): Add new test.
4816 * vec.h (vl_ptr>::reverse): New function.
4818 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4820 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4822 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4825 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4827 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4829 2018-05-28 Richard Biener <rguenther@suse.de>
4831 PR tree-optimization/85933
4832 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4833 look at stmts marked as vectorizable.
4835 2018-05-28 Richard Biener <rguenther@suse.de>
4837 PR tree-optimization/85934
4838 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4839 vector boolean check before scalar optimization.
4841 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4843 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4846 2018-05-28 Mark Wielaard <mark@klomp.org>
4848 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4849 if it is an expression containing a minus sign.
4851 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4853 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4855 2018-05-27 Paul Koning <ni1d@arrl.net>
4857 * config/pdp11/pdp11.md (truncsihi2): Remove.
4859 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4860 Chung-Ju Wu <jasonwucj@gmail.com>
4862 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4864 (unaligned_store_dw): Ditto.
4865 * config/nds32/nds32-memory-manipulation.c
4866 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4867 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4868 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4869 (emit_setmem_word_loop): Rename to ...
4870 (emit_setmem_doubleword_loop): ... this.
4871 (nds32_gen_dup_4_byte_to_word_value): New function.
4872 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4873 (nds32_expand_setmem_loop): Refine implementation.
4874 (nds32_expand_setmem_loop_v3m): Ditto.
4875 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4878 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4880 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4882 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4884 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4885 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4886 machine->attr_no_prologue_p.
4887 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4888 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4889 (nds32_expand_epilogue): Consider attr_naked_p.
4890 (nds32_expand_epilogue_v3pop): Likewise.
4891 (nds32_can_use_return_insn): Likewise.
4892 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4893 attr_no_prologue_p fields.
4894 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4896 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4899 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4901 * config/i386/sse.md
4902 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4904 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4906 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4908 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4910 (*<floatsuffix>floatv2div2sf2): Rename to ...
4911 (*float<floatunssuffix>v2div2sf2): ... this.
4912 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4913 (float<floatunssuffix>v2div2sf2_mask): ... this.
4914 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4915 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4916 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4918 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4920 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4922 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4924 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4926 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4928 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4929 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4930 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4931 gen_ufix_truncv8dfv8si2.
4932 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4933 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4934 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4935 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4936 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4937 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4938 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4939 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4941 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4945 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4947 2018-05-25 Jim Wilson <jimw@sifive.com>
4949 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4950 * config/riscv/riscv.c (struct machine_function): Add
4951 interrupt_handler_p and attribute_checked_p fields.
4952 (riscv_attribute_table): Add interrupt.
4953 (riscv_interrupt_type_p): New.
4954 (riscv_save_reg_p): Save extra regs for interrupt handler.
4955 (riscv_use_save_libcall): Return false for interrupt handler.
4956 (riscv_first_stack_step): Add forward declaration.
4957 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4958 for interrupt handler with large frame. Use it for saved reg list.
4959 (riscv_expand_prologue): Move flag_stack_usage_info support to
4960 eliminate duplication.
4961 (riscv_expand_epilogue): Generate mret for interrupt handler.
4962 (riscv_epilogue_uses): New.
4963 (riscv_can_use_return_insn): Return false for interrupt handler.
4964 (riscv_function_ok_for_sibcall): Likewise.
4965 (riscv_set_current_function): Add interrupt handler support.
4966 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4967 * config/riscv/riscv.md (UNSPECV_MRET): New.
4969 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4971 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4973 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4975 PR tree-optimization/85712
4976 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4977 this candidate has already been replaced in-situ by a copy.
4979 2018-05-25 Jason Merrill <jason@redhat.com>
4981 PR c++/80485 - inline function non-zero address.
4982 * symtab.c (nonzero_address): Check DECL_COMDAT.
4984 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4987 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4988 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4989 extension of SImode operation. Use const123_operand predicate.
4990 (*saddsi_1): Remove.
4991 (*saddl_se_1): Ditto.
4993 (*ssubl_se_1): Ditto.
4994 * config/alpha/predicates.md (const123_operand): New predicate.
4995 * config/alpha/constraints.md (P): Use IN_RANGE.
4997 2018-05-25 Richard Biener <rguenther@suse.de>
4999 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5001 (ref_maybe_used_by_stmt_p): Likewise.
5002 (stmt_may_clobber_ref_p): Likewise.
5003 (stmt_may_clobber_ref_p_1): Likewise.
5004 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5006 (ref_maybe_used_by_stmt_p): Likewise.
5007 (stmt_may_clobber_ref_p): Likewise.
5008 (stmt_may_clobber_ref_p_1): Likewise.
5009 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5010 the alias oracle to disambiguate DRs with stmts DR analysis
5012 (vect_analyze_data_refs): Do not give up on not analyzable
5013 DRs for BB vectorization. Remove code truncating the dataref
5016 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5019 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5020 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5021 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5023 2018-05-25 Richard Biener <rguenther@suse.de>
5025 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5026 function, combining stmt data ref gathering and fatal analysis
5028 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5029 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5030 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5031 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5032 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5034 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5036 PR tree-optimization/85720
5037 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5038 SCC if all partitions are builtins.
5039 (version_loop_by_alias_check): New parameter. Generate cancelable
5040 runtime alias check if all partitions are builtins.
5041 (distribute_loop): Update call to above function.
5043 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5045 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5046 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5047 (parm_default_def_partition_arg): Ditto.
5048 (set_parm_default_def_partition): Ditto.
5049 (get_parm_default_def_partitions): Ditto and make it static.
5050 (get_undefined_value_partitions): Ditto and make it static.
5051 (remove_ssa_form): Refactor call to init_var_map here.
5052 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5053 computation for loop region.
5054 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5055 (register_default_def): Delete.
5056 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5057 (parm_default_def_partition_arg): Ditto.
5058 (set_parm_default_def_partition): Ditto.
5059 (get_parm_default_def_partitions): Ditto and make it static.
5060 (get_undefined_value_partitions): Ditto and make it static.
5061 (coalesce_with_default, coalesce_with_default): Update comment.
5062 (create_coalesce_list_for_region): New func factored out from
5063 create_outofssa_var_map.
5064 (populate_coalesce_list_for_outofssa): New func factored out from
5065 create_outofssa_var_map and coalesce_ssa_name.
5066 (create_outofssa_var_map): Delete.
5067 (coalesce_ssa_name): Refactor to support live range computation.
5068 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5069 (get_parm_default_def_partitions): Delete.
5070 (get_undefined_value_partitions): Ditto.
5071 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5072 computation for loop region.
5073 (new_tree_live_info, loe_visit_block): Ditto.
5074 (live_worklist, set_var_live_on_entry): Ditto.
5075 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5076 * tree-ssa-live.h (struct _var_map): New fields.
5077 (init_var_map): Change decl.
5078 (region_contains_p): New.
5080 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5082 * tree-ssa-live.h (live_merge_and_clear): Delete.
5084 2018-05-25 Richard Biener <rguenther@suse.de>
5087 * tree-dump.c (dequeue_and_dump): Remove access to removed
5088 operand 2 of a SWITCH_EXPR.
5090 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5092 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5093 multiplication and division.
5094 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5095 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5096 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5097 (cond_udiv_optab, cond_umod_optab): New optabs.
5098 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5099 (IFN_COND_RDIV): New internal functions.
5100 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5101 TRUNC_MOD_EXPR and RDIV_EXPR.
5102 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5103 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5105 (SVE_INT_BINARY): Include mult.
5106 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5107 (optab, sve_int_op): Handle mult.
5108 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5110 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5111 for SVE_INT_BINARY_SD.
5113 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5115 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5116 (optab, sve_int_op): Handle div and udiv.
5117 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5118 for SVE_INT_BINARY_SD.
5119 (*<optab><mode>3): New insn for the same.
5121 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5123 * tree-vect-patterns.c: Include predict.h.
5124 (vect_recog_divmod_pattern): Restrict check for division support
5125 to when optimizing for size.
5127 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5129 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5130 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5131 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5132 (gimple_match_op::set_op): Likewise.
5133 (gimple_resimplify4): Declare.
5134 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5135 (expr::gen_transform): Likewise.
5136 (decision_tree::gen): Generate a simplification routine for 4 operands.
5137 * gimple-match-head.c (gimple_simplify): Add an overload for
5138 4 operands. In the top-level function, handle up to 4 call
5139 arguments and call gimple_resimplify4.
5140 (gimple_resimplify4): New function.
5141 (build_call_internal): Pass a fourth operand.
5142 (maybe_push_to_seq): Likewise.
5143 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5144 Fold VEC_COND_EXPRs of an operation and a default value into
5145 an IFN_COND_* function if possible.
5146 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5148 (SVE_COND_FP_BINARY): Include them.
5149 (optab, sve_fp_op): Handle them.
5150 (SVE_INT_BINARY_REV): New code iterator.
5151 (SVE_COND_FP_BINARY_REV): New int iterator.
5152 (commutative): New int attribute.
5153 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5155 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5157 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5158 (*cond_<optab><mode>): New patterns for reversed operands.
5160 2018-05-25 Richard Biener <rguenther@suse.de>
5162 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5163 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5164 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5165 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5166 * tree-vect-data-refs.c (everywhere): Adjust users.
5167 * tree-vect-loop.c (everywhere): Likewise.
5168 * tree-vect-slp.c (everywhere): Likewise.
5169 * tree-vect-stmts.c (everywhere): Likewise.
5170 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5172 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5174 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5176 (gcc_cv_as_section_exclude): ... this.
5177 Try Solaris as #exclude syntax.
5178 * configure: Regenerate.
5179 * config.in: Regenerate.
5180 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5182 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5183 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5185 * varasm.c (default_elf_asm_named_section): Don't check if
5186 HAVE_GAS_SECTION_EXCLUDE is defined.
5188 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5190 * doc/md.texi: Update the documentation of the cond_* optabs
5191 to mention the new final operand. Fix GET_MODE_NUNITS call.
5192 Describe the scalar case too.
5193 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5194 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5196 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5197 (get_conditional_internal_fn): Update comment.
5198 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5199 accumulator value as a final argument to conditional functions.
5200 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5201 a define_expand and add an "else" operand. Assert for now that
5202 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5203 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5204 (*cond_<optab><mode>): New patterns.
5205 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5206 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5207 (UNSPEC_COND_EOR): Delete.
5208 (optab): Remove associated mappings.
5209 (SVE_INT_BINARY): New code iterator.
5210 (sve_int_op): Remove int attribute and add "minus" to the code
5212 (SVE_COND_INT_OP): Delete.
5213 (SVE_COND_FP_OP): Rename to...
5214 (SVE_COND_FP_BINARY): ...this.
5216 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5218 * optabs.c (can_reuse_operands_p): New function.
5219 (maybe_legitimize_operands): Try to reuse the results for
5222 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5224 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5225 Add {q} suffix to insn mnemonic.
5227 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5229 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5230 (msp430_warn_func_return): New.
5232 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5234 * fold-const.c (tree_nonzero_bits): New function.
5235 * fold-const.h (tree_nonzero_bits): Likewise.
5236 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5237 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5239 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5243 * varasm.c (assemble_alias): Check ifunc_resolver only on
5246 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5249 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5250 when memory input operand is handled.
5252 2018-05-24 Luis Machado <luis.machado@linaro.org>
5254 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5256 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5258 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5260 * match.pd: Delay FMA folds until after vectorization.
5262 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5265 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5266 address check not strict.
5268 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5270 * gimple-match.h (gimple_match_op): New class.
5271 (mprts_hook): Replace parameters with a gimple_match_op *.
5272 (maybe_build_generic_op): Likewise.
5273 (gimple_simplified_result_is_gimple_val): Replace parameters with
5274 a const gimple_match_op *.
5275 (gimple_simplify): Replace code_helper * and tree * parameters with
5276 a gimple_match_op * parameter.
5277 (gimple_resimplify1): Replace code_helper *, tree and tree *
5278 parameters with a gimple_match_op * parameter.
5279 (gimple_resimplify2): Likewise.
5280 (gimple_resimplify3): Likewise.
5281 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5282 parameters with a gimple_match_op * parameter.
5283 * gimple-match-head.c (gimple_simplify): Change prototypes of
5284 auto-generated functions to take a gimple_match_op * instead of
5285 separate code_helper * and tree * parameters. Make the same
5286 change in the top-level overload and update calls to the
5287 gimple_resimplify routines. Update calls to the auto-generated
5288 functions and to maybe_push_res_to_seq in the publicly-facing
5289 operation-specific gimple_simplify overloads.
5290 (gimple_match_op::MAX_NUM_OPS): Define.
5291 (gimple_resimplify1): Replace rcode and ops with a single res_op
5292 parameter. Update call to gimple_simplify.
5293 (gimple_resimplify2): Likewise.
5294 (gimple_resimplify3): Likewise.
5295 (mprts_hook): Replace parameters with a gimple_match_op *.
5296 (maybe_build_generic_op): Likewise.
5297 (build_call_internal): Replace type, nargs and ops with
5298 a gimple_match_op *.
5299 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5300 with a single gimple_match_op *. Update calls to mprts_hook,
5301 build_call_internal and gimple_simplified_result_is_gimple_val.
5302 Factor out code that is common to the tree_code and combined_fn cases.
5303 * genmatch.c (expr::gen_transform): Replace tem_code and
5304 tem_ops with a gimple_match_op called tem_op. Update calls
5305 to the gimple_resimplify functions and maybe_push_res_to_seq.
5306 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5307 res_ops. Update call to the gimple_resimplify functions.
5308 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5309 (decision_tree::gen): Make the functions take a gimple_match_op *
5310 called res_op instead of separate res_code and res_ops parameters.
5311 Update call accordingly.
5312 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5313 and ops with a single res_op parameter. Update calls to
5314 maybe_build_generic_op and maybe_push_res_to_seq.
5315 (fold_stmt_1): Update calls to gimple_simplify and
5316 replace_stmt_with_simplification.
5317 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5318 and gimple_simplified_result_is_gimple_val.
5319 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5321 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5322 with a gimple_match_op *.
5323 (vn_nary_build_or_lookup): Likewise. Update call to
5324 vn_nary_build_or_lookup_1.
5325 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5326 gimple_match_op *. Update calls to the gimple_resimplify routines
5327 and to gimple_simplified_result_is_gimple_val.
5328 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5329 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5330 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5331 (visit_nary_op): Likewise.
5332 (visit_reference_op_load): Likewise.
5334 2018-05-23 Luis Machado <luis.machado@linaro.org>
5336 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5337 modifier for printing the step amount.
5339 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5342 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5345 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5347 * doc/sourcebuild.texi (Endianness): New subsubsection.
5349 2018-05-23 Luis Machado <luis.machado@linaro.org>
5351 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5352 <prefetch_dynamic_strides>: New const bool field.
5353 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5354 prefetch_dynamic_strides.
5355 (exynosm1_prefetch_tune): Likewise.
5356 (thunderxt88_prefetch_tune): Likewise.
5357 (thunderx_prefetch_tune): Likewise.
5358 (thunderx2t99_prefetch_tune): Likewise.
5359 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5361 (aarch64_override_options_internal): Update to set
5362 PARAM_PREFETCH_DYNAMIC_STRIDES.
5363 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5364 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5365 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5366 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5367 prefetch-dynamic-strides setting.
5369 2018-05-23 Luis Machado <luis.machado@linaro.org>
5371 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5372 <minimum_stride>: New const int field.
5373 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5374 minimum_stride field defaulting to -1.
5375 (exynosm1_prefetch_tune): Likewise.
5376 (thunderxt88_prefetch_tune): Likewise.
5377 (thunderx_prefetch_tune): Likewise.
5378 (thunderx2t99_prefetch_tune): Likewise.
5379 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5380 <default_opt_level>: Set to 3.
5381 (aarch64_override_options_internal): Update to set
5382 PARAM_PREFETCH_MINIMUM_STRIDE.
5383 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5384 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5385 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5386 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5387 stride is constant and is below the minimum stride threshold.
5389 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5391 * config/arm/arm-cpus.in (mode26): Delete.
5392 (armv4): Delete mode26 reference.
5393 * config/arm/arm.c (arm_configure_build_target): Delete use of
5396 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5398 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5400 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5401 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5403 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5404 (floatunsdidf2): Ditto.
5406 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5408 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5409 (fixuns_trunc<mode>si2_avx512f): Ditto.
5410 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5411 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5412 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5414 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5416 PR rtl-optimization/79985
5417 * df-scan.c (df_insn_refs_collect): Remove special case for
5418 global registers and asm statements.
5420 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5422 * extend.texi (Global Register Variables): Rewrite the bullet list.
5423 Note that the register is available for allocation. Note that access
5424 via inline asm must use constraints. Add note about async-signal
5425 handlers. Remove paragraph about automagic register selection.
5427 2018-05-23 Richard Biener <rguenther@suse.de>
5429 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5430 of fixed offset from memset VN.
5432 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5434 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5436 (alloc_cand_and_find_basis): Initialize first_interp field.
5437 (slsr_process_mul): Modify first_interp field.
5438 (slsr_process_add): Likewise.
5439 (slsr_process_cast): Modify first_interp field for each new
5441 (slsr_process_copy): Likewise.
5442 (dump_candidate): Dump first_interp field.
5443 (replace_mult_candidate): Process all interpretations, not just
5445 (replace_rhs_if_not_dup): Likewise.
5446 (replace_one_candidate): Likewise.
5448 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5450 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5452 (aarch64_needs_frame_chain): New function.
5453 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5455 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5458 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5459 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5460 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5461 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5462 as true for strict-align.
5463 (aarch64_can_inline_p): Perform checks even when callee has no
5464 attributes to check for strict alignment.
5465 * doc/extend.texi (AArch64 Function Attributes): Document
5467 * doc/invoke.texi: (AArch64 Options): Likewise.
5469 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5471 PR tree-optimization/85853
5472 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5473 the handling of the root of the node to...
5474 (vect_slp_analyze_node_operations_1): ...this new function,
5475 and run the whole thing with the child nodes' def types
5476 set according to their SLP node's def type.
5478 2018-05-23 Richard Biener <rguenther@suse.de>
5481 * tree-data-ref.c (create_runtime_alias_checks): Defer
5482 and ignore overflow warnings.
5484 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5486 PR tree-optimization/85822
5487 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5490 2018-05-23 Richard Biener <rguenther@suse.de>
5492 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5493 memset constants via native_interpret_expr.
5495 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5498 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5500 (cgraph_node::create_alias): Likewise.
5501 (cgraph_node::get_availability): Check ifunc_resolver instead
5502 of looking up ifunc attribute.
5503 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5504 * varasm.c (do_assemble_alias): Likewise.
5505 (assemble_alias): Likewise.
5506 (default_binds_local_p_3): Likewise.
5507 * cgraph.h (cgraph_node): Add ifunc_resolver.
5508 (cgraph_node::only_called_directly_or_aliased_p): Return false
5510 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5512 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5513 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5514 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5515 instead of looking up ifunc attribute.
5517 2018-05-22 Luis Machado <luis.machado@linaro.org>
5519 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5521 2018-05-22 Martin Sebor <msebor@redhat.com>
5524 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5525 only when expasion succeeds.
5526 (expand_builtin_strcmp): Same.
5527 (expand_builtin_strncmp): Same.
5529 2018-05-22 Martin Sebor <msebor@redhat.com>
5531 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5533 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5534 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5536 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5537 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5539 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5540 Add check for legitimate_address.
5541 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5542 (aarch64_swap_ldrstr_operands): New.
5543 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5546 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5547 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5549 * config/aarch64/aarch64.md: New patterns to generate stp
5551 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5552 single words and double words.
5553 (load_pair_sw, load_pair_dw): Likewise.
5554 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5556 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5558 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5559 for different mode ldpstp and add peephole for merged zero stores.
5561 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5563 (aarch64_gen_store_pair): Rename calls to match new patterns.
5564 (aarch64_gen_load_pair): Rename calls to match new patterns.
5565 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5566 (load_pair<DREG:mode><DREG2:mode>): ... This.
5567 (store_pair<mode>): Rename to...
5568 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5569 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5571 (V_INT_EQUIV): Handle SImode.
5572 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5575 2018-05-22 Martin Sebor <msebor@redhat.com>
5578 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5579 or ajust the presumed bound on an operation to avoid unnecessary
5582 2018-05-22 Martin Sebor <msebor@redhat.com>
5584 PR tree-optimization/85826
5585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5586 assuming that a DECL necesarily has a constant size.
5588 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5591 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5593 2018-05-22 Richard Biener <rguenther@suse.de>
5595 PR tree-optimization/85834
5596 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5597 non-constant and non-zero memset arguments.
5599 2018-05-22 Martin Liska <mliska@suse.cz>
5602 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5604 2018-05-22 Richard Biener <rguenther@suse.de>
5606 PR tree-optimization/85863
5607 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5608 comparisons when vectype is specified.
5609 (vectorizable_condition): Do not specify vectype for
5610 vect_is_simple_cond when SLP vectorizing.
5612 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5615 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5616 define __ibm128 as long double.
5617 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5618 as a distinct type when IEEE 128-bit support is enabled.
5619 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5620 128-bit types to use the correct functions.
5621 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5622 convert between 128-bit floating point types that have different
5623 modes but the same representation, instead of using gen_lowpart to
5625 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5627 (IFKF_reg): New attributes to give the register constraints for
5629 (extend<mode>tf2_internal): New insns to mark an explicit
5630 conversion between 128-bit floating point types that have a
5631 different mode but share the same representation.
5633 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5635 PR tree-optimization/85814
5636 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5637 a null return from get_strinfo when unsharing the next
5638 strinfo in the chain.
5640 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5643 * varasm.c (weak_finish): Clean up weak_decls.
5645 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5647 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5648 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5649 UNSPEC_UADALP values.
5650 * config/aarch64/iterators.md (ABAL): New int iterator.
5653 (sur): Add mappings for the above.
5654 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5656 (aarch64_<sur>abal<mode>_4): Likewise.
5657 (aarch64_<sur>adalp<mode>_3): Likewise.
5658 (<sur>sadv16qi): New define_expand.
5660 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5662 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5663 (*movdf_internal): Ditto.
5664 (*rcpsf2_sse): Ditto.
5665 (*rsqrtsf2_sse): Ditto.
5666 (*sqrt<mode>2_sse): Ditto.
5668 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5670 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5672 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5673 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5674 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5675 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5677 * config/aarch64/arm_neon.h: Likewise.
5678 * config/aarch64/iterators.md (VQ_I): New.
5680 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5682 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5684 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5687 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5689 * config/nds32/constraints.md (S): New constraint.
5690 * config/nds32/nds32.md (call_internal): Use constraint S.
5691 (call_value_internal): Likewise.
5692 (sibcall_internal): Likewise.
5693 (sibcall_value_internal): Likewise.
5695 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5696 Chung-Ju Wu <jasonwucj@gmail.com>
5698 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5701 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5702 Chung-Ju Wu <jasonwucj@gmail.com>
5704 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5705 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5706 (nds32_rtx_costs_impl): Simplify.
5707 (nds32_address_cost_impl): Simplify.
5708 (nds32_init_rtx_costs): New function.
5709 (nds32_rtx_costs_speed_prefer): Likewise.
5710 (nds32_rtx_costs_size_prefer): Likewise.
5711 (nds32_address_cost_speed_prefer): Likewise.
5712 (nds32_address_cost_speed_fwprop): Likewise.
5713 (nds32_address_cost_size_prefer): Likewise.
5714 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5715 * config/nds32/nds32.c (nds32_option_override): Use
5716 nds32_init_rtx_costs function.
5718 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5720 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5721 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5722 (TARGET_PIPELINE_N8): Likewise.
5723 (TARGET_PIPELINE_N10): Likewise.
5724 (TARGET_PIPELINE_N13): Likewise.
5725 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5727 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5729 * config/nds32/nds32-fpu.md: Update copyright year.
5731 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5733 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5735 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5737 * config/nds32/nds32.c
5738 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5739 * config/nds32/nds32.opt (minline-asm-r15): New option.
5741 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5743 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5745 * config/nds32/nds32.md (abssi2): New pattern.
5747 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5749 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5750 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5751 Merge insn pattern from sse_cvtsi2ss<round_name> and
5752 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5753 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5754 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5755 using SWI48 mode iterator.
5756 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5757 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5758 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5759 pattern from sse_cvttss2si<round_saeonly_name>
5760 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5761 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5762 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5763 using SWI48 mode iterator.
5764 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5765 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5766 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5767 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5768 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5769 using SWI48 mode iterator.
5770 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5771 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5772 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5773 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5774 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5775 SWI48 mode iterator.
5776 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5777 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5778 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5779 pattern from sse_cvttsd2si<round_saeonly_name>
5780 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5782 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5784 * config/nds32/nds32-md-auxiliary.c
5785 (nds32_valid_smw_lwm_base_p): Refine.
5786 (nds32_output_smw_single_word): Refine.
5787 (nds32_output_smw_double_word): New.
5788 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5790 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5792 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5793 (nds32_output_stack_pop): Refine.
5794 (nds32_expand_unaligned_load): Refine.
5795 (nds32_expand_unaligned_store): Refine.
5797 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5798 Chung-Ju Wu <jasonwucj@gmail.com>
5800 * config/nds32/constants.md: Add TP_REGNUM constant.
5801 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5802 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5804 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5805 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5806 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5807 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5808 and PIC code generation.
5809 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5811 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5813 * config/nds32/nds32.md: Support TLS and PIC.
5814 * config/nds32/nds32.c: Support TLS and PIC.
5815 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5816 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5819 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5821 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5822 mode with E_ prefix.
5824 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5825 Chung-Ju Wu <jasonwucj@gmail.com>
5827 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5828 * config/nds32/nds32-md-auxiliary.c
5829 (symbolic_reference_mentioned_p): New.
5830 (nds32_legitimize_ict_address): New.
5831 (nds32_expand_ict_move): New.
5832 (nds32_indirect_call_referenced_p): New.
5833 (nds32_symbol_binds_local_p): Delete.
5834 (nds32_long_call_p): Modify.
5835 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5836 * config/nds32/nds32-protos.h
5837 (symbolic_reference_mentioned_p): Declare.
5838 (nds32_legitimize_ict_address): Declare.
5839 (nds32_expand_ict_move): Declare.
5840 (nds32_indirect_call_referenced_p): Declare.
5841 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5842 (nds32_relax_group): Use nds32_ict_const_p as condition.
5843 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5844 (nds32_asm_file_start): Output ict_model directive in asm code.
5845 (nds32_legitimate_address_p): Consider indirect call.
5846 (nds32_print_operand): Consider indirect call.
5847 (nds32_print_operand_address): Consider indirect call.
5848 (nds32_insert_attributes): Handle "indirect_call" attribute.
5849 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5850 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5851 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5852 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5853 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5854 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5855 (TARGET_ICT_MODEL_SMALL): Define.
5856 (TARGET_ICT_MODEL_LARGE): Define.
5857 * config/nds32/nds32.md (movsi): Consider ict model.
5858 (call, call_value): Consider ict model.
5859 (sibcall, sibcall_value): Consider ict model.
5860 * config/nds32/nds32.opt (mict-model): New option.
5861 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5864 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5865 Monk Chiang <sh.chiang04@gmail.com>
5866 Jim Wilson <jimw@sifive.com>
5868 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5869 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5870 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5871 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5872 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5873 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5874 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5875 compute save_libcall_adjustment properly.
5876 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5877 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5878 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5879 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5880 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5881 (ABI_SPEC): Handle mabi=ilp32e.
5882 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5883 (RVE): Add RVE mask.
5884 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5885 <-march>: Add rv32e as an example.
5887 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5890 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5891 (intra_create_variable_infos): Handle C++ constructors.
5893 2018-05-18 Martin Liska <mliska@suse.cz>
5895 * passes.def: Remove a redundant pass.
5897 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5900 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5902 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5904 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5906 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5908 (armv2, armv2a, armv3, armv3m): Delete architectures.
5909 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5910 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5911 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5913 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5914 (*mulsidi3adddi): Likewise.
5915 (mulsidi3): Likewise.
5916 (*mulsidi3_nov6): Likewise.
5917 (umulsidi3): Likewise.
5918 (umulsidi3_nov6): Likewise.
5919 (umaddsidi4): Likewise.
5920 (*umulsidi3adddi): Likewise.
5921 (smulsi3_highpart): Likewise.
5922 (*smulsi3_highpart_nov6): Likewise.
5923 (umulsi3_highpart): Likewise.
5924 (*umulsi3_highpart_nov6): Likewise.
5925 * config/arm/arm.h (arm_arch3m): Delete.
5926 * config/arm/arm.c (arm_arch3m): Delete.
5927 (arm_option_override_internal): Update armv3-related comment.
5928 (arm_configure_build_target): Delete use of isa_bit_mode32.
5929 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5930 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5931 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5933 (mulusa3): Likewise.
5934 * config/arm/arm-protos.h (arm_arch3m): Delete.
5935 * config/arm/arm-tables.opt: Regenerate.
5936 * config/arm/arm-tune.md: Likewise.
5937 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5938 deleted architectures.
5940 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5942 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5943 (armv5t, armv5te): New features.
5944 (ARMv5, ARMv5e): Delete fgroups.
5945 (ARMv5t, ARMv5te): Adjust for above changes.
5947 (armv5, armv5e): Delete arches.
5948 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5950 (*call_reg_arm): Likewise.
5951 (*call_value_reg_armv5): Likewise.
5952 (*call_value_reg_arm): Likewise.
5953 (*call_symbol): Likewise.
5954 (*call_value_symbol): Likewise.
5955 (*sibcall_insn): Likewise.
5956 (*sibcall_value_insn): Likewise.
5958 (prefetch): Likewise.
5959 (define_split and define_peephole2 dependent on arm_arch5):
5961 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5963 (TARGET_ARM_QBIT): Likewise.
5964 (TARGET_DSP_MULTIPLY): Likewise.
5965 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5966 (arm_arch5, arm_arch5e): Delete.
5967 (arm_arch5t, arm_arch5te): Declare.
5968 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5969 (arm_arch5t): Declare.
5970 (arm_option_reconfigure_globals): Update for the above.
5971 (arm_options_perform_arch_sanity_checks): Update comment, replace
5972 use of arm_arch5 with arm_arch5t.
5973 (use_return_insn): Likewise.
5974 (arm_emit_call_insn): Likewise.
5975 (output_return_instruction): Likewise.
5976 (arm_final_prescan_insn): Likewise.
5977 (arm_coproc_builtin_available): Likewise.
5978 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5979 arm_arch5e with arm_arch5t and arm_arch5te.
5980 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5981 (arm_arch5t, arm_arch5te): Declare.
5982 * config/arm/arm-tables.opt: Regenerate.
5983 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5984 * config/arm/t-multilib: Likewise.
5985 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5986 instead of arm_arch5.
5987 (*call_reg_thumb1): Likewise.
5988 (*call_value_reg_thumb1_v5): Likewise.
5989 (*call_value_reg_thumb1): Likewise.
5990 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5992 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5994 2018-05-18 Martin Liska <mliska@suse.cz>
5996 PR gcov-profile/84846
5997 * doc/gcov.texi: Document -t option of gcov tool.
5999 2018-05-18 Martin Liska <mliska@suse.cz>
6001 PR gcov-profile/84846
6002 * gcov.c (print_usage): Add new -t option.
6003 (process_args): Handle the option.
6004 (generate_results): Use stdout as output when requested by
6007 2018-05-18 Martin Liska <mliska@suse.cz>
6009 PR gcov-profile/84846
6010 * coverage.c (coverage_init): Write PWD to .gcno file.
6011 * doc/gcov.texi: Document how working directory is printed.
6012 * gcov-dump.c (dump_gcov_file): Print PWD.
6013 * gcov.c (output_intermediate_file): Likewise.
6014 (read_graph_file): Read PWD string.
6015 (output_lines): Print PWD.
6017 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6020 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6021 for retval and return false if all args to phi are zero.
6023 2018-05-18 Richard Biener <rguenther@suse.de>
6025 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6027 (evrp_dom_walker::before_dom_children): Call it.
6029 2018-05-18 Richard Biener <rguenther@suse.de>
6031 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6032 results when processing array refs with variable index.
6034 2018-05-18 Toon Moene <toon@moene.org>
6036 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6037 directly after that of -floop-interchange. Indicate that both
6038 options are enabled by default when specifying -O3.
6040 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6042 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6043 iterator. Delete separate integer-mode vec_set<mode> expander.
6044 (aarch64_simd_vec_setv2di): Delete.
6045 (vec_setv2di): Delete.
6046 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6047 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6048 the "w, r" alternative.
6050 2018-05-18 Martin Liska <mliska@suse.cz>
6052 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6053 * tree-pass.h (make_pass_lower_switch_O0): New function.
6054 * tree-switch-conversion.c (node_has_low_bound): Remove.
6055 (node_has_high_bound): Likewise.
6056 (node_is_bounded): Likewise.
6057 (class pass_lower_switch): Make it a template type and create
6059 (pass_lower_switch::execute): Add template argument.
6060 (make_pass_lower_switch): New function.
6061 (make_pass_lower_switch_O0): New function.
6062 (do_jump_if_equal): Remove.
6063 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6064 all the hard work to tree optimization passes.
6066 2018-05-18 Martin Liska <mliska@suse.cz>
6068 * dbgcnt.c (limit_low): Renamed from limit.
6069 (limit_high): New variable.
6070 (dbg_cnt_is_enabled): Check for upper limit.
6071 (dbg_cnt): Adjust dumping.
6072 (dbg_cnt_set_limit_by_index): Add new argument for high
6074 (dbg_cnt_set_limit_by_name): Likewise.
6075 (dbg_cnt_process_single_pair): Parse new format.
6076 (dbg_cnt_process_opt): Use strtok.
6077 (dbg_cnt_list_all_counters): Remove 'value' and add
6079 * doc/invoke.texi: Document changes.
6081 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6083 * doc/sourcebuild.texi (scalar_all_fma): Document.
6084 * tree.def (FMA_EXPR): Delete.
6085 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6086 * internal-fn.c (ternary_direct): New macro.
6087 (expand_ternary_optab_fn): Likewise.
6088 (direct_ternary_optab_supported_p): Likewise.
6089 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6090 * builtins.c (fold_builtin_fma): Delete.
6091 (fold_builtin_3): Don't call it.
6092 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6093 * expr.c (expand_expr_real_2): Likewise.
6094 * fold-const.c (operand_equal_p): Likewise.
6095 (fold_ternary_loc): Likewise.
6096 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6097 * gimple.c (DEFTREECODE): Likewise.
6098 * gimplify.c (gimplify_expr): Likewise.
6099 * optabs-tree.c (optab_for_tree_code): Likewise.
6100 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6101 * tree-eh.c (operation_could_trap_p): Likewise.
6102 (stmt_could_throw_1_p): Likewise.
6103 * tree-inline.c (estimate_operator_cost): Likewise.
6104 * tree-pretty-print.c (dump_generic_node): Likewise.
6105 (op_code_prio): Likewise.
6106 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6107 * tree-ssa-operands.c (get_expr_operands): Likewise.
6108 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6109 * fold-const-call.h (fold_fma): Delete.
6110 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6111 CFN_FNMA and CFN_FNMS.
6113 * genmatch.c (combined_fn): New enum.
6114 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6115 (commutative_op): New function.
6116 (commutate): Use it. Handle more than 2 operands.
6117 (dt_operand::gen_gimple_expr): Use commutative_op.
6118 (parser::parse_expr): Allow :c to be used with non-binary
6119 operators if the commutative operand is known.
6120 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6121 CFN_FMS, CFN_FNMA and CFN_FNMS.
6122 (backprop::process_assign_use): Remove FMA_EXPR handling.
6123 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6124 (gen_hsa_fma): New function.
6125 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6126 IFN_FNMA and IFN_FNMS.
6127 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6128 * gimple-fold.h (follow_all_ssa_edges): Declare.
6129 * gimple-fold.c (follow_all_ssa_edges): New function.
6130 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6131 gimple_build interface and use follow_all_ssa_edges to fold the result.
6132 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6133 instead of checking for optabs directly.
6134 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6135 rather than FMA_EXPRs.
6136 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6137 call to IFN_FMA instead of an FMA_EXPR.
6139 2018-05-17 Jim Wilson <jimw@sifive.com>
6141 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6142 sign extended promoted subreg, and the range does not have the sign bit
6143 set, then do a sign extend.
6145 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6146 test, check for sign extended subreg and/or constant operands, and
6147 do a sign extend in that case.
6149 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6151 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6152 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6154 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6155 Change logics_shift_reg to logics_shift_imm.
6156 (thunderx2t99_fp_loadpair_basic): Delete.
6157 (thunderx2t99_fp_storepair_basic): Delete.
6158 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6159 (thunderx2t99_asimd_polynomial): Delete.
6160 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6161 and neon_fp_mul_d_scalar_q.
6162 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6163 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6164 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6165 (thunderx2t99_asimd_lut): Add missing tbl types.
6166 (thunderx2t99_asimd_ext): Delete.
6167 (thunderx2t99_asimd_load1_1_mult): Delete.
6168 (thunderx2t99_asimd_load1_2_mult): Delete.
6169 (thunderx2t99_asimd_load1_ldp): New.
6170 (thunderx2t99_asimd_load1): New.
6171 (thunderx2t99_asimd_load2): Add missing *load2* types.
6172 (thunderx2t99_asimd_load3): New.
6173 (thunderx2t99_asimd_load4): New.
6174 (thunderx2t99_asimd_store1_1_mult): Delete.
6175 (thunderx2t99_asimd_store1_2_mult): Delete.
6176 (thunderx2t99_asimd_store2_mult): Delete.
6177 (thunderx2t99_asimd_store2_onelane): Delete.
6178 (thunderx2t99_asimd_store_stp): New.
6179 (thunderx2t99_asimd_store1): New.
6180 (thunderx2t99_asimd_store2): New.
6181 (thunderx2t99_asimd_store3): New.
6182 (thunderx2t99_asimd_store4): New.
6184 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6186 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6187 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6189 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6190 Segher Boessenkool <segher@kernel.crashing.org>
6193 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6196 2018-05-17 Richard Biener <rguenther@suse.de>
6198 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6199 for pruning loop and prune defs feeding only already visited PHIs.
6201 2018-05-17 Richard Biener <rguenther@suse.de>
6203 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6205 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6206 Richard Biener <rguenther@suse.de>
6208 PR tree-optimization/85793
6209 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6210 for VMAT_ELEMENTWISE.
6212 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6214 * internal-fn.h (lookup_internal_fn): Declare
6215 * internal-fn.c (lookup_internal_fn): New function.
6216 * gimple.c (gimple_build_call_from_tree): Handle calls to
6218 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6219 internal function names.
6220 * tree-pretty-print.c (dump_generic_node): Likewise.
6221 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6223 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6225 * gimple-fold.h (gimple_build): Make the function forms take
6226 combined_fn rather than built_in_function.
6227 (gimple_simplify): Likewise.
6228 * gimple-match-head.c (gimple_simplify): Likewise.
6229 * gimple-fold.c (gimple_build): Likewise.
6230 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6231 rather than gimple_build_call_internal.
6232 (get_initial_defs_for_reduction): Likewise.
6233 (vect_create_epilog_for_reduction): Likewise.
6234 (vectorizable_live_operation): Likewise.
6236 2018-05-17 Martin Liska <mliska@suse.cz>
6238 * gimple-ssa-sprintf.c (format_directive): Do not use
6239 space in between 'G_' and '('.
6241 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6244 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6245 even if the mask is not all ones.
6248 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6250 (ix86_gimple_fold_builtin): Likewise.
6253 * config/i386/i386.c: Include tree-vector-builder.h.
6254 (ix86_vector_shift_count): New function.
6255 (ix86_fold_builtin): Fold shift builtins by scalar count.
6256 (ix86_gimple_fold_builtin): Likewise.
6258 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6259 _mm512_setzero): New intrinsics.
6261 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6262 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6264 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6265 code generation for cases where splatting a value is not useful.
6266 * simplify-rtx.c (simplify_ternary_operation): Simplify
6267 vec_merge across a vec_duplicate and a paradoxical subreg forming
6268 a vector mode to a vec_concat.
6270 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6272 * config.gcc: Support "goldmont-plus".
6273 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6275 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6276 PROCESSOR_GOLDMONT_PLUS.
6277 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6278 (processor_target_table): Add "goldmont-plus".
6279 (PTA_GOLDMONT_PLUS): Define.
6280 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6281 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6282 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6283 (fold_builtin_cpu): Add "goldmont-plus".
6284 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6285 (ix86_option_override_internal): Add "goldmont-plus".
6286 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6287 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6288 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6289 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6291 2018-05-17 Richard Biener <rguenther@suse.de>
6293 PR tree-optimization/85757
6294 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6295 remove defs that only feed that PHI from further processing.
6297 2018-05-16 Jim Wilson <jimw@sifive.com>
6299 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6301 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6303 2018-05-16 Mark Wielaard <mark@klomp.org>
6305 * dwarf2out.c (count_index_strings): New function.
6306 (output_indirect_strings): Call count_index_strings and generate
6307 header for dwarf_version >= 5.
6309 2018-05-16 Mark Wielaard <mark@klomp.org>
6311 * dwarf2out.c (dwarf_FORM): New function.
6312 (set_indirect_string): Use dwarf_FORM.
6313 (reset_indirect_string): Likewise.
6314 (size_of_die): Likewise.
6315 (value_format): Likewise.
6316 (output_die): Likewise.
6317 (add_skeleton_AT_string): Likewise.
6318 (output_macinfo_op): Likewise.
6319 (index_string): Likewise.
6320 (output_index_string_offset): Likewise.
6321 (output_index_string): Likewise.
6322 (count_index_strings): Likewise.
6324 2018-05-16 Carl Love <cel@us.ibm.com>
6326 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6327 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6329 2018-05-16 Martin Jambor <mjambor@suse.cz>
6331 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6332 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6334 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6336 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6337 (fnma<mode>4): Likewise.
6338 (fms<mode>4): Likewise.
6339 (fnms<mode>4): Likewise.
6340 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6341 (aarch64_fnma<mode>4): Likewise.
6342 (aarch64_fms<mode>4): Likewise.
6343 (aarch64_fnms<mode>4): Likewise.
6344 (aarch64_fnmadd<mode>4): Likewise.
6346 2018-05-16 Jason Merrill <jason@redhat.com>
6348 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6350 2018-05-16 Richard Biener <rguenther@suse.de>
6352 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6353 (dump_stmt_cost): Declare.
6354 (add_stmt_cost): Dump cost we add.
6355 (add_stmt_costs): New function.
6356 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6358 (vect_analyze_stmt): Adjust prototype.
6359 (vectorizable_condition): Likewise.
6360 (vectorizable_live_operation): Likewise.
6361 (vectorizable_reduction): Likewise.
6362 (vectorizable_induction): Likewise.
6363 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6364 cost vector to pass to vectorizable_ and record afterwards.
6365 (vect_model_reduction_cost): Take cost vector argument and adjust.
6366 (vect_model_induction_cost): Likewise.
6367 (vectorizable_reduction): Likewise.
6368 (vectorizable_induction): Likewise.
6369 (vectorizable_live_operation): Likewise.
6370 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6371 SLP_TREE_NUMBER_OF_VEC_STMTS.
6372 (vect_analyze_slp_cost_1): Remove.
6373 (vect_analyze_slp_cost): Likewise.
6374 (vect_slp_analyze_node_operations): Take visited args and
6375 a target cost vector. Avoid processing already visited stmt sets.
6376 (vect_slp_analyze_operations): Use a local cost vector to gather
6377 costs and register those of non-discarded instances.
6378 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6379 (vect_schedule_slp_instance): Remove copying of
6380 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6382 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6383 adding cost. Record cost entry location.
6384 (vect_prologue_cost_for_slp_op): Function to compute cost of
6385 a constant or invariant generated for SLP vect in the prologue,
6386 split out from vect_analyze_slp_cost_1.
6387 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6388 (vect_model_promotion_demotion_cost): Likewise.
6389 (vect_model_store_cost): Likewise, make static.
6390 (vect_model_load_cost): Likewise.
6391 (vectorizable_bswap): Add cost vector arg and adjust.
6392 (vectorizable_call): Likewise.
6393 (vectorizable_simd_clone_call): Likewise.
6394 (vectorizable_conversion): Likewise.
6395 (vectorizable_assignment): Likewise.
6396 (vectorizable_shift): Likewise.
6397 (vectorizable_operation): Likewise.
6398 (vectorizable_store): Likewise.
6399 (vectorizable_load): Likewise.
6400 (vectorizable_condition): Likewise.
6401 (vectorizable_comparison): Likewise.
6402 (can_vectorize_live_stmts): Likewise.
6403 (vect_analyze_stmt): Likewise.
6404 (vect_transform_stmt): Adjust calls to vectorizable_*.
6405 * tree-vectorizer.c: Include gimple-pretty-print.h.
6406 (dump_stmt_cost): New function.
6408 2018-05-16 Richard Biener <rguenther@suse.de>
6410 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6411 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6412 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6413 (check_name): New callback.
6414 (dse_classify_store): Track cycles via a visited bitmap of PHI
6415 defs and simplify handling of in-loop and across loop dead stores
6416 and properly fail for loop-variant refs. Handle byte-tracking with
6417 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6420 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6422 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6423 (vect_get_mask_type_for_stmt): Likewise.
6424 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6426 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6427 to determine the statement's vector type and the vector type that
6428 should be used for calculating nunits. Deal with cases in which
6429 the type has to be deferred.
6430 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6431 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6432 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6433 (vect_determine_vf_for_stmt): New functions, split out from...
6434 (vect_determine_vectorization_factor): ...here.
6435 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6436 (vect_get_mask_type_for_stmt): New functions, split out from
6437 vect_determine_vectorization_factor.
6439 2018-05-16 Richard Biener <rguenther@suse.de>
6441 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6442 verify the [VEC_]COND_EXPR embedded comparison.
6444 2018-05-15 Martin Sebor <msebor@redhat.com>
6446 PR tree-optimization/85753
6447 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6448 RECORD_TYPE in addition to ARRAY_TYPE.
6450 2018-05-15 Martin Sebor <msebor@redhat.com>
6453 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6455 2018-05-15 Richard Biener <rguenther@suse.de>
6457 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6458 add by_clobber_p one. Change algorithm to collect all defs
6459 representing uses we need to walk and try reducing them to
6460 a single one before failing.
6461 (dse_dom_walker::dse_optimize_stmt): Adjust.
6463 2018-05-13 Mark Wielaard <mark@klomp.org>
6465 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6466 (size_of_loc_descr): Likewise.
6467 (output_loc_operands): Likewise.
6468 (output_loc_operands_raw): Likewise.
6469 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6470 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6471 (hash_loc_operands): Likewise.
6472 (compare_loc_operands): Likewise.
6474 2018-05-14 Mark Wielaard <mark@klomp.org>
6476 * dwarf2out.c (count_index_addrs): New function.
6477 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6479 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6481 PR tree-optimization/83648
6482 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6483 return value as malloc candidate.
6485 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6488 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6489 param as true in call to suggest_attribute.
6491 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6494 -mreadonly-in-sdata.
6496 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6498 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6500 (aarch64_crypto_aesd_fused): Likewise.
6502 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6504 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6505 (movsi_aarch64): Likewise.
6506 (load_pairsi): Likewise.
6507 (load_pairdi): Likewise.
6508 (store_pairsi): Likewise.
6509 (store_pairdi): Likewise.
6510 (load_pairsf): Likewise.
6511 (load_pairdf): Likewise.
6512 (store_pairsf): Likewise.
6513 (store_pairdf): Likewise.
6514 (zero_extend): Likewise.
6515 (trunc): Swap alternatives.
6516 (fcvt_target): Add '?' to prefer w over r.
6518 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6521 * config/i386/i386.md: Disallow non-commutative arithmetics in
6522 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6523 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6524 in the peephole2 before it.
6526 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6528 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6529 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6530 (ix86_handle_option): Handle -mcldemote.
6531 * config.gcc: New header.
6532 * config/i386/cldemoteintrin.h: New file.
6533 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6534 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6537 OPTION_MASK_ISA_CLDEMOTE.
6538 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6539 (ix86_valid_target_attribute_inner_p): Ditto.
6540 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6541 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6542 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6543 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6544 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6546 * config/i386/i386.opt: Add -mcldemote.
6547 * config/i386/x86intrin.h: New header.
6548 * doc/invoke.texi: Add -mcldemote.
6550 2018-05-14 Richard Biener <rguenther@suse.de>
6552 * doc/match-and-simplify.texi: Adjust :s documentation.
6554 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6556 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6557 intended memcpy size.
6558 (REORDER_45): Likewise.
6560 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6562 * sort.cc: New file.
6563 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6564 * vec.c (qsort_chk): Use gcc_qsort.
6565 * Makefile.in (OBJS-libcommon): Add sort.o.
6566 (build/sort.o): New target. Use it...
6567 (BUILD_RTL): ... here, and...
6568 (build/gencfn-macros): ... here, and...
6569 (build/genmatch): ... here.
6571 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6572 Chung-Ju Wu <jasonwucj@gmail.com>
6574 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6575 * config/nds32/nds32-graywolf.md: New file.
6576 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6577 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6579 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6580 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6581 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6582 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6583 * config/nds32/pipelines.md: Include n15 settings.
6585 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6586 Chung-Ju Wu <jasonwucj@gmail.com>
6588 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6589 * config/nds32/nds32-n13.md: New file.
6590 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6591 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6593 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6594 * config/nds32/nds32.md (pipeline_model): Add n13.
6595 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6596 * config/nds32/pipelines.md: Include n13 settings.
6598 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6599 Chung-Ju Wu <jasonwucj@gmail.com>
6601 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6602 * config/nds32/nds32-n10.md: New file.
6603 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6604 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6606 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6607 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6608 * config/nds32/nds32.md (pipeline_model): Add n10.
6609 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6610 * config/nds32/pipelines.md: Include n10 settings.
6612 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6613 Kito Cheng <kito.cheng@gmail.com>
6614 Chung-Ju Wu <jasonwucj@gmail.com>
6616 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6617 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6618 Add enum values for DSP extension instructions.
6619 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6621 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6622 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6624 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6625 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6626 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6627 * config/nds32/nds32-intrinsic.md: Likewise.
6628 * config/nds32/nds32_intrinsic.h: Likewise.
6629 * config/nds32/nds32-md-auxiliary.c: Likewise.
6630 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6631 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6632 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6633 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6634 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6635 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6636 TYPE_DMAC in switch statement.
6637 * config/nds32/nds32.c: New checking and implementation for DSP
6638 extension instructions.
6639 * config/nds32/nds32.h: Likewise.
6640 * config/nds32/nds32.md: Likewise.
6641 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6642 * config/nds32/predicates.md: Implement new predicates for DSP
6645 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6647 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6648 Reformat alternatives and attributes so it is easier to identify
6649 which constraints/attributes go with which instruction.
6650 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6651 (mov<mode>_softfloat32, FMOVE64): Likewise.
6652 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6653 (mov<mode>_softfloat64, FMOVE64): Likewise.
6655 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6657 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6659 (Basic PowerPC Built-in Functions): The new name of the
6660 subsection previously known as "PowerPC Built-in Functions".
6661 (Basic PowerPC Built-in Functions Available on all Configurations):
6663 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6664 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6665 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6666 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6668 2018-05-11 Martin Jambor <mjambor@suse.cz>
6671 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6674 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6677 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6679 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6681 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6682 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6683 (ix86_handle_option): Handle -mwaitpkg.
6684 * config.gcc: New header.
6685 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6686 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6687 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6689 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6690 OPTION_MASK_ISA_WAITPKG.
6691 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6692 (ix86_option_override_internal): Add PTA_WAITPKG.
6693 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6694 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6695 IX86_BUILTIN_TPAUSE.
6696 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6697 __builtin_ia32_umwait and __builtin_ia32_tpause.
6698 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6699 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6700 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6701 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6702 UNSPECV_TPAUSE): New.
6703 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6704 * config/i386/i386.opt: Add -mwaitpkg.
6705 * config/i386/waitpkgintrin.h: New file.
6706 * config/i386/x86intrin.h: New header.
6707 * doc/invoke.texi: Add -mwaitpkg.
6709 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6712 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6714 (cortex-m0): Use armv6s-m isa.
6715 (cortex-m0plus): Likewise.
6716 (cortex-m1): Likewise.
6717 (cortex-m0.small-multiply): Likewise.
6718 (cortex-m0plus.small-multiply): Likewise.
6719 (cortex-m1.small-multiply): Likewise.
6721 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6722 Jakub Jelinek <jakub@redhat.com>
6724 PR tree-optimization/85692
6725 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6726 source permute as well.
6728 2018-05-11 Martin Liska <mliska@suse.cz>
6731 * doc/extend.texi: Document LLVM style format for no_sanitize
6734 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6736 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6737 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6738 (mode_supports_vsx_dform_quad): Likewise.
6739 (mode_supports_vmx_dform): Move these functions to be next to the
6740 other mode_supports functions.
6741 (mode_supports_dq_form): Likewise.
6742 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6743 mode_supports_dq_form.
6744 (reg_offset_addressing_ok_p): Likewise.
6745 (offsettable_ok_by_alignment): Likewise.
6746 (rs6000_legitimate_offset_address_p): Likewise.
6747 (legitimate_lo_sum_address_p): Likewise.
6748 (rs6000_legitimize_address): Likewise.
6749 (rs6000_legitimize_reload_address): Likewise.
6750 (rs6000_secondary_reload_inner): Likewise.
6751 (rs6000_preferred_reload_class): Likewise.
6752 (rs6000_output_move_128bit): Likewise.
6754 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6756 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6757 Generate SImode target register for null target.
6758 <case IX86_BUILTIN_XGETBV>: Ditto.
6759 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6760 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6762 2018-05-10 Carl Love <cel@us.ibm.com>
6764 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6765 dcbtt and dcbtstt if operands[2] is 0.
6767 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6770 * config/i386/sse.md (usadv64qi): New expander.
6772 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6774 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6775 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6776 -maltivec=be support.
6777 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6778 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6779 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6780 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6781 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6782 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6783 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6784 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6785 altivec_vsumsws): Adjust.
6786 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6787 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6788 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6790 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6791 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6792 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6793 (altivec_lve<VI_char>x): Delete expand.
6794 (*altivec_lve<VI_char>x_internal): Rename to...
6795 (altivec_lve<VI_char>x): ... this.
6796 (altivec_lvxl_<mode>): Delete expand.
6797 (*altivec_lvxl_<mode>_internal): Rename to ...
6798 (altivec_lvxl_<mode>): ... this.
6799 (altivec_stvxl_<mode>): Delete expand.
6800 (*altivec_stvxl_<mode>_internal): Rename to ...
6801 (altivec_stvxl_<mode>): ... this.
6802 (altivec_stve<VI_char>x): Delete expand.
6803 (*altivec_stve<VI_char>x_internal): Rename to ...
6804 (altivec_stve<VI_char>x): ... this.
6805 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6806 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6807 reduc_plus_scal_<mode>): Adjust.
6808 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6810 (rs6000_cpu_cpp_builtins): Adjust.
6811 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6812 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6813 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6814 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6815 -maltivec=be support.
6816 (rs6000_split_vec_extract_var): Adjust.
6817 (rs6000_split_v4si_init): Adjust.
6818 (swap_selector_for_mode): Delete.
6819 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6820 altivec_expand_stvex_be): Delete.
6821 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6822 -maltivec=be support.
6823 (rs6000_gimple_fold_builtin): Ditto.
6824 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6826 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6827 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6828 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6829 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6830 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6831 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6832 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6833 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6834 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6835 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6836 anonymous split): Adjust.
6837 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6838 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6840 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6842 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6843 when --with-gxx-include-dir is also specified.
6844 * configure: Regenerate.
6846 2018-05-09 Jim Wilson <jimw@sifive.com>
6849 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6850 * config/riscv/t-withmultilib: New.
6851 * config/riscv/withmultilib.h: New.
6852 * doc/install.texi: Document RISC-V --with-multilib-list support.
6854 2018-05-09 Richard Biener <rguenther@suse.de>
6856 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6858 (vect_bb_vectorization_profitable_p): Adjust. Compute
6859 actual scalar cost using the cost vector and the add_stmt_cost
6862 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6864 PR rtl-optimization/85645
6865 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6866 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6868 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6870 PR rtl-optimization/85645
6871 * shrink-wrap.c (spread_components): Return a boolean saying if
6872 anything was changed.
6873 (try_shrink_wrapping_separate): Iterate spread_components until
6874 nothing changes anymore.
6876 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6878 PR rtl-optimization/85645
6879 * regrename.c (build_def_use): Also kill the chains that include the
6880 destination of a REG_CFA_REGISTER note.
6882 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6884 PR rtl-optimization/85645
6885 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6886 insn that has a REG_CFA_REGISTER note.
6888 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6890 * cfgexpand.c (expand_clobber): New function.
6891 (expand_gimple_stmt_1): Use it.
6892 * tree-vect-stmts.c (vect_clobber_variable): New function,
6894 (vectorizable_simd_clone_call): ...here.
6895 (vectorizable_store): Emit a clobber either side of an
6896 IFN_STORE_LANES sequence.
6897 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6899 2018-05-09 Tom de Vries <tom@codesourcery.com>
6902 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6903 (define_insn "trap_if_false"): Add exit after trap.
6905 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6907 PR rtl-optimization/85638
6908 * bb-reorder.c: Include common/common-target.h.
6909 (create_forwarder_block): New function extracted from...
6910 (fix_up_crossing_landing_pad): ...here. Rename into...
6911 (dw2_fix_up_crossing_landing_pad): ...this.
6912 (sjlj_fix_up_crossing_landing_pad): New function.
6913 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6914 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6915 from both partitions and exit the loop after one iteration.
6917 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6920 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6922 (Basic PowerPC Built-in Functions): The new name of the
6923 subsection previously known as "PowerPC Built-in Functions".
6924 (Basic PowerPC Built-in Functions Available on all Configurations):
6926 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6928 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6930 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6932 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6935 2018-05-08 Jim Wilson <jimw@sifive.com>
6937 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6938 (LD_EMUL_SUFFIX): New.
6939 (LINK_SPEC): Use it.
6941 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6943 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6945 (Basic PowerPC Built-in Functions): The new name of the
6946 subsection previously known as "PowerPC Built-in Functions".
6947 (Basic PowerPC Built-in Functions Available on all Configurations):
6949 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6951 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6953 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6955 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6958 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6961 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6962 after cmpelim optimization.
6964 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6966 * config.gcc: Support "goldmont".
6967 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6968 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6970 * config/i386/i386.c (m_GOLDMONT): Define.
6971 (processor_target_table): Add "goldmont".
6972 (PTA_GOLDMONT): Define.
6973 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6974 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6975 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6976 (fold_builtin_cpu): Add "goldmont".
6977 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6978 (ix86_option_override_internal): Add "goldmont".
6979 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6980 (processor_type): Add PROCESSOR_GOLDMONT.
6981 * config/i386/i386.md: Add CPU "glm".
6982 * config/i386/glm.md: New file.
6983 * config/i386/x86-tune.def: Add m_GOLDMONT.
6984 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6986 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6989 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6991 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6992 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6993 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6996 * config/i386/i386.c (ix86_fold_builtin): Handle
6997 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7000 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7001 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7003 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7006 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7007 (check_arch): Likewise.
7008 (check_fpu): Return the result rather than printing it.
7009 (end arch): Fix operator precedence.
7010 (end cpu): Likewise.
7011 (END): Print the result from check_fpu.
7013 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7014 Alan Hayward <alan.hayward@arm.com>
7015 David Sherwood <david.sherwood@arm.com>
7017 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7018 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7019 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7020 (*fcmuo<mode>_and): New patterns.
7022 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7024 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7025 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7026 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7027 (cmp_op, sve_imm_con): New code attributes.
7028 (SVE_COND_INT_CMP, imm_con): Delete.
7029 (cmp_op): Remove above unspecs from int attribute.
7030 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7032 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7033 comparison-specific unspecs.
7034 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7035 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7036 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7037 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7038 (*vec_fcm<cmp_op><mode>): Rename to...
7039 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7040 (*vec_fcmuo<mode>): Rename to...
7041 (*fcmuo<mode>): ...this and adjust likewise.
7042 (*pred_fcm<cmp_op><mode>): New pattern.
7043 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7044 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7046 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7048 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7049 (aarch64_emit_sve_predicated_cond): New function.
7050 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7051 (aarch64_emit_unspec_cond_or): Replace with...
7052 (aarch64_emit_sve_or_conds): ...this new function. Use
7053 aarch64_emit_sve_ptrue_op for the individual comparisons and
7054 aarch64_emit_binop to OR them together.
7055 (aarch64_emit_inverted_unspec_cond): Replace with...
7056 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7057 aarch64_emit_sve_ptrue_op for the comparison and
7058 aarch64_emit_unop to invert the result.
7059 (aarch64_expand_sve_vec_cmp_float): Update after the above
7060 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7062 2018-05-07 Nathan Sidwell <nathan@acm.org>
7064 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7065 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7066 (Backwards Compatibility): Likewise.
7068 2018-05-07 Luis Machado <luis.machado@linaro.org>
7072 2018-05-07 Luis Machado <luis.machado@linaro.org>
7074 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7075 <prefetch_dynamic_strides>: New const bool field.
7076 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7077 prefetch_dynamic_strides.
7078 (exynosm1_prefetch_tune): Likewise.
7079 (thunderxt88_prefetch_tune): Likewise.
7080 (thunderx_prefetch_tune): Likewise.
7081 (thunderx2t99_prefetch_tune): Likewise.
7082 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7084 (aarch64_override_options_internal): Update to set
7085 PARAM_PREFETCH_DYNAMIC_STRIDES.
7086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7090 prefetch-dynamic-strides setting.
7092 2018-05-07 Luis Machado <luis.machado@linaro.org>
7094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7095 <minimum_stride>: New const int field.
7096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7097 minimum_stride field.
7098 (exynosm1_prefetch_tune): Likewise.
7099 (thunderxt88_prefetch_tune): Likewise.
7100 (thunderx_prefetch_tune): Likewise.
7101 (thunderx2t99_prefetch_tune): Likewise.
7102 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7103 (aarch64_override_options_internal): Update to set
7104 PARAM_PREFETCH_MINIMUM_STRIDE.
7105 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7106 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7107 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7108 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7109 stride is constant and is below the minimum stride threshold.
7111 2018-05-07 Luis Machado <luis.machado@linaro.org>
7113 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7116 2018-05-07 Luis Machado <luis.machado@linaro.org>
7118 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7119 <prefetch_dynamic_strides>: New const bool field.
7120 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7121 prefetch_dynamic_strides.
7122 (exynosm1_prefetch_tune): Likewise.
7123 (thunderxt88_prefetch_tune): Likewise.
7124 (thunderx_prefetch_tune): Likewise.
7125 (thunderx2t99_prefetch_tune): Likewise.
7126 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7128 (aarch64_override_options_internal): Update to set
7129 PARAM_PREFETCH_DYNAMIC_STRIDES.
7130 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7131 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7132 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7133 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7134 prefetch-dynamic-strides setting.
7136 2018-05-07 Luis Machado <luis.machado@linaro.org>
7138 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7139 <minimum_stride>: New const int field.
7140 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7141 minimum_stride field.
7142 (exynosm1_prefetch_tune): Likewise.
7143 (thunderxt88_prefetch_tune): Likewise.
7144 (thunderx_prefetch_tune): Likewise.
7145 (thunderx2t99_prefetch_tune): Likewise.
7146 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7147 (aarch64_override_options_internal): Update to set
7148 PARAM_PREFETCH_MINIMUM_STRIDE.
7149 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7150 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7151 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7152 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7153 stride is constant and is below the minimum stride threshold.
7155 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7158 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7159 the type is addressable. Don't force op into register if it has
7162 2018-05-05 Roland McGrath <mcgrathr@google.com>
7165 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7166 any section for which we don't know a specific type it should have,
7167 regardless of name. Previously this was done only for the exact
7168 names ".init_array", ".fini_array", and ".preinit_array".
7169 (default_elf_asm_named_section): Add comment about
7170 relationship with default_section_type_flags and SECTION_NOTYPE.
7171 (get_section): Don't consider it a type conflict if one side has
7172 SECTION_NOTYPE and the other doesn't, as long as neither has the
7173 SECTION_BSS et al used in the default_section_type_flags logic.
7175 2018-05-05 Tom de Vries <tom@codesourcery.com>
7178 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7179 (workaround_barsyncs): New function.
7180 (nvptx_reorg): Use workaround_barsyncs.
7181 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7182 (define_expand "nvptx_membar_cta"): New define_expand.
7183 (define_insn "*nvptx_membar_cta"): New insn.
7185 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7187 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7188 To improve optimization opportunities.
7189 * builtin-types.def: The new needed builtin types for the above.
7191 2018-05-04 Richard Biener <rguenther@suse.de>
7193 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7194 * gimple-ssa-store-merging.c
7195 (imm_store_chain_info::output_merged_store): Remove redundant create,
7196 release split_store vector contents on failure.
7197 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7198 scalar stmt vector on cache hit.
7200 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7202 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7204 * config.gcc (powerpc-xilinx-eabi*): Remove.
7205 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7207 (fusion_addis_mem_combo_load): Ditto.
7208 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7210 (rs6000_cpu_cpp_builtins): Ditto.
7211 * config/rs6000/rs6000-linux.c
7212 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7213 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7214 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7216 (rs6000_setup_reg_addr_masks): Ditto.
7217 (rs6000_init_hard_regno_mode_ok): Ditto.
7218 (rs6000_option_override_internal): Ditto.
7219 (legitimate_lo_sum_address_p): Ditto.
7220 (rs6000_legitimize_address): Ditto.
7221 (rs6000_legitimize_reload_address): Ditto.
7222 (rs6000_legitimate_address_p): Ditto.
7223 (abi_v4_pass_in_fpr): Ditto.
7224 (setup_incoming_varargs): Ditto.
7225 (rs6000_gimplify_va_arg): Ditto.
7226 (rs6000_split_multireg_move): Ditto.
7227 (rs6000_savres_strategy): Ditto.
7228 (rs6000_emit_prologue_components): Ditto.
7229 (rs6000_emit_epilogue_components): Ditto.
7230 (rs6000_emit_prologue): Ditto.
7231 (rs6000_emit_epilogue): Ditto.
7232 (rs6000_elf_file_end): Ditto.
7233 (rs6000_function_value): Ditto.
7234 (rs6000_libcall_value): Ditto.
7235 * config/rs6000/rs6000.h: Ditto.
7236 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7237 (TARGET_MINMAX): ... this. New.
7238 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7239 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7240 (*movsi_internal1_single): Delete.
7241 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7242 mfpu=, mxilinx-fpu): Delete.
7243 * config/rs6000/singlefp.h: Delete.
7244 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7245 * config/rs6000/t-rs6000: Ditto.
7246 * config/rs6000/t-xilinx: Delete.
7247 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7248 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7249 (VStype_simple): Delete.
7250 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7251 * config/rs6000/xfpu.h: Delete.
7252 * config/rs6000/xfpu.md: Delete.
7253 * config/rs6000/xilinx.h: Delete.
7254 * config/rs6000/xilinx.opt: Delete.
7255 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7256 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7258 2018-05-04 Tom de Vries <tom@codesourcery.com>
7261 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7264 2018-05-04 Richard Biener <rguenther@suse.de>
7267 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7268 (expand_complex_div_wide): Likewise.
7269 (expand_complex_operations_1): Likewise.
7270 (expand_complex_libcall): Preserve EH info of the original stmt.
7271 (tree_lower_complex): Handle removed blocks.
7272 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7273 on complex multiplication and division libcall builtins.
7275 2018-05-04 Richard Biener <rguenther@suse.de>
7278 * fold-const.c (negate_expr_p): Restrict negation of operand
7279 zero of a division to when we know that can happen without
7281 (fold_negate_expr_1): Likewise.
7283 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7286 * real.h (real_nextafter): Declare.
7287 * real.c (real_nextafter): New function.
7288 * fold-const-call.c (fold_const_nextafter): New function.
7289 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7290 CASE_CFN_NEXTTOWARD.
7291 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7292 even when arg1_mode is different from arg0_mode.
7294 2018-05-03 Nathan Sidwell <nathan@acm.org>
7296 * doc/extend.texi (Deprecated Features): Remove
7298 (Backwards Compatibility): Likewise.
7299 * doc/invoke.texi (C++ Language Options): Likewise.
7300 (C++ Dialect Options): Likewise.
7302 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7305 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7306 _mm512_mask_mullox_epi64): New intrinsics.
7308 2018-05-03 Tom de Vries <tom@codesourcery.com>
7311 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7312 dump files): Add offload-tree.
7314 2018-05-03 Richard Biener <rguenther@suse.de>
7316 PR tree-optimization/85615
7317 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7318 to loops not nested in BBs loop father to avoid creating multi-entry
7321 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7323 PR tree-optimization/70291
7324 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7325 arguments. Change return type to tree. Emit libcall as a new
7326 statement rather than replacing existing one when inplace_p is true.
7327 (expand_complex_multiplication_components): New function.
7328 (expand_complex_multiplication): Expand floating-point complex
7329 multiplication using the above.
7330 (expand_complex_division): Rename inner_type parameter to type.
7331 Update expand_complex_libcall call-site.
7332 (expand_complex_operations_1): Update expand_complex_multiplication
7333 and expand_complex_division call-sites.
7335 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7338 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7339 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7340 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7341 the highest significant bit of the shift count mask is clear. In
7342 check whether and[sq]i3 is needed verify that all significant bits
7343 of the shift count other than the highest are set.
7345 2018-05-02 Tom de Vries <tom@codesourcery.com>
7348 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7349 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7350 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7351 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7352 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7353 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7354 __builtin_goacc_parlevel_size.
7356 2018-05-02 Richard Biener <rguenther@suse.de>
7358 PR tree-optimization/85597
7359 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7360 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7363 2018-05-02 Tom de Vries <tom@codesourcery.com>
7366 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7367 dump files): Add ltrans-tree.
7369 2018-05-02 Tom de Vries <tom@codesourcery.com>
7372 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7373 dump files): Add wpa-ipa.
7375 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7377 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7378 powerpc*-*-linux*paired* target.
7379 * config/rs6000/750cl.h: Delete.
7380 * config/rs6000/paired.h: Delete.
7381 * config/rs6000/paired.md: Delete.
7382 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7384 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7385 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7386 comment. Remove paired float support.
7387 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7388 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7390 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7391 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7393 * config/rs6000/rs6000.c: Remove paired float support.
7394 (paired_expand_vector_init, paired_expand_vector_move,
7395 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7396 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7397 paired_expand_builtin, paired_expand_predicate_builtin,
7398 paired_init_builtins): Delete.
7399 * config/rs6000/rs6000.h: Remove paired float support.
7400 * config/rs6000/rs6000.md: Remove paired float support.
7401 (move_from_CR_ov_bit): Delete.
7402 * config/rs6000/rs6000.opt (mpaired): Delete.
7403 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7406 2018-05-02 Richard Biener <rguenther@suse.de>
7409 * gimplify.c (gimplify_save_expr): When in SSA form allow
7410 SAVE_EXPRs to compute to SSA vars.
7412 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7415 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7416 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7417 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7418 clobber operands[2], instead use a new pseudo. Formatting fixes.
7420 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7422 PR tree-optimization/85586
7423 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7424 exit early for statements in the same group if the accesses are
7427 2018-05-02 Tom de Vries <tom@codesourcery.com>
7430 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7433 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7435 PR tree-optimization/85143
7436 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7438 2018-05-01 Tom de Vries <tom@codesourcery.com>
7441 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7442 not found" error message.
7444 2018-05-01 Tom de Vries <tom@codesourcery.com>
7447 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7448 * vec.c (test_ordered_remove_if): New function.
7449 (vec_c_tests): Call test_ordered_remove_if.
7450 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7451 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7452 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7453 VEC_ORDERED_REMOVE_IF.
7455 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7457 PR tree-optimization/82665
7458 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7459 pointer subtraction where arguments come from a memchr call.
7461 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7463 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7464 --push-state --as-needed and --pop-state instead of --as-needed and
7465 --no-as-needed if ld supports it.
7466 * configure: Regenerated.
7469 * doc/install.texi2html: Replace _002d with - and _002a with * in
7470 generated html files using sed.
7472 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7475 * gcc-rich-location.c (blank_line_before_p): New function.
7476 (use_new_line): New function.
7477 (gcc_rich_location::add_fixit_insert_formatted): New function.
7478 * gcc-rich-location.h
7479 (gcc_rich_location::add_fixit_insert_formatted): New function.
7481 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7483 * selftest.c (assert_streq): Rename "expected" and "actual" to
7484 "val1" and "val2". Extend NULL-handling to cover both inputs
7485 symmetrically, while still requiring both to be non-NULL for a pass.
7486 * selftest.h (assert_streq): Rename "expected" and "actual" to
7488 (ASSERT_EQ): Likewise.
7489 (ASSERT_EQ_AT): Likewise.
7490 (ASSERT_KNOWN_EQ): Likewise.
7491 (ASSERT_KNOWN_EQ_AT): Likewise.
7492 (ASSERT_NE): Likewise.
7493 (ASSERT_MAYBE_NE): Likewise.
7494 (ASSERT_MAYBE_NE_AT): Likewise.
7495 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7496 the assertion to pass.
7497 (ASSERT_STREQ_AT): Likewise.
7499 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7501 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7502 interaction with -pie.
7504 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7506 * selftest.h: Fix alphabetization of per-source-file selftest
7509 2018-04-30 Jason Merrill <jason@redhat.com>
7511 PR c++/61982 - dead stores to destroyed objects.
7512 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7515 2018-04-30 Jason Merrill <jason@redhat.com>
7517 * tree.c (build_clobber): New.
7518 * tree.h: Declare it.
7519 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7521 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7523 * diagnostic-show-locus.c (layout::layout): Update for
7524 location_get_source_line returning a char_span.
7525 (struct char_span): Move to input.h.
7526 (struct correction): Update for fields in char_span becoming
7528 (struct source_line): Update for location_get_source_line
7529 returning a char_span.
7530 (layout::print_line): Likewise.
7531 * edit-context.c (edited_file::print_content): Likewise.
7532 (edited_file::print_diff_hunk): Likewise.
7533 (edited_file::print_run_of_changed_lines): Likewise.
7534 (edited_file::get_num_lines): Likewise.
7535 (edited_line::edited_line): Likewise.
7536 * final.c (asm_show_source): Likewise.
7537 * input.c (location_get_source_line): Convert return type
7538 from const char * to char_span, losing the final "line_len"
7540 (dump_location_info): Update for the above.
7541 (get_substring_ranges_for_loc): Likewise. Use a char_span
7542 when handling the literal within the line.
7543 (test_reading_source_line): Update for location_get_source_line
7544 returning a char_span.
7545 * input.h (class char_span): Move here from
7546 diagnostic-show-locus.c, converting from a struct to a class.
7547 Make data members private.
7548 (char_span::operator bool): New.
7549 (char_span::length): New.
7550 (char_span::get_buffer): New.
7551 (char_span::operator[]): New.
7552 (char_span::subspan): Make const.
7553 (char_span::xstrdup): New.
7554 (location_get_source_line): Convert return type from const char *
7555 to char_span, losing the final "line_size" param.
7557 2018-04-30 Jan Hubicka <jh@suse.cz>
7559 * lto-wrapper.c (ltrans_priorities): New static var.
7560 (cmp_priority): New.
7561 (run_gcc): Read priorities and if doing parallel build order
7562 the Makefile by them.
7564 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7566 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7568 2018-04-30 Richard Biener <rguenther@suse.de>
7570 * tree-cfg.c (verify_address): Remove base argument, add
7571 flag whether to check TREE_ADDRESSABLE and do that.
7572 (verify_expr): Remove.
7573 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7574 (verify_gimple_assign_single): Likewise.
7575 (verify_gimple_switch): Likewise.
7576 (verify_expr_location_1): Dereference tp once. Add (disabled)
7577 piece from verify_expr.
7578 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7580 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7582 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7584 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7586 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7587 (small_data_pattern): Likewise.
7588 (arc_rewrite_small_data): Likewise.
7589 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7590 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7591 (get_symbol_alignment): New function.
7592 (legitimate_small_data_address_p): Likewise.
7593 (legitimate_scaled_address): Update, call
7594 legitimate_small_data_address_p.
7595 (output_sdata): New static variable.
7596 (arc_print_operand): Update how we handle small data operands.
7597 (arc_print_operand_address): Likewise.
7598 (arc_legitimate_address_p): Update, use
7599 legitimate_small_data_address_p.
7600 (arc_rewrite_small_data_p): Remove.
7601 (arc_rewrite_small_data_1): Likewise.
7602 (arc_rewrite_small_data): Likewise.
7603 (small_data_pattern): Likewise.
7604 (compact_sda_memory_operand): Update to use
7605 legitimate_small_data_address_p and get_symbol_alignment.
7606 (prepare_move_operands): Don't rewite sdata pattern.
7607 (prepare_extend_operands): Remove.
7608 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7610 (zero_extendqisi2): Likewise.
7611 (zero_extendhisi2): Likewise.
7612 (extendqihi2): Likewise.
7613 (extendqisi2): Likewise.
7614 (extendhisi2): Likewise.
7618 * config/arc/constraints.md (Usd): Change it to memory constraint.
7620 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7622 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7623 as source of std instructions.
7624 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7625 allow 6-bit constants as source for store instructions.
7626 (movdi_insn): Update instruction pattern to allow 6-bit constants
7627 as source for store instructions.
7629 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7631 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7633 2018-04-30 Nathan Sidwell <nathan@acm.org>
7634 Sandra Loosemore <sandra@codesourcery.com>
7636 * dumpfile.c (dump_open): Allow '-' for stdout.
7637 * doc/invoke.texi (Developer Options): Document dump filename
7638 determination early. Document stdin/stdout selection.
7640 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7642 Microblaze Target: PIC data text relative
7644 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7645 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7647 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7648 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7649 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7650 New addressing mode for data-text relative position indepenedent code.
7651 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7652 'ADDRESS_SYMBOLIC_TXT_REL'.
7653 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7654 (microblaze_legitimate_pic_operand): Exclude function calls from
7655 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7656 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7658 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7659 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7660 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7661 for 'address + offset'.
7662 (microblaze_expand_prologue): Add new function prologue call for
7664 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7665 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7666 table in case of TARGET_PIC_DATA_TEXT_REL.
7667 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7668 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7669 Add new macros 'UNSPEC_TEXT',
7670 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7671 + exclude function calls from 'UNSPEC_PLT' in case of data text
7673 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7674 new target hook for generating address diff vector tables in case of
7676 * doc/tm.texi : Regenerate.
7677 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7678 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7679 of addr diff vector generation.
7680 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7681 target hook definition.
7682 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7683 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7684 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7685 option for data text relative.
7687 2018-04-30 Richard Biener <rguenther@suse.de>
7689 * tree-chrec.h (evolution_function_is_constant_p): Remove
7691 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7693 2018-04-30 Richard Biener <rguenther@suse.de>
7696 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7698 2018-04-30 Richard Biener <rguenther@suse.de>
7700 PR tree-optimization/28364
7701 PR tree-optimization/85275
7702 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7703 copying first exit test.
7705 2018-04-28 Mark Wielaard <mark@klomp.org>
7707 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7709 (dwarf_AT): Handle DW_AT_addr_base.
7710 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7712 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7715 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7716 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7717 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7718 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7720 2018-04-28 Richard Biener <rguenther@suse.de>
7722 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7723 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7724 to reflect use. Only add interesting stmts.
7726 2018-04-27 Martin Jambor <mjambor@suse.cz>
7729 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7730 the jump function allows for passing through aggregate values.
7732 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7734 * input.h (in_system_header_at): Convert from macro to inline
7736 (from_macro_expansion_at): Likewise.
7737 (from_macro_definition_at): Likewise.
7739 2018-04-27 Jeff Law <law@redhat.com>
7741 * config.gcc: Mark tile* targets as deprecated/obsolete.
7743 2018-04-27 Richard Biener <rguenther@suse.de>
7745 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7748 2018-04-27 Richard Biener <rguenther@suse.de>
7750 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7752 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7754 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7755 with Yd constraint. Set "preferred_for_speed" attribute from
7756 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7758 (*movdi_internal): Ditto.
7759 (movti_interunit splitters): Remove
7760 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7761 (movdi_interunit splitters): Ditto.
7762 * config/i386/constraints.md (Ye): Remove.
7763 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7765 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7768 * config/aarch64/constraints.md (Usg): Limit to 31.
7771 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7773 PR tree-optimization/85529
7774 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7775 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7776 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7777 zero extension or masking of the MSB bit.
7778 (optimize_range_tests): Add FIRST_BB argument, pass it through
7779 to optimize_range_tests_var_bound.
7780 (maybe_optimize_range_tests, reassociate_bb): Adjust
7781 optimize_range_tests callers.
7783 2018-04-26 Richard Biener <rguenther@suse.de>
7784 Jakub Jelinek <jakub@redhat.com>
7786 * cgraph.h (symbol_table): Just declare debug method here.
7787 * symtab.c (symbol_table::debug): Define.
7789 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7791 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7793 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7795 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7796 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7797 (*movdi_internal): Substitute Yi and Yj constraint with x
7798 and Ym and Yn constraint with y constraint. Update "isa"
7799 attribute and set "preferred_for_speed" attribute from
7800 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7801 (*movsi_internal): Ditto.
7802 (*movdf_internal): Ditto.
7803 (*movsf_internal): Ditto.
7804 (*zero_extendsidi2): Ditto.
7805 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7806 (sse2_loadld): Ditto.
7807 (*vec_extract<ssevecmodelower>_0): Ditto.
7808 (*vec_extractv4si_0_zext_sse4): Ditto.
7809 (vec_concatv2di): Ditto.
7810 (*vec_dup<mode>): Ditto.
7811 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7812 * config/i386/constraints.md (Yi): Remove.
7817 2018-04-26 Nathan Sidwell <nathan@acm.org>
7819 * dumpfile.c (dump_open): New.
7820 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7821 (dump_finish): Detect stdio/stderr by value not name.
7823 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7825 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7827 2018-04-26 Tom de Vries <tom@codesourcery.com>
7830 * config/nvptx/nvptx.c (verify_neutering_jumps)
7831 (verify_neutering_labels): New function
7832 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7834 2018-04-26 Tom de Vries <tom@codesourcery.com>
7837 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7838 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7841 2018-04-26 Richard Biener <rguenther@suse.de>
7842 Tom de Vries <tom@codesourcery.com>
7845 * lto-streamer-out.c (output_function): Fixup loops if required to match
7846 discovery done in the reader.
7848 2018-04-26 Richard Biener <rguenther@suse.de>
7850 PR tree-optimization/85116
7851 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7852 have a loop exit from the single latch predecessor. Remove
7853 case of header with just condition.
7854 (ch_base::copy_headers): Exclude infinite loops from any
7856 (pass_ch::execute): Record exits.
7858 2018-04-26 Richard Biener <rguenther@suse.de>
7860 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7861 prologue cost vector and pass it to vect_get_load_cost.
7862 (vect_get_peeling_costs_all_drs): Likewise.
7863 (vect_peeling_hash_get_lowest_cost): Likewise.
7864 (vect_enhance_data_refs_alignment): Likewise.
7866 2018-04-26 Richard Biener <rguenther@suse.de>
7869 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7870 checking of integer<->pointer conversions.
7871 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7872 sign-/zero-extending pointer types.
7873 (expand_omp_for_static_chunk): Likewise.
7875 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7876 Jean Lee <xiaoyur347@gmail.com>
7878 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7879 (TARGET_ASAN_SHADOW_OFFSET): Define.
7880 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7881 true for -fsanitize=address.
7883 2018-04-25 Mark Wielaard <mark@klomp.org>
7885 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7888 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7890 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7891 than "alu", remove explicit "memory" and "imm_disp" attributes.
7892 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7895 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7896 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7897 gen_lowpart_no_emit.
7899 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7902 * config/i386/i386.c (ix86_expand_builtin): Change memory
7903 operand to XI, extend p0 to Pmode.
7904 * config/i386/i386.md: Change unspec volatile and operand
7905 1 mode to XI, change operand 0 mode to P.
7907 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7909 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7910 GET_MODE_MASK before any checking.
7911 (nds32_can_use_bset_p): Likewise.
7912 (nds32_can_use_btgl_p): Likewise.
7914 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7916 * config/nds32/nds32-doubleword.md: New define_split pattern for
7917 illegal register number.
7919 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7921 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7923 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7925 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7927 2018-04-25 Richard Biener <rguenther@suse.de>
7929 * lto-streamer.h (LTO_major_version): Bump to 8.
7931 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7933 * BASE-VER: Set to 9.0.0.
7935 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7937 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7938 in __abskf2 and __powikf2.
7940 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7943 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7944 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7945 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7946 Use the above on operand 2. Reindent.
7947 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7949 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7952 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7954 * config/i386/i386.opt (mcet): Removed.
7955 * doc/install.texi: Remove -mcet documentation.
7956 * doc/invoke.texi: Likewise.
7958 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7961 * doc/install.texi: Remove -mcet from bootstrap-cet.
7963 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7966 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7967 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7971 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7972 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7973 containing a CONST_VECTOR.
7975 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7977 * doc/install.texi: Update newlib dependency for nvptx.
7979 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7982 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7983 instead of INTVAL when shifting x left.
7985 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7987 PR tree-optimization/85478
7988 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7989 vect_grouped_store_supported for single element vectors.
7991 2018-04-24 Richard Biener <rguenther@suse.de>
7994 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7995 load cost increase to the case of non-constant step.
7997 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8000 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8001 destination if any_malformed_asm.
8003 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8006 * expr.c (store_field): In the bitfield case, if the value comes from
8007 a function call and is returned in registers by means of a PARALLEL,
8008 do not change the mode of the temporary unless BLKmode and VOIDmode.
8010 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8012 PR rtl-optimization/85423
8013 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8014 dependencies to debug insns when the previous insn is non-debug.
8016 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8018 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8019 enums into a single definition.
8020 (fls): Fix predicates and printing.
8023 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8025 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8026 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8027 and short u6 immediate.
8028 (check_if_valid_sleep_operand): Remove.
8029 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8031 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8033 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8034 flag_always_save_lp condition.
8035 * config/nds32/nds32.opt (malways-save-lp): New option.
8037 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8039 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8040 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8041 * config/nds32/nds32.h
8042 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8043 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8045 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8047 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8048 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8050 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8051 Chung-Ju Wu <jasonwucj@gmail.com>
8053 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8055 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8056 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8058 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8060 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8062 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8064 * config/nds32/nds32-protos.h (nds32_data_alignment,
8065 nds32_local_alignment): Declare.
8066 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8067 nds32_local_alignment): New functions.
8068 (TARGET_CONSTANT_ALIGNMENT): Define.
8069 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8071 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8073 * config/nds32/nds32.c
8074 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8075 (TARGET_MODES_TIEABLE_P): Likewise.
8077 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8079 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8082 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8083 Chung-Ju Wu <jasonwucj@gmail.com>
8085 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8086 for unaligned access.
8087 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8089 * config/nds32/nds32-intrinsic.md: Likewise.
8090 * config/nds32/nds32_intrinsic.h: Likewise.
8091 * config/nds32/nds32.h (nds32_builtins): Likewise.
8092 * config/nds32/nds32.opt (munaligned-access): New option.
8093 * config/nds32/nds32.c (nds32_asm_file_start): Display
8094 flag_unaligned_access status.
8096 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8098 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8099 -mno-relax is present.
8100 * config/riscv/linux.h (LINK_SPEC): Ditto.
8102 2018-04-20 Martin Sebor <msebor@redhat.com>
8105 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8107 (gimple_fold_builtin_stxcpy_chk): Same.
8108 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8110 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8113 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8114 __powikf2 when long double is IEEE 128-bit.
8116 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8118 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8119 step to make sure stack always aligned.
8121 2018-04-20 Carl Love <cel@us.ibm.com>
8124 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8125 size check for arg0.
8127 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8128 Tom de Vries <tom@codesourcery.com>
8131 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8132 Emit insns for calls too.
8133 (nvptx_find_par): Always look for worker-level predecessor insn.
8134 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8136 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8137 (nvptx_process_pars): Propagate frames for calls.
8139 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8142 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8144 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8145 (ix86_handle_option): Don't handle OPT_mibt.
8146 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8148 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8150 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8151 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8152 (ix86_target_macros): Define __CET__ with flag_cf_protection
8153 for -fcf-protection.
8154 * config/i386/i386.c (isa2_opts): Remove -mibt.
8155 * config/i386/i386.h (TARGET_IBT): Removed.
8156 (TARGET_IBT_P): Likewise.
8157 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8158 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8159 * config/i386/i386.opt (mcet): Update help message.
8162 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8163 -mcet as an alias for -mshstk.
8165 2018-04-20 Richard Biener <rguenther@suse.de>
8168 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8169 complexity by forcing a single use of the multiply operand.
8171 2018-04-20 Martin Jambor <mjambor@suse.cz>
8174 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8175 recursion dependency to only apply to non-clones.
8177 2018-04-20 Martin Jambor <mjambor@suse.cz>
8180 * ipa-cp.c (create_specialized_node): Check that clones of
8181 self-recursive edges exist during IPA-CP.
8183 2018-04-19 Toon Moene <toon@moene.org>
8185 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8188 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8190 PR tree-optimization/85467
8191 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8192 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8193 VECTOR_CST element to type.
8195 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8198 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8199 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8200 (builtin_longjmp): Likewise.
8201 (save_stack_nonlocal): New pattern.
8202 (restore_stack_nonlocal): Likewise.
8204 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8207 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8208 Replace ASM_OUTPUT_LABEL with fprintf.
8210 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8213 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8214 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8215 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8216 define __IBT__ and __SHSTK__ for -fcf-protection.
8217 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8219 (ix86_trampoline_init): Likewise.
8220 (x86_output_mi_thunk): Likewise.
8221 (ix86_notrack_prefixed_insn_p): Likewise.
8222 (ix86_option_override_internal): Don't disallow -fcf-protection.
8223 * config/i386/i386.md (rdssp<mode>): Also enable for
8225 (incssp<mode>): Likewise.
8226 (nop_endbr): Likewise.
8227 * config/i386/i386.opt (mcet): Change help message to built-in
8231 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8232 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8233 enable CET built-in functions.
8235 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8237 * common/config/i386/i386-common.c
8238 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8239 OPTION_MASK_ISA_MOVDIRI_UNSET,
8240 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8241 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8242 * config.gcc (movdirintrin.h): New header.
8243 * config/i386/cpuid.h (bit_MOVDIRI,
8244 bit_MOVDIR64B): New bits.
8245 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8247 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8248 (VOID, PVOID, PCVOID)): New function types.
8249 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8250 __builtin_ia32_directstoreu_u64,
8251 __builtin_ia32_movdir64b): New builtins.
8252 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8253 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8255 (ix86_valid_target_attribute_inner_p): Ditto.
8256 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8257 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8258 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8259 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8260 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8261 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8262 (movdiri<mode>, movdir64b_<mode>): New.
8263 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8264 * config/i386/immintrin.h: Include movdirintrin.h.
8265 * config/i386/movdirintrin.h: New file.
8266 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8268 2018-04-19 Richard Biener <rguenther@suse.de>
8271 * cfg.c (clear_bb_flags): When loop state says we have
8272 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8274 2018-04-19 Richard Biener <rguenther@suse.de>
8276 PR tree-optimization/84737
8277 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8278 copying restrict info.
8279 (vect_setup_realignment): Use it.
8280 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8281 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8282 the first DR to all generated stores.
8283 (vectorizable_load): Likewise for loads.
8285 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8287 PR tree-optimization/85446
8288 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8289 the integral and pointer types to have the same precision.
8291 * doc/install.texi: Document --disable-cet being the default and
8294 2018-04-18 Martin Liska <mliska@suse.cz>
8296 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8299 2018-04-18 Martin Liska <mliska@suse.cz>
8302 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8305 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8306 arguments if they are comparable.
8308 2018-04-18 Martin Liska <mliska@suse.cz>
8311 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8314 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8317 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8320 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8321 ENDBR after calling __morestack.
8323 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8326 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8327 by using gcc_base_ver to generate a gcc_driver_version, and use
8328 it when generating GCC_DRIVER_NAME.
8329 * configure: Regenerate.
8331 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8334 * config.gcc: Obsolete powerpc*-*-*spe*.
8336 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8339 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8340 (stabstr_D): Change type of unum from unsigned int to
8341 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8344 2018-04-17 Jim Wilson <jimw@sifive.com>
8347 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8348 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8349 Set arg_pointer_offset after using pretend_args_size.
8351 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8353 PR rtl-optimization/85431
8354 * dse.c (record_store): Ignore zero width stores.
8357 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8358 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8359 __builtin_stack_restore rather than after it.
8360 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8361 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8362 argument instead of virtual_dynamic_stack_rtx.
8364 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8366 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8368 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8369 Add note to error message to explain internal mapping of overloaded
8370 built-in function name to non-overloaded built-in function name.
8371 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8374 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8377 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8378 where the inputs overlap with the output.
8380 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8382 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8383 (=v, v) alternative and explicit "memory" attribute.
8384 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8385 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8387 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8388 "sselog1" type instead of "sselog".
8389 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8390 "sselog". Remove explicit "memory" attribute.
8391 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8392 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8394 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8395 "sselog1" type instead of "sselog". Remove explicit "memory"
8397 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8398 use "sselog1" type instead of "sselog". Remove explicit "memory"
8400 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8401 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8403 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8404 "sselog1" type instead of "sselog". Remove explicit "memory"
8406 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8407 use "sselog1" type instead of "sselog". Remove explicit "memory"
8411 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8414 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8417 2018-04-17 Martin Jambor <mjambor@suse.cz>
8420 * ipa-cp.c (create_specialized_node): Call
8421 expand_all_artificial_thunks if necessary.
8423 2018-04-17 Martin Liska <mliska@suse.cz>
8426 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8427 in message, remote space in between '_G' and '('.
8429 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8432 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8433 avx512f_vmcmp<mode>3<round_saeonly_name>,
8434 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8435 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8436 avx512f_rndscale<mode><round_saeonly_name>,
8437 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8438 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8439 Use %<iptr>2 instead of %2 for -masm=intel.
8440 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8441 avx512f_vcvttss2usi<round_saeonly_name>,
8442 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8444 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8445 avx512f_vcvttsd2usi<round_saeonly_name>,
8446 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8447 Use %q1 instead of %1 for -masm=intel.
8448 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8449 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8450 of %3 for -masm=intel.
8451 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8453 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8455 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8457 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8458 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8460 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8462 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8463 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8464 %g1 and one with %0 and %1.
8465 (avx512er_vmrcp28<mode><round_saeonly_name>,
8466 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8468 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8469 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8470 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8471 of %0 and %{%4%} for -masm=intel.
8472 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8473 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8474 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8475 order of %0 and %{%5%}%{z%} for -masm=intel.
8477 2018-04-17 Jan Hubicka <jh@suse.cz>
8480 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8482 2018-04-17 Martin Liska <mliska@suse.cz>
8485 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8486 for target_clone error message. Make default implementation
8487 clone to be a local declaration.
8488 (separate_attrs): Add new argument and check for an empty
8490 (expand_target_clones): Handle it.
8491 (ipa_target_clone): Make redirection just for target_clones
8494 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8495 Tom de Vries <tom@codesourcery.com>
8498 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8499 tiled basic blocks without omp continue statements.
8501 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8504 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8505 vec_extract expression as having side effects to make sure it gets
8508 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8511 * config/i386/i386.c (get_builtin_code_for_version): Check
8514 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8517 * gcc/config.gcc: Support "skylake".
8518 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8520 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8521 (processor_target_table): Add "skylake".
8522 (ix86_option_override_internal): Add "skylake".
8523 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8524 PROCESSOR_CANNONLAKE.
8525 (get_builtin_code_for_version): Fix priority for
8526 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8527 PROCESSOR_SKYLAKE-AVX512.
8528 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8529 (processor_type): Add PROCESSOR_SKYLAKE.
8531 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8532 Jason Merrill <jason@redhat.com>
8535 * convert.c (convert_to_integer_1): Use direct recursion for
8536 enumeral types and types with a precision less than the number
8537 of bits in their mode.
8539 2018-04-16 Julia Koval <julia.koval@intel.com>
8542 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8543 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8545 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8548 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8549 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8550 and -mno-direct-move.
8552 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8555 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8556 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8557 Ensure negative shifts result in {0}.
8559 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8561 PR rtl-optimization/79916
8562 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8563 regs (if any) to define how to gnerate SD moves when LRA is in
8566 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8568 PR rtl-optimization/85393
8569 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8570 * except.c (expand_dw2_landing_pad_for_region): Make static.
8571 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8572 a label and unconditional jump to old_bb, rather than
8573 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8576 PR rtl-optimization/85376
8577 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8578 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8579 instead of a specific value.
8581 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8582 Bin Cheng <bin.cheng@arm.com>
8584 PR tree-optimization/82965
8585 PR tree-optimization/83991
8586 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8587 by_profile_only parameter.
8588 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8589 information if the loop was predicted to iterate too many times.
8590 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8592 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8595 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8598 2018-04-13 Martin Liska <mliska@suse.cz>
8599 Jakub Jelinek <jakub@redhat.com>
8602 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8603 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8604 * builtins.c (expand_builtin_memory_copy_args): Use
8605 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8606 handle dest_addr == pc_rtx.
8608 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8611 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8612 asked to not generate direct moves.
8613 (fix_trunc<mode>si2_stfiwx): Similar.
8614 (fix_trunc<mode>si2_internal): Similar.
8616 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8619 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8620 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8621 lookup if dest in some wider mode is known to be const0_rtx and
8622 if so, record permanent equivalence for it to be ZERO_EXTEND of
8623 the narrower mode destination.
8625 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8627 * lto-streamer-out.c (output_function): Revert 259346.
8628 * omp-expand.c (expand_oacc_for): Likewise.
8630 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8632 PR rtl-optimization/85354
8633 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8634 * sel-sched.c (sel_global_init): ... here.
8636 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8639 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8640 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8641 mode for PE-COFF targets.
8642 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8643 (i386_pe_asm_lto_end): Likewise.
8644 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8645 (TARGET_ASM_LTO_END): Likewise.
8646 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8647 (i386_pe_asm_lto_start): New function.
8648 (i386_pe_asm_lto_end): Likewise.
8650 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8651 Richard Biener <rguenther@suse.de>
8654 * lto-streamer-out.c (output_function): Fix CFG loop state before
8656 * omp-expand.c (expand_oacc_for): Handle calls to internal
8657 functions like regular functions.
8659 2018-04-12 Richard Biener <rguenther@suse.de>
8662 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8663 for the early LTO debug to properly generate references to it
8664 during DIE emission. Do not re-use that for the skeleton for
8666 (dwarf2out_early_finish): Likewise.
8668 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8671 * config/i386/sse.md
8672 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8673 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8674 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8675 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8676 and output is a reg, avoid creating invalid lowpart subreg, but
8677 instead split into a 512-bit move. Don't split if not AVX512VL,
8678 input is xmm16+ reg and output is a mem.
8679 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8680 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8681 xmm16+ reg and output is a mem.
8683 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8685 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8686 also for flag_dwarf2_cfi_asm.
8688 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8690 PR rtl-optimization/85342
8691 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8692 a bool scalar var inside of the loop instead. Don't try to update
8693 recog_data.operand after failed apply_change_group.
8695 2018-04-12 Tom de Vries <tom@codesourcery.com>
8698 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8699 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8700 array with flexible array member as array without given dimension.
8701 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8702 argument for undefined param to true.
8704 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8707 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8708 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8709 from PowerPC section.
8710 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8711 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8712 help text that is too long.
8713 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8714 help text that is too long.
8715 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8716 help text that is too long.
8718 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8720 * config/alpha/alpha.md (stack_probe_internal): Rename
8721 from "probe_stack". Update all callers.
8723 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8725 PR rtl-optimization/84566
8726 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8727 sched_macro_fuse_insns.
8729 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8732 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8733 (compute_block_dependences): ... from here.
8735 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8737 PR tree-optimization/85331
8738 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8739 from int to HOST_WIDE_INT.
8741 2018-04-11 Martin Jambor <mjambor@suse.cz>
8744 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8745 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8746 not the same as the source val.
8747 (cgraph_edge_brings_value_p): New parameter.
8748 (gather_edges_for_value): Pass destination value to
8749 cgraph_edge_brings_value_p.
8750 (perhaps_add_new_callers): Likewise.
8751 (get_info_about_necessary_edges): Likewise and exclude values brought
8752 only by self-recursive edges.
8753 (create_specialized_node): Redirect only clones of self-calling edges.
8754 (+self_recursive_pass_through_p): New function.
8755 (find_more_scalar_values_for_callers_subset): Use it.
8756 (find_aggregate_values_for_callers_subset): Likewise.
8757 (known_aggs_to_agg_replacement_list): Removed.
8758 (decide_whether_version_node): Re-calculate known constants for all
8759 remaining context clones.
8761 2018-04-11 Richard Biener <rguenther@suse.de>
8764 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8765 from early DWARF output.
8766 (dwarf2out_early_finish): Output line info unconditionally into
8767 early DWARF and add reference to it.
8769 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8772 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8773 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8774 other than V2DFmode using iptr mode attribute.
8775 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8777 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8779 PR rtl-optimization/84659
8780 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8782 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8785 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8787 (output_loc_list): Pass address of a dummy size variable even in the
8788 locview handling loop.
8789 (index_location_lists): Add comment on why skip_loc_list_entry can't
8792 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8795 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8798 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8801 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8802 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8803 and -mstring-compare-inline-limit.
8805 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8808 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8809 for stack clash protection in a register whenever we need it to be in
8812 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8814 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8815 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8817 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8820 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8822 (mlong-double-): Ditto.
8823 * config/rs6000/sysv4.opt (msdata=): Ditto.
8824 (mtls-size=): Ditto.
8826 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8828 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8829 erroneous entries for
8830 "vector int vec_ldl (int, long int *)", and
8831 "vector unsigned int vec_ldl (int, unsigned long int *)".
8832 Add comments and entries for
8833 "vector bool char vec_ldl (int, bool char *)",
8834 "vector bool short vec_ldl (int, bool short *)",
8835 "vector bool int vec_ldl (int, bool int *)",
8836 "vector bool long long vec_ldl (int, bool long long *)",
8837 "vector pixel vec_ldl (int, pixel *)",
8838 "vector long long vec_ldl (int, long long *)",
8839 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8840 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8841 type tree bool_long_long_type_node and correct definition of
8842 bool_V2DI_type_node to make reference to this new type tree.
8843 (rs6000_mangle_type): Replace erroneous reference to
8844 bool_long_type_node with bool_long_long_type_node.
8845 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8846 comments to emphasize sign distinctions for char and int types and
8847 replace RS6000_BTI_bool_long constant with
8848 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8849 use of RS6000_BTI_pixel.
8850 (bool_long_type_node): Remove this macro definition.
8851 (bool_long_long_type_node): New macro definition
8853 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8855 PR rtl-optimization/85300
8856 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8857 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8858 simplify_unary_operation fails.
8860 2018-04-10 Martin Liska <mliska@suse.cz>
8862 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8863 cgraph_edge and ipa_ref.
8865 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8869 * config/i386/sse.md
8870 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8871 computation of the VEC_MERGE selector from mask.
8872 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8873 Fix decoding of the VEC_MERGE selector into mask.
8875 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8877 PR tree-optimization/85286
8878 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8880 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8882 * final.c (final_1): Set insn_last_address as well as
8883 insn_current_address.
8885 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8888 * explow.c (emit_stack_probe): Call validize_mem on memory location
8889 before passing it to gen_probe_stack. Create address operand and
8890 legitimize it for the probe_stack_address case.
8892 2018-04-09 Jan Hubicka <jh@suse.cz>
8895 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8896 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8897 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8899 (free_lang_data): Rebuild type inheritance graph.
8901 2018-04-09 Martin Sebor <msebor@redhat.com>
8903 * invoke.texi (-finline-small-functions): Mention other optimization
8905 (-findirect-inlining, -fpartial-inlining): Same.
8906 (-finline-functions-called-once): Same.
8907 (-freorder-blocks-and-partition): Same.
8909 2018-04-09 Jan Hubicka <jh@suse.cz>
8912 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8913 jumps; choose last target that matches the criteria (i.e.
8914 no partition changes for non-crossing jumps).
8915 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8916 support for redirecting crossing jumps to non-crossing.
8918 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8920 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8921 also for naked functions.
8923 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8925 * config/arc/arc.md (add_shift): New pattern.
8926 (add_shift2): Likewise.
8927 (sub_shift): Likewise.
8928 (sub_shift_cmp0_noout): Likewise.
8929 (compare_si_ashiftsi): Likewise.
8930 (xbfu_cmp0_noout): New combine pattern.
8931 (xbfu_cmp0"): Likewise.
8932 (movsi_set_cc_insn): Place the predicable variant first.
8933 (commutative_binary_cmp0_noout): Remove clobber.
8934 (commutative_binary_cmp0): New pattern.
8935 (noncommutative_binary_cmp0): Likewise.
8936 (noncommutative_binary_cmp0_noout): Likewise.
8937 (noncommutative_binary_comparison_result_used): Removed.
8938 (rsub_cmp0): New pattern.
8939 (rsub_cmp0_noout): Likewise.
8940 (extzvsi): Changed, keep only meaningful variants.
8941 (SQH, SEZ): New iterators.
8942 (SQH_postfix): New mode attribute.
8943 (SEZ_prefix): New code attribute.
8944 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8945 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8946 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8948 (noncommutative_operator): Check the availability of barrel
8951 2018-04-09 Richard Biener <rguenther@suse.de>
8953 PR tree-optimization/85284
8954 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8955 Only use the niter constraining form of simple_iv when the exit
8958 2018-04-09 Tom de Vries <tom@codesourcery.com>
8961 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8962 (define_expand "*memory_barrier"): New define_expand.
8963 (define_insn "memory_barrier"): New insn.
8965 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8967 PR rtl-optimization/80463
8968 PR rtl-optimization/83972
8969 PR rtl-optimization/83480
8971 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8972 correct producer for the insn.
8973 (tidy_control_flow): Fixup seqnos in case of debug insns.
8975 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8977 PR rtl-optimization/83913
8979 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8980 different sched-times when merging exprs.
8982 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8984 PR rtl-optimization/83962
8986 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8987 tidy_fallthru_edge and tidy_control_flow.
8989 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8991 PR rtl-optimization/83530
8993 * sel-sched.c (force_next_insn): New global variable.
8994 (remove_insn_for_debug): When force_next_insn is true, also leave only
8995 next insn in the ready list.
8996 (sel_sched_region): When the region wasn't scheduled, make another pass
8997 over it with force_next_insn set to 1.
8999 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9001 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9003 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9004 for interrupt control.
9005 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9006 functions for interrupt control.
9007 * config/nds32/nds32-intrinsic.md: Likewise.
9008 * config/nds32/nds32_intrinsic.h: Likewise.
9009 * config/nds32/nds32.h (nds32_builtins): Likewise.
9011 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9013 * config/nds32/nds32.c (nds32_init_machine_status,
9014 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9015 strict_aligned_p field.
9016 (nds32_expand_to_rtl_hook): New function.
9017 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9018 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9020 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9021 Chung-Ju Wu <jasonwucj@gmail.com>
9023 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9024 * config/nds32/nds32-n7.md: New file.
9025 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9026 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9028 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9029 * config/nds32/nds32.md (pipeline_model): Add n7.
9030 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9031 * config/nds32/pipelines.md: Include n7 settings.
9033 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9034 Chung-Ju Wu <jasonwucj@gmail.com>
9036 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9037 * config/nds32/nds32-e8.md: New file.
9038 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9039 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9041 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9042 * config/nds32/nds32.md (pipeline_model): Add e8.
9043 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9044 * config/nds32/pipelines.md: Include e8 settings.
9046 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9047 Chung-Ju Wu <jasonwucj@gmail.com>
9049 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9050 * config/nds32/nds32-n8.md: New file.
9051 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9052 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9054 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9055 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9056 * config/nds32/nds32.md (pipeline_model): Add n8.
9057 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9058 * config/nds32/pipelines.md: Include n8 settings.
9060 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9061 Chung-Ju Wu <jasonwucj@gmail.com>
9063 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9064 * config/nds32/nds32-n9-2r1w.md: New file.
9065 * config/nds32/nds32-n9-3r2w.md: New file.
9066 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9067 nds32_register_ports): New or modify for cpu n9.
9068 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9070 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9071 * config/nds32/nds32-utils.c: New file.
9072 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9073 TARGET_MUL_SLOW): Define.
9074 * config/nds32/nds32.md (pipeline_model): New attribute.
9075 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9076 New options that support cpu n9.
9077 * config/nds32/pipelines.md: Include n9 settings.
9078 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9080 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9082 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9083 information if necessary.
9084 (output_cond_branch_compare_zero): Likewise.
9085 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9086 (nds32_target_alignment): Refine for alignment.
9087 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9088 (FUNCTION_BOUNDARY): Modify.
9089 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9091 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9093 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9095 * config/nds32/constants.md (unspec_volatile_element): Add values for
9096 TLB operation and data prefetch.
9097 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9098 functions for TLB operation and data prefetch.
9099 * config/nds32/nds32-intrinsic.md: Likewise.
9100 * config/nds32/nds32_intrinsic.h: Likewise.
9101 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9102 (nds32_print_operand): Likewise.
9103 * config/nds32/nds32.h (nds32_builtins): Likewise.
9105 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9106 Andrew Pinski <pinsika@gcc.gnu.org>
9109 * match.pd: Use constant_boolean_node of correct type instead of
9110 boolean_true_node or boolean_false_node for simplifying
9111 pointer comparisons to zero.
9113 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9115 PR tree-optimization/80021
9116 * tree.c (verify_type_variant): Make error call in verify_variant_match
9117 translatable and remove final full stop.
9119 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9121 * config/nds32/constants.md (unspec_volatile_element): Add
9122 UNSPEC_VOLATILE_EH_RETURN.
9123 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9124 nds32_output_stack_pop): Support dwarf exception handling process.
9125 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9126 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9127 exception handling process.
9128 (nds32_compute_stack_frame): Likewise.
9129 (nds32_return_addr_rtx): Likewise.
9130 (nds32_initial_elimination_offset): Likewise.
9131 (nds32_expand_prologue): Likewise.
9132 (nds32_expand_epilogue): Likewise.
9133 (nds32_dynamic_chain_address): New function.
9134 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9136 (DYNAMIC_CHAIN_ADDRESS): Define.
9137 (EH_RETURN_DATA_REGNO): Define.
9138 (EH_RETURN_STACKADJ_RTX): Define.
9139 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9140 patterns for dwarf exception handling.
9142 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9144 * config/nds32/nds32.h: Clean up obsolete macros.
9146 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9148 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9149 Add enum values for particular instructions.
9150 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9151 particular intrinsic functions.
9152 * config/nds32/nds32-intrinsic.md: Likewise.
9153 * config/nds32/nds32_intrinsic.h: Likewise.
9154 * config/nds32/nds32.h (nds32_builtins): Likewise.
9155 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9156 (btst, ave): New patterns for particular instructions.
9158 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9160 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9161 Add enum values for atomic load/store and memory sync.
9162 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9164 * config/nds32/nds32-intrinsic.md: Likewise.
9165 * config/nds32/nds32_intrinsic.h: Likewise.
9166 * config/nds32/nds32.h (nds32_builtins): Likewise.
9168 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9170 PR tree-optimization/85257
9171 * fold-const.c (native_encode_vector): If not all elts could fit
9172 and off is -1, return 0 rather than offset.
9173 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9174 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9175 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9176 adjust buffer in native_interpret_expr call.
9178 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9180 * config/nds32/constants.md (unspec_volatile_element): Add cache
9181 control enum values.
9182 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9183 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9184 * config/nds32/nds32.c (nds32_cctl_names): New.
9185 (nds32_print_operand): Handle cache control register names.
9186 * config/nds32/nds32.h (nds32_builtins): New enum values.
9187 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9189 * config/nds32/nds32.md (type): Add mmu.
9190 * config/nds32/pipelines.md (simple_insn): Add mmu.
9192 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9194 * config/nds32/nds32.md (type): Remove call.
9195 * config/nds32/pipelines.md (simple_insn): Likewise.
9197 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9199 * config/nds32/constants.md (unspec_volatile_element): Add
9200 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9201 UNSPEC_VOLATILE_FMFCFG.
9202 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9203 description for fmfcfg and fmfcsr.
9204 (bdesc_1arg): Add fmtcsr.
9205 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9206 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9207 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9208 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9209 unspec_fmfcfg): New patterns.
9210 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9211 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9212 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9213 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9214 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9215 __nds32__fmfcfg): Define.
9217 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9219 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9220 intrinsic register names.
9221 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9222 intrinsic register enum values and macros.
9224 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9226 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9227 for load/store addressing form.
9228 (nds32_print_operand_address): Likewise.
9230 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9233 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9234 based on LABEL_REF. Remove useless assertion.
9235 (pic_address_needs_scratch): Fix formatting.
9236 (sparc_legitimize_pic_address): Minor tweaks.
9237 (sparc_delegitimize_address): Adjust assertion accordingly.
9238 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9239 into symbolic_operand.
9240 (movsi_high_pic_label_ref): Likewise.
9241 (movsi_lo_sum_pic_label_ref): Likewise.
9242 (movdi_pic_label_ref): Likewise.
9243 (movdi_high_pic_label_ref): Likewise.
9244 (movdi_lo_sum_pic_label_ref): Likewise.
9246 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9248 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9249 custom LIB_SPEC setup.
9251 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9252 Kito Cheng <kito.cheng@gmail.com>
9254 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9255 * config/riscv/freebsd.h: New.
9257 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9259 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9260 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9263 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9264 Kito Cheng <kito.cheng@gmail.com>
9266 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9267 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9268 * config/nds32/nds32-protos.h (nds32_output_call,
9269 nds32_output_return): Declare.
9270 * config/nds32/nds32.md: Refine all the call and return patterns.
9272 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9275 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9276 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9278 PR rtl-optimization/84872
9279 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9280 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9283 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9285 * expr.c (copy_blkmode_to_reg): Revert 254862.
9286 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9288 2018-04-06 Richard Biener <rguenther@suse.de>
9291 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9292 after seeing a component reference with an adjacent field. Treat
9293 refs to arrays at struct end of external decls similar to
9294 refs to unconstrained commons.
9296 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9299 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9300 look through SAVE_EXPRs with non-side-effects argument. Adjust
9302 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9305 2018-04-06 Richard Biener <rguenther@suse.de>
9308 * alias.c (find_base_term): New wrapper around find_base_term
9309 unwinding CSELIB_VAL_PTR changes.
9310 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9313 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9315 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9317 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9318 constant definitions.
9319 ("nop"): lr 0,0 -> nopr r0
9320 ("nop_lr0", "nop_lr1"): New insn definitions.
9322 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9324 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9325 NDS32_V3PUSH_AVAILABLE_P macro.
9327 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9328 Chung-Ju Wu <jasonwucj@gmail.com>
9330 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9331 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9332 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9333 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9334 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9335 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9336 * config/nds32/constraints.md: New constraints and checking for hard
9337 float configuration.
9338 * config/nds32/iterators.md: New mode iterator and attribute for hard
9339 float configuration.
9340 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9342 * config/nds32/nds32-fpu.md: New file.
9343 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9344 deal with hard float code generation.
9345 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9347 (abi_type, float_reg_number): New enum type.
9348 * config/nds32/nds32-predicates.c: New predicates for hard float.
9349 * config/nds32/nds32-protos.h: Declare functions for hard float.
9350 * config/nds32/nds32.c: Implementation for hard float configuration.
9351 * config/nds32/nds32.h: Definitions for hard float configuration.
9352 * config/nds32/nds32.md: Include hard float machine description and
9353 modify patterns for hard float configuration.
9354 * config/nds32/nds32.opt: New options for hard float configuration.
9355 * config/nds32/predicates.md: New predicates for hard float
9358 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9360 * common/config/nds32/nds32-common.c
9361 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9363 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9366 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9367 CONSTRUCTOR_ELT (ctor, ...)->value.
9369 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9372 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9374 2018-04-05 Tom de Vries <tom@codesourcery.com>
9377 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9380 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9381 Kito Cheng <kito.cheng@gmail.com>
9383 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9384 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9385 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9386 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9388 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9389 Kito Cheng <kito.cheng@gmail.com>
9391 * config/nds32/constraints.md (Ufe): New memory constraint.
9392 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9393 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9394 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9396 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9397 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9399 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9401 * config/nds32/nds32.md: Use optimize_size in the condition for
9402 alu-shift instructions.
9404 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9406 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9408 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9410 * config/nds32/nds32.md (negsi2): Refine pattern.
9412 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9413 Chung-Ju Wu <jasonwucj@gmail.com>
9415 * config/nds32/iterators.md (shift_rotate): New code iterator.
9416 (shift): New code attribute.
9417 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9418 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9419 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9420 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9421 bit-wise operations.
9422 (andsi3, *andsi3): Ditto.
9423 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9424 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9425 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9426 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9427 nds32_ior_operand, nds32_xor_operand): New predicates.
9429 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9431 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9432 (addsi3, subsi3): ... this.
9434 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9436 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9438 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9440 * config/nds32/nds32.md: Adjust indention.
9442 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9444 * config/nds32/nds32.md (feature): New attribute.
9446 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9448 * config/nds32/nds32.md (subtype): New attribute.
9450 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9453 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9454 expansion to perform a bitwise AND of the argument followed by a
9455 boolean negation of the result.
9457 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9459 PR rtl-optimization/84878
9460 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9461 the basic block. Assert the use reference is not artificial and that
9462 it has an associated insn.
9464 2018-04-04 Michael Matz <matz@suse.de>
9466 * builtins.c (compute_objsize): Pass correct operand
9467 to array_at_struct_end_p.
9469 2018-04-04 Richard Biener <rguenther@suse.de>
9472 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9473 from contexts for DINFO_LEVEL_TERSE and below.
9475 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9477 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9478 resiter_operand condition.
9479 * config/nds32/nds32.md (*move<mode>): Ditto.
9481 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9482 Monk Chiang <sh.chiang04@gmail.com>
9484 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9486 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9488 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9490 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9491 Kito Cheng <kito.cheng@gmail.com>
9493 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9494 nds32_cond_code_str, output_cond_branch,
9495 output_cond_branch_compare_zero, nds32_expand_cbranch,
9496 nds32_expand_cstore, nds32_expand_movcc,
9497 nds32_output_cbranchsi4_equality_zero,
9498 nds32_output_cbranchsi4_equality_reg,
9499 nds32_output_cbranchsi4_equality_reg_or_const_int,
9500 nds32_output_cbranchsi4_greater_less_zero: New functions.
9501 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9502 nds32_expand_cstore, nds32_expand_movcc,
9503 nds32_output_cbranchsi4_equality_zero,
9504 nds32_output_cbranchsi4_equality_reg,
9505 nds32_output_cbranchsi4_equality_reg_or_const_int,
9506 nds32_output_cbranchsi4_greater_less_zero): Declare.
9507 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9508 nds32_rimm11s_operand): New predicates.
9509 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9510 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9513 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9515 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9516 * config/nds32/nds32.md: Ditto.
9517 * config/nds32/pipelines.md: Ditto.
9519 2018-04-04 Richard Biener <rguenther@suse.de>
9521 PR tree-optimization/85168
9522 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9523 propagating abnormals.
9525 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9527 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9529 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9530 Kito Cheng <kito.cheng@gmail.com>
9532 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9533 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9534 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9535 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9536 * config/nds32/nds32.md (sibcall_internal): New.
9537 (sibcall_register): Remove.
9538 (sibcall_immediate): Remove.
9539 (sibcall_value_internal): New.
9540 (sibcall_value_register): Remove.
9541 (sibcall_value_immediate): Remove.
9542 * config/nds32/predicates.md (nds32_general_register_operand): New.
9543 (nds32_call_address_operand): New.
9545 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9547 PR rtl-optimization/85167
9548 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9549 bb_defs if *split_p, instead preinitialize it to NULL.
9551 PR tree-optimization/85156
9552 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9553 evaluating the argument multiple times.
9555 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9557 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9559 (_mm_cvtpd_ps): Likewise.
9560 (_mm_cvttpd_epi32): Likewise.
9561 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9562 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9563 vector, pixel, and bool following altivec.h include.
9565 2018-04-03 Martin Sebor <msebor@redhat.com>
9567 * doc/extend.texi (Common Function Attributes): Clarify.
9568 (const attribute): Likewise.
9569 (pure attribute): Likewise.
9571 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9574 * config/i386/i386.c (ix86_expand_vector_set): Use
9575 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9577 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9579 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9580 instructions when changing rounding bits to preserve precision bits
9581 in the x87 control word.
9583 2018-04-03 Martin Liska <mliska@suse.cz>
9585 PR tree-optimization/82491
9586 * rtl.h (strip_offset_and_add): Replace += suboffset with
9587 poly_uint64 () + suboffset.
9589 2018-03-29 Martin Liska <mliska@suse.cz>
9590 Martin Jambor <mjambor@suse.cz>
9593 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9594 param_type is not an integral or pointer type.
9596 2018-04-03 Richard Biener <rguenther@suse.de>
9598 * sese.h (recompute_all_dominators): Remove.
9600 2018-04-02 Martin Sebor <msebor@redhat.com>
9602 * doc/invoke.texi (-Wrestrict): Fix typos.
9604 2018-04-02 Jim Wilson <jimw@sifive.com>
9606 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9607 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9608 (<optab>di3, <optab>si3_extend): Likewise.
9609 (<optab>si3_mask, <optab>si3_mask_1): New.
9610 (<optab>di3_mask, <optab>di3_mask_1): New.
9611 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9612 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9613 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9615 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9617 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9620 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9622 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9623 (nds32_canonicalize_comparison): New function.
9625 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9626 Kito Cheng <kito.cheng@gmail.com>
9627 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9629 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9630 * config/nds32/constants.md (unspec_volatile_element): Add
9631 UNSPEC_VOLATILE_RELAX_GROUP.
9632 * config/nds32/nds32-relax-opt.c: New file.
9633 * config/nds32/nds32-predicates.c
9634 (nds32_symbol_load_store_p): New function.
9635 * config/nds32/nds32-protos.h
9636 (nds32_symbol_load_store_p): Declare function.
9637 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9638 * config/nds32/nds32.c
9639 (nds32_register_pass): New function to register pass.
9640 (nds32_register_passes): New function to register passes.
9641 * config/nds32/nds32.md (relax_group): New pattern.
9642 * config/nds32/nds32.opt (mrelax-hint): New option.
9643 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9645 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9647 * config/nds32/t-nds32: Modify files dependency.
9649 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9651 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9652 (PROFILE_HOOK): Define its implementation.
9654 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9656 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9657 type and 32-bit size.
9659 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9662 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9663 (V_128_256): New mode iterator.
9664 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9665 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9666 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9668 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9669 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9671 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9674 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9675 NaN inputs correctly.
9677 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9680 * config/rs6000/vsx.md (??r): New mode attribute.
9681 (*vsx_mov<mode>_64bit): Use it.
9682 (*vsx_mov<mode>_32bit): Likewise.
9684 2018-03-30 Martin Sebor <msebor@redhat.com>
9686 PR tree-optimization/84818
9687 * builtins.c (check_access): Use warning_n.
9689 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9692 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9694 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9697 2018-03-30 Julia Koval <julia.koval@intel.com>
9700 * x86-tune.def (movx, partial_reg_dependency): Enable for
9703 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9706 * lra-constraints.c (process_alt_operands): Move setting
9707 this_alternative_matches below.
9709 2018-03-29 Martin Liska <mliska@suse.cz>
9712 * doc/invoke.texi: Document how LTO works with debug info.
9713 Describe auto-load support of binutils. Mention 'x86-64'
9714 as valid option value of -march option.
9716 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9718 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9721 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9722 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9723 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9726 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9729 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9730 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9731 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9732 for RS6000_BTM_POWERPC64.
9733 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9734 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9735 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9740 2018-03-28 Carl Love <cel@us.ibm.com>
9743 2017-09-27 Carl Love <cel@us.ibm.com>
9745 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9746 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9747 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9750 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9752 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9753 instead of __vector bool.
9754 (_mm_max_pu8): Likewise.
9755 (_mm_min_pi16): Likewise.
9757 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9760 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9761 (DIVWEUO): Likewise.
9763 (DIVDEUO): Likewise.
9764 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9765 DIVWEUO and DIVDEUO.
9766 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9767 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9768 (div_extend): Likewise.
9769 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9771 (__builtin_divweuo): Likewise.
9772 (__builtin_divdeo): Likewise.
9773 (__builtin_divdeuo): Likewise.
9775 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9778 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9779 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9781 PR tree-optimization/82004
9782 * gimple-match-head.c (optimize_pow_to_exp): New function.
9783 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9784 Don't fold to exp if optimize_pow_to_exp is false.
9786 2018-03-28 Martin Liska <mliska@suse.cz>
9789 * calls.c (initialize_argument_information): Fix trailing space.
9790 * common.opt: Fix typo and provide better explanation for
9791 -fsanitize-coverage option.
9792 * config/i386/i386.opt: Fix typo.
9794 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9795 Martin Liska <mliska@suse.cz>
9798 * gimplify.c (asan_poison_variable): Don't do the check for
9799 gimplify_omp_ctxp here.
9800 (gimplify_decl_expr): Do it here.
9801 (gimplify_target_expr): Likewise.
9803 2018-03-28 Martin Liska <mliska@suse.cz>
9806 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9807 chkp_type_bounds_count if MPX is not enabled.
9809 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9811 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9813 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9816 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9817 function to create the function decl for complex long double
9818 multiply and divide for -mabi=ieeelongdouble.
9819 (init_float128_ieee): Call it.
9821 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9824 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9825 -fcf-protection=branch -mibt.
9826 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9828 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9831 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9832 arm_word_relocations.
9834 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9837 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9838 extern array declarations.
9840 2018-03-27 Richard Biener <rguenther@suse.de>
9843 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9844 explicit single_use checks.
9846 2018-03-27 Richard Biener <rguenther@suse.de>
9848 PR tree-optimization/85082
9849 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9852 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9854 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9855 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9856 Turn on fasynchronous-unwind-tables and funwind-tables.
9858 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9861 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9862 (*bmi_blsr_<mode>_ccz): Ditto.
9864 2018-03-26 Tom de Vries <tom@codesourcery.com>
9866 PR tree-optimization/85063
9867 * omp-general.c (offloading_function_p): New function. Factor out
9869 * omp-offload.c (pass_omp_target_link::gate): ... here.
9870 * omp-general.h (offloading_function_p): Declare.
9871 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9872 with attribute omp declare target for offloading functions.
9874 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9876 PR tree-optimization/84005
9877 * tree-data-ref.h (get_base_for_alignment): Declare.
9878 * tree-data-ref.c (get_base_for_alignment_1): New function.
9879 (get_base_for_alignment): Likewise.
9880 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9881 get_base_for_alignment to find a suitable base object, instead
9882 of always using drb->base_address.
9884 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9887 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9888 known size by default.
9890 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9893 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9894 and non BLKmode operands.
9896 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9899 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9900 Clean up attributes.
9902 2018-03-23 Richard Biener <rguenther@suse.de>
9905 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9906 we are going to emit early debug for LTO.
9908 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9911 * function.c (match_asm_constraints_1): Don't optimize if input
9912 doesn't satisfy general_operand predicate for output's mode.
9915 * alias.c (write_dependence_p): Don't require for x_canonicalized
9916 non-VOIDmode if x has VOIDmode.
9919 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9920 just don't try to optimize it rather than assert it never happens.
9922 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9924 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9925 macro expansions for definition of ST_INTERNAL_<mode> and
9926 LD_INTERNAL_<mode> builtins.
9927 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9929 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9931 (altivec_expand_st_builtin): Likewise.
9932 (altivec_expand_builtin): Remove calls to deleted functions.
9933 (rs6000_address_for_altivec): Delete this function.
9934 * config/rs6000/vector.md: Remove expands for
9935 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9937 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9940 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9941 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9942 re-computing once computed.
9943 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9944 (arm_init_machine_status): Initialize
9945 machine->static_chain_stack_bytes.
9947 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9950 * doc/extend.texi: Add four new prototypes for vec_ld.
9951 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9952 definitions for more logical presentation.
9953 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9954 entries for V1TI variants of __builtin_altivec_ld builtin.
9955 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9956 handling of V1TI variant of LVX icode pattern.
9957 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9958 (rs6000_gimple_fold_builtin): Likewise.
9959 (altivec_init_builtins): Add code to define
9960 __builtin_altivec_lvx_v1ti function.
9962 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9965 * function.c (match_asm_constraints_1): Don't do the optimization
9966 if input isn't a REG, SUBREG, MEM or constant.
9968 2018-03-22 Tom de Vries <tom@codesourcery.com>
9970 PR tree-optimization/84956
9971 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9972 bb_has_abnormal_pred.
9974 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9977 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9978 DECL_INITIAL (decl) to decl at the end.
9979 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9982 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9984 * doc/extend.texi (__builtin_tgmath): Document when complex
9985 integer types are treated as _Complex _Float64.
9987 2018-03-21 Tom de Vries <tom@codesourcery.com>
9989 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9991 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9993 PR tree-optimization/84960
9994 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9995 if it is ENTRY block, move them into single succ of ENTRY in that case.
9997 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9999 PR tree-optimization/84811
10000 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10001 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10002 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10003 (known_subrange_p): Update accordingly. Cast each value involved
10004 in the size comparison, rather than casting the result of the
10007 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10009 PR tree-optimization/84982
10010 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10011 by flipping the least significant bit rather than all bits from
10012 bitpos to bitpos + bitsize - 1.
10014 2018-03-21 Nathan Sidwell <nathan@acm.org>
10016 * doc/extend.texi (Deprecated Features): Remove mention of
10017 long-deleted deprecations.
10019 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10022 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10023 * configure: Regenerate.
10025 2018-03-21 Tom de Vries <tom@codesourcery.com>
10027 PR tree-optimization/83126
10028 * tree-parloops.c (num_phis): New function.
10029 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10031 2018-03-21 Nathan Sidwell <nathan@acm.org>
10033 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10034 mention anon-struct/union members and trailing attributes.
10036 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10038 PR tree-optimization/84969
10039 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10040 builtin memset partitions if they set different rhs values.
10042 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10044 PR rtl-optimization/84989
10045 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10046 VEC_DUPLICATE with scalar result mode.
10048 2018-03-21 Martin Liska <mliska@suse.cz>
10051 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10052 not intended return statement.
10054 2018-03-21 Martin Liska <mliska@suse.cz>
10057 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10058 (chkp_find_bound_slots_1): Limit number of iterations.
10060 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10063 * Minor grammar fixes for x86 options.
10065 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10068 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10069 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10071 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10074 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10075 (altivec_lvx_<mode>_1op): Likewise.
10076 (altivec_stvx_<mode>_2op): Likewise.
10077 (altivec_stvx_<mode>_1op): Likewise.
10078 (altivec_lvx_<VM2:mode>): New define_expand.
10079 (altivec_stvx_<VM2:mode>): Likewise.
10080 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10081 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10082 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10083 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10084 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10085 (rs6000_gen_lvx): Likewise.
10086 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10087 (altivec_expand_stv_builtin): Likewise.
10088 (altivec_expand_builtin): Likewise.
10089 * config/rs6000/vector.md: Likewise.
10091 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10094 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10097 2018-03-20 Richard Biener <rguenther@suse.de>
10100 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10101 sign-conversions as zero, fall back to standard scalar_stmt
10104 2018-03-20 Martin Liska <mliska@suse.cz>
10107 * predict.c (rebuild_frequencies): Handle case when we have
10108 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10110 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10113 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10114 flag_section_anchors.
10115 * varasm.c (use_blocks_for_decl_p): Remove hack for
10116 dw2_force_const_mem.
10119 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10121 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10122 be created, use lowpart_subreg of operands[0] rather than operands[0]
10124 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10125 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10126 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10127 and n constraint instead of aarch64_shift_imm_di and Usd.
10128 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10129 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10131 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10134 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10135 to favor GPR over NEON registers.
10136 (<shift>di3_neon): Likewise.
10138 2018-03-20 Tom de Vries <tom@codesourcery.com>
10141 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10142 (nvptx_process_pars): Emit bar.sync asap and alap.
10144 2018-03-20 Tom de Vries <tom@codesourcery.com>
10147 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10148 seen_label if seen_label is already set.
10150 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10153 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10154 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10155 Use 1U instead of 1. Formatting fixes.
10158 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10159 instead of TREE_TYPE (s1) for the return value.
10161 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10163 PR tree-optimization/84946
10164 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10165 bitsize + bitsize in poly_uint64 rather than poly_int64.
10168 * dwarf2asm.c: Include fold-const.c.
10169 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10170 of decl rather than decl itself.
10172 PR rtl-optimization/84643
10173 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10175 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10178 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10179 calling assemble_variable.
10181 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10184 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10185 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10187 2018-03-19 Jim Wilson <jimw@sifive.com>
10190 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10191 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10192 (riscv_first_stack_step): Likewise.
10193 (riscv_option_override): Use STACK_BOUNDARY instead of
10194 MIN_STACK_BOUNDARY.
10195 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10196 MIN_STACK_BOUNDARY.
10197 (BIGGEST_ALIGNMENT): Set to 128.
10198 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10199 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10202 2018-03-19 Richard Biener <rguenther@suse.de>
10204 PR tree-optimization/84933
10205 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10206 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10208 2018-03-19 Richard Biener <rguenther@suse.de>
10210 PR tree-optimization/84859
10211 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10212 (cond_if_else_store_replacement): Perform sinking operation on
10213 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10214 Generalize what a BB with a single eligible store is.
10216 2018-03-19 Richard Biener <rguenther@suse.de>
10218 PR tree-optimization/84929
10219 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10220 chrec_is_positive against non-chrec arg.
10222 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10225 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10227 2018-03-18 Martin Liska <mliska@suse.cz>
10229 PR rtl-optimization/84635
10230 * regrename.c (build_def_use): Use matches_mode only when
10233 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10235 PR tree-optimization/84913
10236 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10237 vectorize chains of COND_EXPRs.
10239 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10241 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10243 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10245 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10247 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10249 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10251 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10252 Kito Cheng <kito.cheng@gmail.com>
10254 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10255 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10256 (nds32_adjust_reg_alloc_order): New function.
10257 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10259 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10261 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10262 nds32_print_operand, nds32_print_operand_address): Use
10263 HOST_WIDE_INT_PRINT_DEC instead.
10265 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10267 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10269 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10272 * config/i386/i386.c (initial_ix86_tune_features,
10273 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10274 unsigned long long.
10275 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10276 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10277 rather than 1u << ix86_tune. Formatting fix.
10278 (ix86_option_override_internal): Change ix86_arch_mask from
10279 unsigned int to unsigned HOST_WIDE_INT, initialize to
10280 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10281 (ix86_function_specific_restore): Likewise.
10283 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10286 * postreload.c (reload_combine_recognize_pattern): Perform
10287 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10288 truncate_int_for_mode the result for the destination's mode.
10291 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10292 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10294 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10297 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10300 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10301 regno_allocno_class_array and sorted_pseudos.
10302 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10303 insns where regno is used.
10305 2018-03-16 Martin Liska <mliska@suse.cz>
10308 * multiple_target.c (create_dispatcher_calls): Redirect
10309 reference in the symbol table.
10311 2018-03-16 Martin Liska <mliska@suse.cz>
10314 * multiple_target.c (create_dispatcher_calls): Redirect also
10317 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10321 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10322 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10323 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10325 2018-03-16 Julia Koval <julia.koval@intel.com>
10327 * doc/invoke.texi (Skylake Server): Add CLWB.
10328 Cannonlake): Remove CLWB.
10330 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10332 PR tree-optimization/84841
10333 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10335 (FLOAT_ONE_CONST_TYPE): Define.
10336 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10337 (sort_by_operand_rank): Put entries with higher constant_type last
10338 rather than first to match comments.
10340 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10342 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10345 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10348 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10349 check and use address of target always.
10351 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10354 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10355 (indirect_thunk_bnd_needed): Likewise.
10356 (indirect_thunks_used): Likewise.
10357 (indirect_thunks_bnd_used): Likewise.
10358 (indirect_return_needed): New.
10359 (indirect_return_bnd_needed): Likewise.
10360 (output_indirect_thunk_function): Add a bool argument for
10362 (output_indirect_thunk_function): Don't generate alias for
10363 function return thunk.
10364 (ix86_code_end): Call output_indirect_thunk_function to generate
10365 function return thunks.
10366 (ix86_output_function_return): Set indirect_return_bnd_needed
10367 and indirect_return_needed instead of indirect_thunk_bnd_needed
10368 and indirect_thunk_needed.
10370 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10372 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10373 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10374 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10376 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10377 Paul Hua <paul.hua.gm@gmail.com>
10380 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10382 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10384 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10385 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10386 resp. SFmode cases.
10388 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10391 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10392 instead of GET_MODE_SIZE when comparing Units.
10394 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10397 * varasm.c (hash_section): Return an unchangeble hash value
10398 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10399 Return !aarch64_can_use_per_function_literal_pools_p ().
10401 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10404 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10405 rather than address of cmode as last argument to prepare_cmp_insn.
10407 2018-03-15 Julia Koval <julia.koval@intel.com>
10409 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10410 F_AVX512VNNI, F_AVX512BITALG): New.
10412 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10415 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10416 insn for floating-point loads and stores.
10418 2018-03-14 Carl Love <cel@us.ibm.com>
10420 * config/rs6000/rs6000-c.c: Add macro definitions for
10421 ALTIVEC_BUILTIN_VEC_PERMXOR.
10422 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10423 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10424 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10425 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10427 * config/doc/extend.texi: Add prototypes for vec_permxor.
10429 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10432 * diagnostic-show-locus.c (class layout_point): Convert m_line
10433 from int to linenum_type.
10434 (line_span::comparator): Use linenum "compare" function when
10435 comparing line numbers.
10436 (test_line_span): New function.
10437 (layout_range::contains_point): Convert param "row" from int to
10439 (layout_range::intersects_line_p): Likewise.
10440 (layout::will_show_line_p): Likewise.
10441 (layout::print_source_line): Likewise.
10442 (layout::should_print_annotation_line_p): Likewise.
10443 (layout::print_annotation_line): Likewise.
10444 (layout::print_leading_fixits): Likewise.
10445 (layout::annotation_line_showed_range_p): Likewise.
10446 (struct line_corrections): Likewise for field m_row.
10447 (line_corrections::line_corrections): Likewise for param "row".
10448 (layout::print_trailing_fixits): Likewise.
10449 (layout::get_state_at_point): Likewise.
10450 (layout::get_x_bound_for_row): Likewise.
10451 (layout::print_line): Likewise.
10452 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10453 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10454 * input.c (selftest::test_linenum_comparisons): New function.
10455 (selftest::input_c_tests): Call it.
10456 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10457 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10458 * selftest.h (ASSERT_GT): New macro.
10459 (ASSERT_GT_AT): New macro.
10460 (ASSERT_LT): New macro.
10461 (ASSERT_LT_AT): New macro.
10463 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10465 PR rtl-optimization/84780
10466 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10468 2018-03-14 Martin Liska <mliska@suse.cz>
10470 * tree.c (record_node_allocation_statistics): Use
10471 get_stats_node_kind.
10472 (get_stats_node_kind): New function extracted from
10473 record_node_allocation_statistics.
10474 (free_node): Use get_stats_node_kind.
10476 2018-03-14 Richard Biener <rguenther@suse.de>
10478 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10479 that the value-set of ANTIC_IN doesn't grow.
10482 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10484 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10485 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10487 2018-03-14 Julia Koval <julia.koval@intel.com>
10489 * config.gcc (icelake-client, icelake-server): New.
10491 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10492 (initial_ix86_arch_features): Ditto.
10493 (PTA_SKYLAKE): Add SGX.
10494 (PTA_ICELAKE): Remove.
10495 (PTA_ICELAKE_CLIENT): New.
10496 (PTA_ICELAKE_SERVER): New.
10497 (ix86_option_override_internal): Split up icelake on icelake client and
10499 (get_builtin_code_for_version): Ditto.
10500 (fold_builtin_cpu): Ditto.
10501 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10502 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10503 * config/i386/i386.h (processor_type): Ditto.
10504 * doc/invoke.texi: Ditto.
10506 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10509 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10510 INTEGER_CST offset, add it together with bitpos / 8 and
10511 sign extend based on POINTER_SIZE.
10515 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10518 * config/i386/constraints.md (Yc): New register constraint.
10519 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10520 Use Yc constraint for alternative 2 of operand 0. Remove
10521 preferred_for_speed attribute.
10523 2018-03-14 Richard Biener <rguenther@suse.de>
10525 PR tree-optimization/84830
10526 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10527 with the old one to avoid oscillations.
10529 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10532 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10534 (assign_by_spills): Return a flag of reload assignment failure.
10535 Do not process the reload assignment failures. Do not spill other
10536 reload pseudos if they has the same reg class. Update n if
10538 (lra_assign): Add a return arg. Set up from the result of
10539 assign_by_spills call.
10540 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10541 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10542 usage_insns if it is not NULL.
10543 (spill_hard_reg_in_range): New function.
10544 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10545 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10546 function prototypes.
10547 (lra_assign): Change prototype.
10548 * lra.c (lra): Add code to deal with fails by splitting hard reg
10551 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10553 * config/riscv/riscv.opt (mrelax): New option.
10554 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10555 "norelax" when riscv_mrelax is disabled.
10556 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10558 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10561 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10562 reassociation for int modes.
10564 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10566 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10567 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10569 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10570 * config/aarch64/aarch64-sve.md
10571 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10572 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10573 (*extend<mode><Vwide>2): Rename to...
10574 (aarch64_sve_extend<mode><Vwide>2): ...this.
10575 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10576 renaming the old pattern to...
10577 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10579 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10580 define_expand, renaming the old pattern to...
10581 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10582 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10583 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10584 account when deciding which SVE instruction the optab should use.
10585 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10587 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10589 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10590 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10591 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10592 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10593 (tlsdesc_small_<mode>): Turn a define_expand and use
10594 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10595 (tlsdesc_small_advsimd_<mode>): ...this.
10596 (tlsdesc_small_sve_<mode>): New pattern.
10598 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10600 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10601 (UNSPEC_UMUL_HIGHPART): New constants.
10602 (MUL_HIGHPART): New int iteraor.
10603 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10604 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10606 (*<su>mul<mode>3_highpart): New define_insn.
10608 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10611 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10614 2018-03-13 Martin Liska <mliska@suse.cz>
10617 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10619 (sem_item_optimizer::~sem_item_optimizer): Release it.
10620 (sem_item_optimizer::merge_classes): Register variable aliases.
10621 (sem_item_optimizer::fixup_pt_set): New function.
10622 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10623 * ipa-icf.h: Declare new variables and functions.
10625 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10627 PR middle-end/84834
10628 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10629 integer_pow2p@2 and test integer_pow2p in condition.
10630 (A < 0 ? C : 0): Similarly for @1.
10632 PR middle-end/84831
10633 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10634 characters starting at p contain '\0' character, don't look beyond
10638 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10639 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10642 * reg-stack.c (change_stack): Change update_end var from int to
10643 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10644 also call set_block_for_insn on the newly added insns and rescan.
10647 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10648 on the last operand.
10651 * tree.c (stabilize_reference_1): Return save_expr (e) for
10652 STATEMENT_LIST even if it doesn't have side-effects.
10654 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10656 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10658 2018-03-12 Renlin Li <renlin.li@arm.com>
10660 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10661 aarch64_output_scalar_simd_mov_immediate.
10663 2018-03-12 Martin Sebor <msebor@redhat.com>
10665 PR tree-optimization/83456
10666 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10667 for perfectly overlapping calls to memcpy.
10668 (gimple_fold_builtin_memory_chk): Same.
10669 (gimple_fold_builtin_strcpy): Handle no-warning.
10670 (gimple_fold_builtin_stxcpy_chk): Same.
10671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10673 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10675 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10676 parameter. Use it for SFmode.
10677 (rs6000_function_arg_advance_1): Adjust.
10678 (rs6000_function_arg): Adjust.
10679 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10681 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10683 PR rtl-optimization/84169
10684 PR rtl-optimization/84780
10685 * combine.c (can_combine_p): Check for a 2-insn combination whether
10686 the destination register is used between the two insns, too.
10688 2018-03-12 Richard Biener <rguenther@suse.de>
10690 PR tree-optimization/84803
10691 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10692 for refs DR analysis didn't process.
10694 2018-03-12 Richard Biener <rguenther@suse.de>
10696 PR tree-optimization/84777
10697 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10698 force-vectorize loops ignore whether we are optimizing for size.
10700 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10702 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10703 (TARGET_MD_ASM_ADJUST): Define.
10705 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10706 Kito Cheng <kito.cheng@gmail.com>
10707 Chung-Ju Wu <jasonwucj@gmail.com>
10709 * config/nds32/nds32.c (nds32_compute_stack_frame,
10710 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10711 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10712 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10713 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10714 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10715 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10716 * config/nds32/nds32.md (prologue, epilogue): Use macro
10717 NDS32_V3PUSH_AVAILABLE_P to do checking.
10719 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10722 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10723 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10724 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10725 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10726 addition of most attributes on !orig_type_die or the attribute not
10727 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10729 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10730 Chung-Ju Wu <jasonwucj@gmail.com>
10732 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10733 __NDS32_VH__ macro.
10734 * config/nds32/nds32.opt (mvh): New option.
10736 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10737 Chung-Ju Wu <jasonwucj@gmail.com>
10739 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10741 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10742 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10745 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10746 Chung-Ju Wu <jasonwucj@gmail.com>
10748 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10750 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10751 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10753 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10754 Kito Cheng <kito.cheng@gmail.com>
10755 Chung-Ju Wu <jasonwucj@gmail.com>
10757 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10758 UNSPEC_FFMISM and UNSPEC_FLMISM.
10759 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10760 for ffb, ffmism and flmism.
10761 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10762 (unspec_ffmism): Ditto.
10763 (unspec_flmism): Ditto.
10764 (nds32_expand_builtin_impl): Check if string extension is available.
10765 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10766 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10768 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10771 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10774 * lra-assigns.c (assign_by_spills): Return a flag of reload
10775 assignment failure. Do not process the reload assignment
10776 failures. Do not spill other reload pseudos if they has the same
10778 (lra_assign): Add a return arg. Set up from the result of
10779 assign_by_spills call.
10780 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10781 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10782 usage_insns if it is not NULL.
10783 (spill_hard_reg_in_range): New function.
10784 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10785 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10786 function prototypes.
10787 (lra_assign): Change prototype.
10788 * lra.c (lra): Add code to deal with fails by splitting hard reg
10791 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10794 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10796 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10799 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10800 (dw_val_node): Add val_symbolic_view.
10801 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10802 (symview_upper_bound): New.
10803 (new_line_info_table): Initialize symviews_since_reset.
10804 (dwarf2out_source_line): Count symviews_since_reset and set
10805 symview_upper_bound.
10806 (dw_val_equal_p): Handle symview.
10807 (add_AT_symview): New.
10808 (print_dw_val): Handle symview.
10809 (attr_checksum, attr_checksum_ordered): Likewise.
10810 (same_dw_val_p, size_of_die): Likewise.
10811 (value_format, output_die): Likewise.
10812 (add_high_low_attributes): Use add_AT_symview for entry_view.
10813 (dwarf2out_finish): Reset symview_upper_bound, clear
10816 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10819 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10820 Add strict argument and use it.
10821 (rs6000_split_multireg_move): Update for new strict argument.
10822 (mem_operand_gpr): Disallow all non-offsettable addresses.
10823 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10825 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10828 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10829 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10830 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10833 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10834 decl, use remap_type if we want to use the type.
10836 2018-03-09 Martin Sebor <msebor@redhat.com>
10838 PR tree-optimization/84526
10839 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10841 (builtin_access::generic_overlap): Be prepared to handle non-array
10844 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10846 PR rtl-optimization/84682
10847 * lra-constraints.c (process_address_1): Check is_address flag
10848 for address constraints.
10849 (process_alt_operands): Likewise.
10850 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10851 preprocess_constraints.
10852 * recog.h (preprocess_constraints): Add oploc parameter.
10854 * recog.c (preprocess_constraints): Test address_operand for
10855 CT_ADDRESS constraints.
10857 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10860 * lra-assigns.c (assign_by_spills): Return a flag of reload
10861 assignment failure. Do not process the reload assignment
10862 failures. Do not spill other reload pseudos if they has the same
10864 (lra_assign): Add a return arg. Set up from the result of
10865 assign_by_spills call.
10866 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10867 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10868 usage_insns if it is not NULL.
10869 (spill_hard_reg_in_range): New function.
10870 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10871 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10872 function prototypes.
10873 (lra_assign): Change prototype.
10874 * lra.c (lra): Add code to deal with fails by splitting hard reg
10877 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10880 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10881 Accept complain bool parameter. Only emit errors if it is true.
10882 (arm_parse_cpu_option_name): Likewise.
10883 (arm_target_thumb_only): Adjust callers of the above.
10884 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10885 prototype to take a default true bool parameter.
10886 (arm_parse_arch_option_name): Likewise.
10888 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10889 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10893 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10894 * configure: Regenerate.
10895 * configure.ac ("linker --version-script option"): New.
10896 ("linker soname option"): New.
10898 2018-03-09 Richard Biener <rguenther@suse.de>
10900 PR tree-optimization/84775
10901 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10902 immediate uses of predicate stmts and mark them modified.
10905 PR tree-optimization/84178
10906 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10908 (version_loop_for_if_conversion): Delay update_ssa call.
10909 (tree_if_conversion): Delay update_ssa until after predicate
10912 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10915 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10916 when the function accesses prior frames.
10918 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10921 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10922 gen_llsym, otherwise call maybe_gen_llsym.
10924 PR inline-asm/84742
10925 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10926 has ',' character inside of it.
10928 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10931 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10932 as clobbering CC_REGNUM.
10934 2018-03-08 Richard Biener <rguenther@suse.de>
10936 PR middle-end/84552
10937 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10938 (follow_copies_to_constant): Do not follow SSA names registered
10941 2018-03-08 Richard Biener <rguenther@suse.de>
10943 PR tree-optimization/84178
10944 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10946 (version_loop_for_if_conversion): Delay update_ssa call.
10947 (tree_if_conversion): Delay update_ssa until after predicate
10950 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10952 PR tree-optimization/84178
10953 * tree-if-conv.c (release_bb_predicate): Remove the
10954 the assertion that the stmts have NULL use_ops.
10955 Discard the statements, asserting that they haven't
10956 yet been added to a BB.
10958 2018-03-08 Richard Biener <rguenther@suse.de>
10960 PR tree-optimization/84746
10961 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10962 (phi_translate): Pass in destination ANTIC_OUT set.
10963 (phi_translate_1): Likewise. For a simplified result lookup
10964 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10965 (phi_translate_set): Adjust.
10966 (do_pre_regular_insertion): Likewise.
10967 (do_pre_partial_partial_insertion): Likewise.
10969 2018-03-08 Martin Liska <mliska@suse.cz>
10971 PR gcov-profile/84735
10972 * doc/gcov.texi: Document usage of profile files.
10973 * gcov-io.h: Document changes in the format.
10975 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10979 * dwarf2out.c (struct dw_line_info_table): Update comments for
10981 (FORCE_RESET_NEXT_VIEW): New.
10982 (FORCE_RESETTING_VIEW_P): New.
10983 (RESETTING_VIEW_P): Check for -1 too.
10984 (ZERO_VIEW_P): Likewise.
10985 (new_line_info_table): Force-reset next view.
10986 (dwarf2out_begin_function): Likewise.
10987 (dwarf2out_source_line): Simplify zero_view_p initialization.
10988 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10989 view directly. Omit view when omitting .loc at line 0.
10991 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10993 PR tree-optimization/84740
10994 * tree-switch-conversion.c (process_switch): Call build_constructors
10995 only if info.phi_count is non-zero.
10997 PR tree-optimization/84739
10998 * tree-tailcall.c (find_tail_calls): Check call arguments against
10999 DECL_ARGUMENTS (current_function_decl) rather than
11000 DECL_ARGUMENTS (func) when checking for tail recursion.
11002 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11004 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11005 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11006 Volker Reichelt's entry and add entries for people that perform
11007 GCC fuzzy testing and report numerous bugs.
11009 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11012 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11013 readonly data in sdata, if that is disabled.
11014 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11015 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11016 -mreadonly-in-sdata option.
11018 2018-03-07 Martin Sebor <msebor@redhat.com>
11020 PR tree-optimization/84468
11021 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11022 basic block when looking for nul assignment.
11024 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11027 * except.h (output_function_exception_table): Adjust prototype.
11028 * except.c (output_function_exception_table): Remove FNNAME parameter
11029 and add SECTION parameter. Ouput one part of the table at a time.
11030 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11031 the first part of the exception table and emit unwind directives.
11032 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11033 (i386_pe_seh_cold_init): Likewise.
11034 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11035 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11036 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11037 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11038 * config/i386/winnt.c: Include except.h.
11039 (struct seh_frame_state): Add reg_offset, after_prologue and
11040 in_cold_section fields.
11041 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11042 (i386_pe_seh_cold_init): New function.
11043 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11044 to seh->in_cold_section.
11045 (seh_emit_push): Record the offset of the push.
11046 (seh_emit_save): Record the offet of the save.
11047 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11048 Test seh->after_prologue to disregard the epilogue.
11049 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11050 (i386_pe_end_cold_function): New function.
11052 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11055 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11056 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11059 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11061 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11062 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11064 PR middle-end/84723
11065 * multiple_target.c: Include tree-inline.h and intl.h.
11066 (expand_target_clones): Diagnose and fail if node->definition and
11067 !tree_versionable_function_p (node->decl).
11069 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11071 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11073 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11074 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11075 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11077 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11080 * combine.c (try_combine): Use reg_or_subregno instead of handling
11081 just paradoxical SUBREGs and REGs.
11083 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11085 * config/arc/arc.c (arc_finalize_pic): Remove function.
11086 (arc_must_save_register): We use single base PIC register, remove
11087 checks to save/restore the PIC register.
11088 (arc_expand_prologue): Likewise.
11089 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11091 (arc_verify_short): Likewise.
11092 (arc_attr_type): Likewise.
11093 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11094 (walk_stores): Likewise.
11095 (arc_address_cost): Make it static.
11096 (arc_verify_short): Likewise.
11097 (branch_dest): Likewise.
11098 (arc_attr_type): Likewise.
11099 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11100 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11101 (arc_final_prescan_insn): Remove inserting the nops due to
11102 hardware hazards. It is done in reorg step.
11103 (insn_length_variant_t): Remove.
11104 (insn_length_parameters_t): Likewise.
11105 (arc_insn_length_parameters): Likewise.
11106 (arc_get_insn_variants): Likewise.
11107 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11109 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11111 PR inline-asm/84683
11112 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11115 PR tree-optimization/84687
11116 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11118 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11120 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11122 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11123 Rename to ppc_speculation_barrier.
11124 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11125 __builtin_ppc_speculation_barrier.
11127 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11130 * combine.c (combine_simplify_rtx): Don't try to simplify if
11131 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11134 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11136 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11137 to 32 bytes when compiling for POWER9.
11139 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11142 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11143 regparm >= 3 with no arg reg available also for calls with
11144 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11147 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11149 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11151 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11154 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11156 2018-03-05 Richard Biener <rguenther@suse.de>
11158 PR tree-optimization/84486
11159 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11160 When inserting a __builtin_assume_aligned call set the LHS
11161 SSA name alignment info accordingly.
11163 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11165 PR tree-optimization/84114
11166 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11167 Avoid reassociation of FLOAT_MODE addition.
11169 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11171 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11172 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11173 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11174 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11175 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11176 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11177 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11179 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11180 __builtin_ia32_wbinvd): New builtins.
11181 (SPECIAL_ARGS2): New.
11182 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11183 (SPECIAL_ARGS2): New.
11184 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11185 (ix86_valid_target_attribute_inner_p): Ditto.
11186 (ix86_init_mmx_sse_builtins): Add special_args2.
11187 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11188 TARGET_WBNOINVD_P): New.
11189 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11190 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11191 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11192 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11193 * config/i386/pconfigintrin.h: New file.
11194 * config/i386/wbnoinvdintrin.h: Ditto.
11195 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11197 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11199 2018-03-05 Richard Biener <rguenther@suse.de>
11201 PR tree-optimization/84670
11202 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11204 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11205 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11206 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11207 doesn't grow if all successors (recursively) were visited at least
11210 2018-03-05 Richard Biener <rguenther@suse.de>
11212 PR tree-optimization/84650
11213 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11214 if executed in the loop pipeline.
11216 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11218 * doc/configfiles.texi (Configuration Files): Move info about
11219 conditionalizing $target-protos.h to...
11220 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11221 differs from $target-protos.h.
11223 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11224 Chung-Ju Wu <jasonwucj@gmail.com>
11226 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11227 * config/nds32/nds32-multiple.md (setmemsi): Define.
11228 * config/nds32/nds32-memory-manipulation.c
11229 (nds32_gen_dup_4_byte_to_word_value): New.
11230 (emit_setmem_word_loop): New.
11231 (emit_setmem_byte_loop): New.
11232 (nds32_expand_setmem_loop): New.
11233 (nds32_expand_setmem_loop_v3m): New.
11234 (nds32_expand_setmem_unroll): New.
11235 (nds32_expand_setmem): New.
11237 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11238 Chung-Ju Wu <jasonwucj@gmail.com>
11240 * config/nds32/nds32-memory-manipulation.c
11241 (nds32_emit_load_store): New.
11242 (nds32_emit_post_inc_load_store): New.
11243 (nds32_emit_mem_move): New.
11244 (nds32_emit_mem_move_block): New.
11245 (nds32_expand_movmemsi_loop_unknown_size): New.
11246 (nds32_expand_movmemsi_loop_known_size): New.
11247 (nds32_expand_movmemsi_loop): New.
11248 (nds32_expand_movmemsi_unroll): New.
11249 (nds32_expand_movmemqi): Rename ...
11250 (nds32_expand_movmemsi): ... to this.
11251 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11252 (movmemsi): ... to this.
11253 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11254 (nds32_expand_movmemsi): ... to this.
11256 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11257 Monk Chiang <sh.chiang04@gmail.com>
11258 Chung-Ju Wu <jasonwucj@gmail.com>
11260 * config/nds32/nds32-protos.h
11261 (nds32_expand_load_multiple): New arguments.
11262 (nds32_expand_store_multiple): Ditto.
11263 (nds32_valid_multiple_load_store): Rename ...
11264 (nds32_valid_multiple_load_store_p): ... to this.
11265 * config/nds32/nds32-memory-manipulation.c
11266 (nds32_expand_load_multiple): Refine implementation.
11267 (nds32_expand_store_multiple): Ditto.
11268 * config/nds32/nds32-multiple.md
11269 (load_multiple): Update nds32_expand_load_multiple interface.
11270 (store_multiple): Update nds32_expand_store_multiple interface.
11271 * config/nds32/nds32-predicates.c
11272 (nds32_valid_multiple_load_store): Rename ...
11273 (nds32_valid_multiple_load_store_p): ... to this and refine
11275 * config/nds32/predicates.md
11276 (nds32_load_multiple_and_update_address_operation): New predicate.
11277 (nds32_store_multiple_and_update_address_operation): New predicate.
11279 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11280 Chung-Ju Wu <jasonwucj@gmail.com>
11282 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11283 (combo): New attribute.
11284 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11286 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11288 * config/nds32/nds32.opt: Change -mcmodel= default value.
11290 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11291 Monk Chiang <sh.chiang04@gmail.com>
11292 Chung-Ju Wu <jasonwucj@gmail.com>
11294 * config/nds32/constants.md (unspec_element): New enum.
11295 * config/nds32/constraints.md (Umw): New constraint.
11296 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11297 * config/nds32/nds32-intrinsic.md: Likewise.
11298 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11299 (nds32_valid_smw_lwm_base_p): New.
11300 (nds32_output_smw_single_word): New.
11301 (nds32_output_lmw_single_word): New.
11302 (nds32_expand_unaligned_load): New.
11303 (nds32_expand_unaligned_store): New.
11304 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11305 (nds32_output_smw_single_word): Declare.
11306 (nds32_output_lmw_single_word): Declare.
11307 (nds32_expand_unaligned_load): Declare.
11308 (nds32_expand_unaligned_store): Declare.
11309 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11310 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11311 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11312 NDS32_BUILTIN_UASTORE_DW.
11313 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11316 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11317 Kito Cheng <kito.cheng@gmail.com>
11318 Chung-Ju Wu <jasonwucj@gmail.com>
11320 * config/nds32/nds32-intrinsic.c
11321 (nds32_expand_builtin_null_ftype_reg): Delete.
11322 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11323 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11324 (nds32_read_argument): New.
11325 (nds32_legitimize_target): Ditto.
11326 (nds32_legitimize_argument): Ditto.
11327 (nds32_check_constant_argument): Ditto.
11328 (nds32_expand_unop_builtin): Ditto.
11329 (nds32_expand_unopimm_builtin): Ditto.
11330 (nds32_expand_binop_builtin): Ditto.
11331 (nds32_builtin_decl_impl): Ditto.
11332 (builtin_description): Ditto.
11333 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11334 (nds32_init_builtins_impl): Ditto.
11335 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11336 (nds32_builtin_decl): New.
11337 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11338 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11340 2018-03-02 Jeff Law <law@redhat.com>
11342 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11343 (redundant_insn, fill_simple_delay_slots): Likewise.
11344 (fill_slots_from_thread): Likewise.
11345 * resource.c (mark_referenced_resources): Likewise.
11346 (mark_set_resources, find_dead_or_set_registers): Likewise.
11348 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11350 * substring-locations.h (format_warning_va): Formatting fix for
11351 ATTRIBUTE_GCC_DIAG.
11352 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11354 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11355 * substring-locations.c: Include intl.h.
11356 (format_warning_va): Turned into small wrapper around
11357 format_warning_n_va, renamed to ...
11358 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11359 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11361 (format_warning_at_substring_n): New function.
11362 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11363 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11364 format_warning_at_substring with just a shorter name instead of
11365 const function pointer.
11366 (fmtwarn_n): New function.
11367 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11368 appropriate, get rid of all the fmtstr temporaries, move conditionals
11369 with G_() wrapped string literals directly into fmtwarn arguments,
11370 cast dir.len to (int), formatting fixes.
11372 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11374 * doc/invoke.texi: Remove "Cilk Plus" references.
11376 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11377 Richard Biener <rguenther@suse.de>
11380 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11381 for error or warning attributes if CALL_FROM_THUNK_P is set.
11384 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11387 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11388 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11391 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11392 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11394 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11395 instead of -1U in last predictors element's probability member.
11397 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11400 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11401 arguments if they are comparable.
11403 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11405 PR tree-optimization/84634
11406 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11407 masks and masked_loop_p with a single loop_masks, making sure it's
11408 null for bb vectorization.
11410 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11412 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11413 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11414 loop->force_vectorize to check whether there is no alias.
11416 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11419 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11421 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11423 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11424 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11425 instead of a loop around prev_real_insn.
11426 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11429 PR inline-asm/84625
11430 * config/i386/i386.c (ix86_print_operand): Use conditional
11431 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11434 2018-03-02 Richard Biener <rguenther@suse.de>
11436 PR tree-optimization/84427
11437 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11438 (bitmap_set_subtract_values): Rewrite to handle multiple
11441 (prune_clobbered_mems): Likewise.
11442 (phi_translate): Take edge instead of pred/phiblock.
11443 (phi_translate_1): Likewise.
11444 (phi_translate_set): Likewise. Insert all translated
11445 exprs for a value into the set, keeping possibly multiple
11446 expressions per value.
11447 (compute_antic_aux): Adjust for phi_translate changes.
11448 When intersecting union the expressions and prune those
11449 not in the final value set, keeping possibly multiple
11450 expressions per value. Do not use value-insertion
11451 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11452 all expressions. Add verification that the value-sets
11453 only shrink during iteration.
11454 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11455 (do_pre_regular_insertion): Likewise.
11456 (do_pre_partial_partial_insertion): Likewise.
11458 2018-03-02 Richard Biener <rguenther@suse.de>
11461 * config/darwin.c (saved_debug_info_level): New static global.
11462 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11463 (darwin_asm_lto_end): Restore debug info generation settings.
11465 2018-03-01 Martin Liska <mliska@suse.cz>
11468 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11469 volatile arguments.
11471 2018-03-01 Richard Biener <rguenther@suse.de>
11474 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11475 type annotation with LTO when debug was disabled at compile-time.
11477 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11479 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11481 (mips_final_postscan_insn): Likewise.
11483 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11485 PR rtl-optimization/84528
11486 * alias.c (init_alias_target): Add commentary.
11487 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11488 a unique base value if the frame pointer is not eliminated
11489 to the stack pointer.
11491 2018-03-01 Tom de Vries <tom@codesourcery.com>
11493 PR rtl-optimization/83327
11494 * lra-int.h (hard_regs_spilled_into): Declare.
11495 * lra.c (hard_regs_spilled_into): Define.
11496 (init_reg_info): Init hard_regs_spilled_into.
11497 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11498 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11499 (process_bb_lives): Handle hard_regs_spilled_into.
11500 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11501 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11503 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11505 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11506 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11507 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11508 * config/rs6000/aix72.h: New file.
11510 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11512 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11513 instead of warning_at with conditional singular and plural messages
11517 * stor-layout.c (update_alignment_for_field): For
11518 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11519 && !DECL_PACKED (field), do the alignment update, just use
11520 only desired_align instead of MAX (type_align, desired_align)
11522 (place_field): Don't do known_align < desired_align handling
11523 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11524 is non-NULL, instead do it after rli->prev_field handling and
11525 only if not within a bitfield word. For DECL_PACKED (field)
11526 use type_align of BITS_PER_UNIT.
11528 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11530 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11531 superfluous parentheses and trailing spaces.
11533 2018-02-28 Richard Biener <rguenther@suse.de>
11535 PR tree-optimization/84584
11536 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11537 SCoPs with fake exit edge.
11539 2018-02-28 Martin Liska <mliska@suse.cz>
11542 * timevar.c (timer::print): Fix format to properly print 100%
11545 2018-02-28 Richard Biener <rguenther@suse.de>
11547 PR middle-end/84607
11548 * genmatch.c (capture_info::walk_match): Do not mark
11549 captured expressions without operands as expr_p given
11550 they act more like predicates and should be subject to
11551 "lost tail" side-effect preserving.
11553 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11555 PR rtl-optimization/81611
11556 * auto-inc-dec.c (attempt_change): Move dead note from
11557 mem_insn if it's the next use of regno
11558 (find_address): Take address use of reg holding
11559 non-incremented value. Add parm to limit search to the named
11561 (merge_in_block): Attempt to use a mem insn that is the next
11562 use of the original regno.
11564 2018-02-27 Martin Sebor <msebor@redhat.com>
11567 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11568 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11570 2018-02-27 Martin Sebor <msebor@redhat.com>
11572 PR translation/84207
11573 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11574 n argument to unsigned HOST_WIDE_INT.
11575 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11576 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11577 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11578 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11580 2018-02-27 Richard Biener <rguenther@suse.de>
11582 PR tree-optimization/84512
11583 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11584 Do not use the estimate returned from record_stmt_cost for
11585 the scalar iteration cost but sum properly using add_stmt_cost.
11587 2018-02-27 Richard Biener <rguenther@suse.de>
11589 PR tree-optimization/84466
11590 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11591 Adjust last change to less strictly validate use operands.
11593 2018-02-27 Martin Liska <mliska@suse.cz>
11595 PR gcov-profile/84548
11596 * gcov.c (process_file): Allow partial overlap and consider it
11597 also as group functions.
11598 (output_lines): Properly calculate range of lines for a group.
11600 2018-02-27 Martin Liska <mliska@suse.cz>
11602 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11603 'ggc' suffixes. Change first column width.
11604 (timer::print): Fix formatting of the column.
11606 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11608 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11609 preserve inline entry blocks for the sake of debug inline
11610 entry point markers alone.
11611 (remove_unused_locals): Suggest in comments a better place to
11612 force the preservation of inline entry blocks that are
11613 otherwise unused, but do not preserve them.
11615 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11617 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11619 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11622 * config/i386/constraints.md (Bs): Replace
11623 ix86_indirect_branch_register with
11624 TARGET_INDIRECT_BRANCH_REGISTER.
11626 * config/i386/i386.md (indirect_jump): Likewise.
11627 (tablejump): Likewise.
11628 (*sibcall_memory): Likewise.
11629 (*sibcall_value_memory): Likewise.
11630 Peepholes of indirect call and jump via memory: Likewise.
11631 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11632 (*sibcall_value_GOT_32): Likewise.
11633 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11634 (GOT_memory_operand): Likewise.
11635 (call_insn_operand): Likewise.
11636 (sibcall_insn_operand): Likewise.
11637 (GOT32_symbol_operand): Likewise.
11638 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11640 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11642 PR rtl-optimization/83496
11643 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11644 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11645 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11646 redundant insn, if any.
11647 (relax_delay_slots): Likewise.
11648 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11650 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11652 PR tree-optimization/83965
11653 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11654 that grouped statements are part of a reduction chain. Return
11655 true if the statement is not marked as a reduction itself but
11656 is part of a group.
11657 (vect_recog_dot_prod_pattern): Don't check whether the statement
11658 is part of a group here.
11659 (vect_recog_sad_pattern): Likewise.
11660 (vect_recog_widen_sum_pattern): Likewise.
11662 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11665 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11667 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11670 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11672 (ix86_output_indirect_function_return): New prototype.
11673 (ix86_split_simple_return_pop_internal): Likewise.
11674 * config/i386/i386.c (indirect_return_via_cx): New.
11675 (indirect_return_via_cx_bnd): Likewise.
11676 (indirect_thunk_name): Handle return va CX_REG.
11677 (output_indirect_thunk_function): Create alias for
11678 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11679 (ix86_output_indirect_jmp): Remove the bool argument.
11680 (ix86_output_indirect_function_return): New function.
11681 (ix86_split_simple_return_pop_internal): Likewise.
11682 * config/i386/i386.md (*indirect_jump): Don't pass false
11683 to ix86_output_indirect_jmp.
11684 (*tablejump_1): Likewise.
11685 (simple_return_pop_internal): Change it to define_insn_and_split.
11686 Call ix86_split_simple_return_pop_internal to split it for
11687 -mfunction-return=.
11688 (simple_return_indirect_internal): Call
11689 ix86_output_indirect_function_return instead of
11690 ix86_output_indirect_jmp.
11692 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11695 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11696 memset and value initialization afterwards.
11698 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11700 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11702 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11705 * common/config/aarch64/aarch64-common.c
11706 (aarch_option_optimization_table[]): Switch
11707 off fomit-frame-pointer
11709 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11710 Chung-Ju Wu <jasonwucj@gmail.com>
11712 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11714 (store_multiple): Ditto.
11716 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11718 * config.gcc: Add --with-cpu support for nds32 target.
11719 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11720 * config/nds32/nds32.opt: Add -mcpu= option.
11722 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11724 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11725 isel=yes): Warn for these deprecated options.
11727 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11729 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11730 ISA_2_5_MASKS_EMBEDDED.
11732 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11734 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11735 p->max as pointers rather than using iterative_hash_expr.
11737 2018-02-23 Carl Love <cel@us.ibm.com>
11739 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11740 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11742 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11743 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11744 P8V_BUILTIN_VEC_VUNSIGNED2.
11746 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11749 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11750 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11752 * lra-constraints.c (curr_insn_transform): Set up
11753 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11755 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11756 * lra-eliminations.c (spill_pseudos): Ditto.
11757 (process_insn_for_elimination): Ditto.
11758 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11759 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11760 LRA_NON_CLOBBERED_ALT.
11762 2018-02-22 Martin Sebor <msebor@redhat.com>
11764 PR tree-optimization/84480
11765 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11766 to maybe_diag_stxncpy_trunc. Call it.
11767 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11768 from gimple_fold_builtin_strcpy. Print inlining stack.
11769 (handle_builtin_stxncpy): Print inlining stack.
11770 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11772 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11775 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11776 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11777 and -fcheck-pointer-bounds are used together.
11778 (indirect_thunk_prefix): New enum.
11779 (indirect_thunk_need_prefix): New function.
11780 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11781 "_nt" instead of "_bnd" for NOTRACK prefix.
11782 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11783 (output_indirect_thunk_function): Likewise.
11785 (ix86_code_end): Update output_indirect_thunk_function calls.
11786 (ix86_output_indirect_branch_via_reg): Replace
11787 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11788 (ix86_output_indirect_branch_via_push): Likewise.
11789 (ix86_output_function_return): Likewise.
11790 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11791 incompatible with -fcf-protection=branch and
11792 -fcheck-pointer-bounds.
11794 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11797 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11798 Change gcc_assert call to output_operand_lossage.
11800 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11802 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11804 2018-02-22 DJ Delorie <dj@redhat.com>
11805 Sebastian Perta <sebastian.perta@renesas.com>
11806 Oleg Endo <olegendo@gcc.gnu.org>
11808 * config/rx/rx.c (rx_rtx_costs): New function.
11809 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11811 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11813 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11815 2018-02-22 Martin Liska <mliska@suse.cz>
11818 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11819 Add "native" as a possible value.
11821 2018-02-22 Martin Liska <mliska@suse.cz>
11824 * config/i386/i386.c (ix86_option_override_internal):
11825 Add "native" as a possible value for -march and -mtune.
11827 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11830 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11831 to all type variants.
11833 PR tree-optimization/84503
11834 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11835 width as info->bitpos + info->bitsize - start.
11836 (merged_store_group::merge_overlapping): Simplify width computation.
11837 (check_no_overlap): New function.
11838 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11839 start + width and last_order of the group, fail if check_no_overlap
11841 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11842 to group if check_no_overlap fails.
11844 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11846 * config/rs6000/altivec.md: Delete contraint arguments to
11847 define_expand, define_split, and define_peephole2, and in
11848 define_insn_and_split if always unused.
11849 * config/rs6000/darwin.md: Ditto.
11850 * config/rs6000/dfp.md: Ditto.
11851 * config/rs6000/rs6000.md: Ditto.
11852 * config/rs6000/sync.md: Ditto.
11853 * config/rs6000/vector.md: Ditto.
11854 * config/rs6000/vsx.md: Ditto.
11856 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11858 * config/rs6000/altivec.md: Write output control strings as braced
11859 blocks instead of double-quoted strings.
11860 * config/rs6000/darwin.md: Ditto.
11861 * config/rs6000/rs6000.md: Ditto.
11862 * config/rs6000/vector.md: Ditto.
11863 * config/rs6000/vsx.md: Ditto.
11865 2018-02-21 Jason Merrill <jason@redhat.com>
11867 PR c++/84314 - ICE with templates and fastcall attribute.
11868 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11870 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11872 * ipa-cp.c (determine_versionability): Fix comment typos.
11874 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11877 * ipa-cp.c (determine_versionability): Do not version functions caling
11880 2018-02-21 Martin Liska <mliska@suse.cz>
11883 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11884 Add "native" as a possible value.
11885 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11886 the macro when native cpu detection is available.
11888 2018-02-21 Martin Liska <mliska@suse.cz>
11891 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11892 Add "native" as a possible value.
11893 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11894 when native cpu detection is available.
11896 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11897 Martin Sebor <msebor@redhat.com>
11899 PR tree-optimization/84478
11900 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11902 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11903 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11904 support which is conservatively correct, for 2 only stay conservative
11905 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11906 argument to the 2 argument get_range_strlen, adjust 6 arg
11907 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11909 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11910 (gimple_fold_builtin_strlen): Pass true as last argument to
11913 2018-02-20 Martin Sebor <msebor@redhat.com>
11915 PR middle-end/84095
11916 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11917 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11918 (builtin_memref::builtin_memref): Factor out parts into
11919 set_base_and_offset and call it.
11921 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11923 PR middle-end/84406
11924 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11925 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11926 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11927 search at the associated MODE_INT.
11929 2018-02-20 Jeff Law <law@redhat.com>
11931 PR middle-end/82123
11932 PR tree-optimization/81592
11933 PR middle-end/79257
11934 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11935 for range data rather than using global data.
11936 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11937 range data rather than using global data.
11938 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11939 pass it to children as needed.
11940 (struct directive::fmtresult): Similarly.
11941 (struct directive::set_width): Similarly.
11942 (struct directive::set_precision): Similarly.
11943 (format_integer, format_directive, parse_directive): Similarly.
11944 (format_none): Accept unnamed vr_values parameter.
11945 (format_percent, format_floating, format_character): Similarly.
11946 (format_string, format_plain): Similarly.
11947 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11948 the EVRP range analyzer for range data rather than using global data.
11949 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11950 gimple-ssa-evrp-analyze.h
11951 (class sprintf_dom_walker): Add after_dom_children member function.
11952 Add evrp_range_analyzer member.
11953 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11954 range analyzer as needed.
11955 (sprintf_dom_walker::after_dom_children): New member function.
11956 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11958 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11959 (evrp_range_analyzer::pop_to_marker): Likewise.
11961 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11963 PR tree-optimization/84419
11964 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11965 with the required type if its current type is compatible but
11968 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11970 PR middle-end/82004
11971 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11972 after vectorization.
11974 2018-02-20 Martin Liska <mliska@suse.cz>
11977 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11978 possible values if we don't have a hint.
11980 2018-02-20 Martin Liska <mliska@suse.cz>
11984 * final.c (shorten_branches): Build align_tab array with one
11986 * opts.c (finish_options): Add alignment option limit check.
11987 (MAX_CODE_ALIGN): Likewise.
11988 (MAX_CODE_ALIGN_VALUE): Likewise.
11989 * doc/invoke.texi: Document maximum allowed option value for
11990 all -falign-* options.
11992 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11995 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11996 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11997 * var-tracking.c (emit_note_insn_var_location): Remove all references
11998 to NOTE_INSN_CALL_ARG_LOCATION.
11999 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12000 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12001 Use copy_rtx_if_shared.
12002 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12003 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12004 (dwarf2out_var_location): Remove handling of
12005 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12007 * final.c (final_scan_insn): Remove all references to
12008 NOTE_INSN_CALL_ARG_LOCATION.
12009 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12010 before dumping final insns.
12011 * except.c (emit_note_eh_region_end): Remove all references to
12012 NOTE_INSN_CALL_ARG_LOCATION.
12013 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12014 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12015 * config/arc/arc.c (hwloop_optimize): Likewise.
12016 * config/arm/arm.c (create_fix_barrier): Likewise.
12017 * config/s390/s390.c (s390_chunkify_start): Likewise.
12018 * config/sh/sh.c (find_barrier): Likewise.
12019 * config/i386/i386.c (rest_of_insert_endbranch,
12020 ix86_seh_fixup_eh_fallthru): Likewise.
12021 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12022 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12023 * config/frv/frv.c (frv_function_prologue): Likewise.
12024 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12026 (note_outside_basic_block_p): Remove all references to
12027 NOTE_INSN_CALL_ARG_LOCATION.
12028 * gengtype.c (adjust_field_rtx_def): Likewise.
12029 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12031 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12032 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12035 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12038 PR tree-optimization/84452
12039 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12040 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12043 2018-02-19 Martin Liska <mliska@suse.cz>
12046 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12048 2018-02-19 Martin Liska <mliska@suse.cz>
12049 Richard Sandiford <richard.sandiford@linaro.org>
12051 PR tree-optimization/82491
12052 * gimple-fold.c (get_base_constructor): Make earlier bail out
12055 2018-02-19 Carl Love <cel@us.ibm.com>
12057 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12058 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12060 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12061 P8V_BUILTIN_VEC_NEG.
12063 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12065 * config/rl78/rl78.md (movdf): New define expand.
12067 2018-02-19 Martin Liska <mliska@suse.cz>
12070 * doc/invoke.texi: Fix typo.
12071 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12073 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12075 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12076 handle rs6000_single_float and rs6000_double_float specially for
12079 2018-02-16 Jeff Law <law@redhat.com>
12081 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12082 (pop_regs): Use it.
12084 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12087 * ipa-inline.c (inline_small_functions): Fix a typo.
12089 2018-02-16 Nathan Sidwell <nathan@acm.org>
12091 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12093 2018-02-16 Carl Love <cel@us.ibm.com>
12095 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12096 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12097 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12098 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12099 expansion to P8V_BUILTIN_VEC_FLOAT2.
12101 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12103 PR rtl-optimization/70023
12104 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12105 src_regno into account.
12107 2018-02-16 Carl Love <cel@us.ibm.com>
12109 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12110 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12111 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12112 * config/rs6000/rs6000.c: Remove case statements for
12113 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12114 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12115 and P9V_BUILTIN_VEC_VINSERT4B.
12116 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12117 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12118 * config/rs6000/vsx.md:
12119 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12122 2018-02-16 Carl Love <cel@us.ibm.com>
12124 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12126 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12128 * config/rs6000/rs6000-c.c: Add the definitions for
12129 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12130 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12131 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12132 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12133 definition for insert4b and define insn *insert3b_internal.
12134 * doc/extend.texi: Add documentation for vec_extract4b.
12136 2018-02-16 Nathan Sidwell <nathan@acm.org>
12138 * doc/extend.texi (Backwards Compatibility): Mention friend
12139 injection. Note for-scope is deprecated.
12140 * doc/invoke.texi (-ffriend-injection): Deprecate.
12142 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12144 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12145 that moved to I2, also allow destinations that are a paradoxical
12146 subreg (instead of a normal reg).
12148 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12151 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12154 2018-02-16 Richard Biener <rguenther@suse.de>
12156 PR tree-optimization/84037
12157 PR tree-optimization/84016
12159 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12160 Adjust vec_construct for the fact we need additional higher latency
12161 128bit inserts for AVX256 and AVX512 vector builds.
12162 (ix86_add_stmt_cost): Scale vector construction cost for
12165 2018-02-16 Richard Biener <rguenther@suse.de>
12167 PR tree-optimization/84417
12168 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12169 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12170 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12172 2018-02-16 Martin Liska <mliska@suse.cz>
12175 * internal-fn.def (ASAN_CHECK): Set proper flags.
12176 (ASAN_MARK): Likewise.
12178 2018-02-16 Julia Koval <julia.koval@intel.com>
12180 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12181 from PTA_CANNONLAKE.
12183 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12186 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12187 Use ++iter rather than iter++ for std::list iterators.
12188 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12189 defer deleting them until all nodes in the forest are processed. Do
12190 free even leaf nodes. Change to_process into auto_vec.
12193 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12194 * vec.h (vec_default_construct): Use memset instead of placement new
12195 if BROKEN_VALUE_INITIALIZATION is defined.
12196 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12197 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12200 PR rtl-optimization/83723
12201 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12202 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12203 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12205 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12207 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12209 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12211 PR rtl-optimization/81443
12212 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12213 from inner REGs to paradoxical SUBREGs.
12215 2018-02-16 Richard Biener <rguenther@suse.de>
12217 PR tree-optimization/84399
12218 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12219 For operands we can analyze at their definition make sure we can
12220 analyze them at each use as well.
12222 2018-02-16 Richard Biener <rguenther@suse.de>
12224 PR tree-optimization/84190
12225 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12226 volatile accesses if the decl isn't volatile.
12228 2018-02-15 Jason Merrill <jason@redhat.com>
12230 PR c++/84314 - ICE with templates and fastcall attribute.
12231 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12232 TYPE_CANONICAL on an existing type.
12234 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12236 PR tree-optimization/84383
12237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12238 dstoff nor call operand_equal_p if dstbase is NULL.
12240 PR tree-optimization/84334
12241 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12242 also a CONSTANT_CLASS_P, punt.
12244 2018-02-14 Jim Wilson <jimw@sifive.com>
12246 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12247 first SMALL_OPERAND check. New local min_second_step. Move assert
12248 to where locals are set. Add TARGET_RVC support.
12249 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12251 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12253 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12255 2018-02-14 Martin Sebor <msebor@redhat.com>
12257 PR tree-optimization/83698
12258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12259 arrays constrain the offset range to their bounds.
12260 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12261 (builtin_access::overlap): Avoid setting the size of overlap if it's
12263 (maybe_diag_overlap): Also consider arrays when deciding what values
12264 of offsets to include in diagnostics.
12266 2018-02-14 Martin Sebor <msebor@redhat.com>
12269 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12270 that correspond to the kind of a declaration.
12272 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12275 * config/pa/pa.md: Load address of PIC label using the linkage table
12276 if the label is nonlocal.
12278 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12280 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12281 warning message if user requests -maltivec=be.
12282 * doc/invoke.texi: Document deprecation of -maltivec=be.
12284 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12287 * config/rs6000/rs6000-c.c: Update definitions for
12288 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12289 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12291 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12294 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12295 add _get_ssp intrinsics. Remove argument from
12296 __builtin_ia32_rdssp[d|q].
12297 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12298 * config/i386/i386-builtin.def: Remove argument from
12299 __builtin_ia32_rdssp[d|q].
12300 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12301 ix86_expand_special_args_builtin for _rdssp[d|q].
12302 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12303 Clear register before usage.
12304 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12305 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12307 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12309 PR tree-optimization/84357
12310 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12311 operand 1 of an ARRAY_REF too.
12313 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12316 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12317 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12319 (set_of_reg): New struct.
12320 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12321 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12322 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12324 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12325 Split into bitclr, bitset, bitinvert patterns if appropriate.
12326 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12327 use rx_fuse_in_memory_bitop.
12328 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12329 to named insn, correct maximum insn length.
12331 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12334 * machmode.def: Define a complex mode for PARTIAL_INT.
12335 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12337 * doc/rtl.texi: Document CSPImode.
12338 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12340 (msp430_hard_regno_nregs_with_padding): Likewise.
12342 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12345 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12347 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12349 PR rtl-optimization/84169
12350 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12351 we generated a parallel as new i3 and we split that to new i2 and i3
12352 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12353 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12354 those to i2, not i1. Partially rewrite this scan code.
12356 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12359 * stor-layout.c (place_field): For variable length fields, adjust
12360 offset_align afterwards not just based on the field's alignment,
12361 but also on the size.
12363 PR middle-end/84309
12364 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12365 of exps and logs in the use_exp2 case.
12367 2018-02-13 Jeff Law <law@redhat.com>
12369 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12370 entry for "vector".
12372 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12375 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12379 * common.opt (gas-loc-support, gas-locview-support): New.
12380 (ginline-points, ginternal-reset-location-views): New.
12381 * doc/invoke.texi: Document them. Use @itemx where intended.
12382 (gvariable-location-views): Adjust.
12383 * target.def (reset_location_view): New.
12384 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12385 (TARGET_RESET_LOCATION_VIEW): New.
12386 * doc/tm.texi: Rebuilt.
12387 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12388 (dwarf2out_default_as_locview_support): New.
12389 (output_asm_line_debug_info): Use option variables.
12390 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12391 (output_loc_list): Likewise.
12392 (add_high_low_attributes): Check option variables.
12393 Don't output entry view attribute in strict mode.
12394 (gen_inlined_subroutine_die): Check option variables.
12395 (dwarf2out_inline_entry): Likewise.
12396 (init_sections_and_labels): Likewise.
12397 (dwarf2out_early_finish): Likewise.
12398 (maybe_reset_location_view): New, from...
12399 (dwarf2out_var_location): ... here. Call it.
12400 * debug.h (dwarf2out_default_as_loc_support): Declare.
12401 (dwarf2out_default_as_locview_support): Declare.
12402 * hooks.c (hook_int_rtx_insn_0): New.
12403 * hooks.h (hook_int_rtx_insn_0): Declare.
12404 * toplev.c (process_options): Take -gas-loc-support and
12405 -gas-locview-support from dwarf2out. Enable
12406 -gvariable-location-views by default only with locview
12407 assembler support. Enable -ginternal-reset-location-views by
12408 default only if the target defines the corresponding hook.
12409 Enable -ginline-points by default if location views are
12410 enabled; force it disabled if statement frontiers are
12412 * tree-inline.c (expand_call_inline): Check option variables.
12413 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12415 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12417 PR tree-optimization/84321
12418 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12419 handling. Also check whether the anti-range contains any values
12420 that satisfy the mask; switch to a VR_RANGE if not.
12422 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12425 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12427 2018-02-13 Martin Jambor <mjambor@suse.cz>
12430 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12431 of call statements, also set location of a load to a temporary.
12433 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12435 * config/rl78/rl78.c (add_vector_labels): New function.
12436 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12437 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12438 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12439 which checks that no arguments are passed.
12440 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12441 * doc/extend.texi: Documentation for the new attribute.
12443 2018-02-13 Andreas Schwab <schwab@suse.de>
12445 * config/riscv/linux.h (CPP_SPEC): Define.
12447 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12450 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12451 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12452 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12453 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12454 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12455 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12456 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12457 temporarily for AES and PCLMUL builtins.
12459 PR tree-optimization/84339
12460 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12461 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12464 PR middle-end/84309
12465 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12466 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12467 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12469 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12471 * omp-simd-clone.h: New file.
12472 * omp-simd-clone.c: Include omp-simd-clone.h.
12473 (expand_simd_clones): No longer static.
12474 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12475 cgraph.h and omp-simd-clone.h.
12476 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12477 (vect_recog_widen_shift_pattern): Formatting fix.
12478 (vect_pattern_recog_1): Don't check optab for calls.
12481 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12482 operands[2] into a REG before using gen_lowpart on it.
12484 2018-02-12 Jeff Law <law@redhat.com>
12487 * config/sh/sh.c (find_barrier): Consider a sibling call
12490 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12491 successfully back substituting a reg.
12493 2018-02-12 Richard Biener <rguenther@suse.de>
12495 PR tree-optimization/84037
12496 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12497 parameter, move visited init to caller.
12498 (vect_slp_analyze_operations): Separate cost from validity
12499 check, initialize visited once for all instances.
12500 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12502 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12504 (vect_model_promotion_demotion_cost): Likewise.
12505 (vectorizable_bswap): Guard cost modeling with !slp_node
12506 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12508 (vectorizable_call): Likewise.
12509 (vectorizable_conversion): Likewise.
12510 (vectorizable_assignment): Likewise.
12511 (vectorizable_shift): Likewise.
12512 (vectorizable_operation): Likewise.
12513 (vectorizable_store): Likewise.
12514 (vectorizable_load): Likewise.
12515 (vectorizable_condition): Likewise.
12516 (vectorizable_comparison): Likewise.
12518 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12521 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12522 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12523 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12525 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12527 PR middle-end/83665
12528 * params.def (inline-min-speedup): Increase from 8 to 15.
12529 (max-inline-insns-auto): Decrease from 40 to 30.
12530 * ipa-split.c (consider_split): Add some buffer for function to
12531 be considered inlining candidate.
12532 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12535 2018-02-12 Richard Biener <rguenther@suse.de>
12537 PR tree-optimization/84037
12538 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12539 matched stmts if we cannot swap the non-matched ones.
12541 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12543 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12544 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12545 _mm_maskz_scalef_round_ss): New intrinsics.
12546 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12547 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12548 __builtin_ia32_scalefss_round): Remove.
12549 (__builtin_ia32_scalefsd_mask_round,
12550 __builtin_ia32_scalefss_mask_round): New intrinsics.
12551 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12552 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12553 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12554 "<round_constraint>")): Changed to ...
12555 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12556 "<round_scalar_constraint>")): ... this.
12557 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12558 %0, %1, %2<round_op3>}"): Changed to ...
12559 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12560 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12561 %2<round_scalar_mask_op3>}"): ... this.
12562 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12564 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12566 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12567 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12568 (_mm_maskz_sqrt_round_ss): New intrinsics.
12569 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12570 (__builtin_ia32_sqrtsd_mask_round)
12571 (__builtin_ia32_sqrtss_mask_round): New builtins.
12572 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12573 (__builtin_ia32_sqrtss_round): Remove.
12574 (__builtin_ia32_sqrtsd_mask_round)
12575 (__builtin_ia32_sqrtss_mask_round): New builtins.
12576 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12577 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12578 ((match_operand:VF_128 1 "vector_operand"
12579 "xBm,<round_constraint>")): Changed to ...
12580 ((match_operand:VF_128 1 "vector_operand"
12581 "xBm,<round_scalar_constraint>")): ... this.
12582 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12583 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12584 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12585 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12586 %<iptr>1<round_scalar_mask_op3>}): ... this.
12587 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12588 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12590 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12593 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12594 Cast vec_cmpeq result to correct type.
12595 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12596 Cast vec_cmpgt result to correct type.
12598 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12600 * final.c (final_scan_insn_1): Renamed from...
12601 (final_scan_insn): ... this. New wrapper, to recover
12602 seen from the outermost call in recursive ones.
12603 * config/sparc/sparc.c (output_return): Drop seen from call.
12604 (output_sibcall): Likewise.
12605 * config/visium/visium.c (output_branch): Likewise.
12607 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12609 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12612 2018-02-10 Alan Modra <amodra@gmail.com>
12615 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12616 Specify LR as an input.
12618 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12621 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12622 remove_member_access_dummy_vars): New functions.
12623 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12624 lower_omp_1, execute_lower_omp): Use them.
12626 PR rtl-optimization/84308
12627 * shrink-wrap.c (spread_components): Release todo vector.
12629 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12631 PR rtl-optimization/57193
12632 * ira-color.c (struct allocno_color_data): Add member
12633 conflict_allocno_hard_prefs.
12634 (update_conflict_allocno_hard_prefs): New.
12635 (bucket_allocno_compare_func): Add a preference based on
12636 conflict_allocno_hard_prefs.
12637 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12638 (color_allocnos): Remove a dead code. Initiate
12639 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12641 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12644 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12645 constraint from =wa to wa. Avoid a subreg on the output operand,
12646 instead use a pseudo and subreg it in a move.
12647 (p9_xxbrd_<mode>): Changed to ...
12648 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12649 (p9_xxbrd_v2df): New expander.
12650 (p9_xxbrw_<mode>): Changed to ...
12651 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12652 (p9_xxbrw_v4sf): New expander.
12654 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12656 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12658 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12661 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12662 multiply in 32-bit mode.
12663 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12664 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12667 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12669 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12670 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12671 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12672 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12674 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12677 * dwarf2out.c (is_trivial_indirect_ref): New function.
12678 (dwarf2out_late_global_decl): Do not generate a location
12679 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12680 and that are not defined in the current unit.
12682 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12684 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12685 instead of a libcall for UNORDERED.
12687 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12690 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12691 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12693 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12696 * config/s390/s390.c (s390_set_current_function): Invoke
12697 s390_indirect_branch_settings also if fndecl didn't change.
12699 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12701 * config/rs6000/rs6000.md (blockage): Set length to zero.
12703 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12705 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12707 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12710 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12711 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12715 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12716 PARALLEL incoming that failed vt_get_decl_and_offset check.
12718 PR middle-end/84237
12719 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12720 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12722 (get_variable_section): For decls in named .bss* sections pass true as
12723 second argument to bss_initializer_p.
12725 2018-02-09 Marek Polacek <polacek@redhat.com>
12726 Jakub Jelinek <jakub@redhat.com>
12729 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12730 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12731 Sync some changes from cxx_fold_indirect_ref.
12733 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12735 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12737 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12738 (BLOCK_INLINE_ENTRY_LABEL): New.
12739 (dwarf2out_var_location): Disregard inline entry markers.
12740 (inline_entry_data): New struct.
12741 (inline_entry_data_hasher): New hashtable type.
12742 (inline_entry_data_hasher::hash): New.
12743 (inline_entry_data_hasher::equal): New.
12744 (inline_entry_data_table): New variable.
12745 (add_high_low_attributes): Add DW_AT_entry_pc and
12746 DW_AT_GNU_entry_view attributes if a pending entry is found
12747 in inline_entry_data_table. Add old entry_pc attribute only
12748 if debug nonbinding markers are disabled.
12749 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12750 markers are enabled.
12751 (block_within_block_p, dwarf2out_inline_entry): New.
12752 (dwarf2out_finish): Check that no entries remained in
12753 inline_entry_data_table.
12754 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12755 (final_scan_insn, notice_source_line): Likewise.
12756 (rest_of_clean_state): Skip inline entry markers.
12757 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12759 * gimple.c (gimple_build_debug_inline_entry): New.
12760 * gimple.h (enum gimple_debug_subcode): Add
12761 GIMPLE_DEBUG_INLINE_ENTRY.
12762 (gimple_build_debug_inline_entry): Declare.
12763 (gimple_debug_inline_entry_p): New.
12764 (gimple_debug_nonbind_marker_p): Adjust.
12765 * insn-notes.def (INLINE_ENTRY): New.
12766 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12767 inline entry marker notes.
12768 (print_insn): Likewise.
12769 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12770 (INSN_DEBUG_MARKER_KIND): Likewise.
12771 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12772 * tree-inline.c (expand_call_inline): Build and insert
12773 debug_inline_entry stmt.
12774 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12775 inline entry blocks early, if nonbind markers are enabled.
12776 (dump_scope_block): Dump fragment info.
12777 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12778 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12779 (gimple_build_debug_inline_entry): New.
12780 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12781 Enable/disable inline entry points too.
12782 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12783 (DEBUG_INSN): Describe inline entry markers.
12785 * common.opt (gvariable-location-views): New.
12786 (gvariable-location-views=incompat5): New.
12787 * config.in: Rebuilt.
12788 * configure: Rebuilt.
12789 * configure.ac: Test assembler for view support.
12790 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12791 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12792 * dwarf2out.c (var_loc_view): New typedef.
12793 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12794 (dwarf2out_locviews_in_attribute): New.
12795 (dwarf2out_locviews_in_loclist): New.
12796 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12797 (enum dw_line_info_opcode): Add LI_adv_address.
12798 (struct dw_line_info_table): Add view.
12799 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12800 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12801 (zero_view_p): New variable.
12802 (ZERO_VIEW_P): New macro.
12803 (output_asm_line_debug_info): New.
12804 (struct var_loc_node): Add view.
12805 (add_AT_view_list, AT_loc_list): New.
12806 (add_var_loc_to_decl): Add view param. Test it against last.
12807 (new_loc_list): Add view params. Record them.
12808 (AT_loc_list_ptr): Handle loc and view lists.
12809 (view_list_to_loc_list_val_node): New.
12810 (print_dw_val): Handle dw_val_class_view_list.
12811 (size_of_die): Likewise.
12812 (value_format): Likewise.
12813 (loc_list_has_views): New.
12814 (gen_llsym): Set vl_symbol too.
12815 (maybe_gen_llsym, skip_loc_list_entry): New.
12816 (dwarf2out_maybe_output_loclist_view_pair): New.
12817 (output_loc_list): Output view list or entries too.
12818 (output_view_list_offset): New.
12819 (output_die): Handle dw_val_class_view_list.
12820 (output_dwarf_version): New.
12821 (output_compilation_unit_header): Use it.
12822 (output_skeleton_debug_sections): Likewise.
12823 (output_rnglists, output_line_info): Likewise.
12824 (output_pubnames, output_aranges): Update version comments.
12825 (output_one_line_info_table): Output view numbers in asm comments.
12826 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12827 Call maybe_gen_llsym.
12828 (loc_list_from_tree_1): Adjust.
12829 (add_AT_location_description): Create view list attribute if
12830 needed, check it's absent otherwise.
12831 (convert_cfa_to_fb_loc_list): Adjust.
12832 (maybe_emit_file): Call output_asm_line_debug_info for test.
12833 (dwarf2out_var_location): Reset views as needed. Precompute
12834 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12835 attribute. Set view.
12836 (new_line_info_table): Reset next view.
12837 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12838 (dwarf2out_source_line): Likewise. Output view resets and labels to
12839 the assembler, or select appropriate line info opcodes.
12840 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12841 (optimize_string_length): Catch it. Adjust.
12842 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12843 dw_val_class_view_list, and remove it if no longer needed.
12844 (hash_loc_list): Hash view numbers.
12845 (loc_list_hasher::equal): Compare them.
12846 (optimize_location_lists): Check whether a view list symbol is
12847 needed, and whether the locview attribute is present, and
12848 whether they match. Remove the locview attribute if no longer
12850 (index_location_lists): Call skip_loc_list_entry for test.
12851 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12852 Use output_dwarf_version.
12853 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12854 (struct dw_val_node): Add val_view_list.
12855 * final.c (SEEN_NEXT_VIEW): New.
12856 (set_next_view_needed): New.
12857 (clear_next_view_needed): New.
12858 (maybe_output_next_view): New.
12859 (final_start_function): Rename to...
12860 (final_start_function_1): ... this. Take pointer to FIRST,
12861 add SEEN parameter. Emit param bindings in the initial view.
12862 (final_start_function): Reintroduce SEEN-less interface.
12863 (final): Rename to...
12864 (final_1): ... this. Take SEEN parameter. Output final pending
12865 next view at the end.
12866 (final): Reintroduce seen-less interface.
12867 (final_scan_insn): Output pending next view before switching
12868 sections or ending a block. Mark the next view as needed when
12869 outputting variable locations. Notify debug backend of section
12870 changes, and of location view changes.
12871 (rest_of_handle_final): Adjust.
12872 * toplev.c (process_options): Autodetect value for debug variable
12873 location views option. Warn on incompat5 without -gdwarf-5.
12874 * doc/invoke.texi (gvariable-location-views): New.
12875 (gvariable-location-views=incompat5): New.
12876 (gno-variable-location-views): New.
12878 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12880 PR tree-optimization/84136
12881 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12882 that the result of find_edge is non-NULL.
12884 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12887 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12888 storing integer register in SImode. Fix cost of 256 and 512
12889 byte aligned SSE register store.
12891 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12893 * config/i386/i386.c (ix86_multiplication_cost): Fix
12894 multiplication cost for TARGET_AVX512DQ.
12896 2018-02-08 Marek Polacek <polacek@redhat.com>
12898 PR tree-optimization/84238
12899 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12902 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12904 PR tree-optimization/84265
12905 * tree-vect-stmts.c (vectorizable_store): Don't treat
12906 VMAT_CONTIGUOUS accesses as grouped.
12907 (vectorizable_load): Likewise.
12909 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12911 PR tree-optimization/81635
12912 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12913 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12914 (test_round_for_mask): New functions.
12915 (wide_int_cc_tests): Call test_round_for_mask.
12916 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12917 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12918 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12919 range returned by get_range_info.
12921 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12924 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12925 * symtab.c: Include builtins.h
12926 (symtab_node::output_to_lto_symbol_table_p): Move here
12927 from lto-streamer-out.c:output_symbol_p.
12928 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12929 (output_symbol_p): Move all logic to symtab.c
12930 (produce_symtab): Update.
12932 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12934 * config/s390/s390-opts.h (enum indirect_branch): Define.
12935 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12936 (s390_indirect_branch_via_thunk)
12937 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12938 (enum s390_indirect_branch_type): Define.
12939 * config/s390/s390.c (struct s390_frame_layout, struct
12940 machine_function): Remove.
12941 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12942 (indirect_branch_table_label_no, indirect_branch_table_name):
12944 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12945 (enum s390_indirect_branch_option): Define.
12946 (s390_return_addr_from_memory): New function.
12947 (s390_handle_string_attribute): New function.
12948 (s390_attribute_table): Add new attribute handler.
12949 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12950 (s390_indirect_branch_via_thunk): New function.
12951 (s390_indirect_branch_via_inline_thunk): New function.
12952 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12953 sibling call optimization for non z10 compiles.
12954 (s390_emit_call): Force indirect branch target to be a single
12955 register. Add r1 clobber for non-z10 compiles.
12956 (s390_emit_epilogue): Emit return jump via return_use expander.
12957 (s390_reorg): Handle JUMP_INSNs as execute targets.
12958 (s390_option_override_internal): Perform validity checks for the
12959 new command line options.
12960 (s390_indirect_branch_attrvalue): New function.
12961 (s390_indirect_branch_settings): New function.
12962 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12963 (s390_output_indirect_thunk_function): New function.
12964 (s390_code_end): Implement target hook.
12965 (s390_case_values_threshold): Implement target hook.
12966 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12968 * config/s390/s390.h (struct s390_frame_layout)
12969 (struct machine_function): Move here from s390.c.
12970 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12971 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12972 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12973 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12974 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12975 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12976 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12977 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12978 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12979 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12980 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12981 (mnemonic attribute): Add values which aren't recognized
12983 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12984 pattern for branch conversion. Fix mnemonic attribute.
12985 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12986 indirect branch via thunk if requested.
12987 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12988 ("*indirect_jump"): Disable for branch conversion using out of
12990 ("indirect_jump_via_thunk<mode>_z10")
12991 ("indirect_jump_via_thunk<mode>")
12992 ("indirect_jump_via_inlinethunk<mode>_z10")
12993 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12994 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12995 ("casesi_jump_via_inlinethunk<mode>_z10")
12996 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12997 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12998 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12999 ("*indirect2_jump"): Disable for branch conversion.
13000 ("casesi_jump"): Turn into expander and expand patterns for branch
13002 ("return_use"): New expander.
13003 ("*return"): Emit return via thunk and rename it to ...
13004 ("*return<mode>"): ... this one.
13005 * config/s390/s390.opt: Add new options and and enum for the
13008 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13010 * lra-constraints.c (match_reload): Unconditionally use
13011 gen_lowpart_SUBREG, rather than selecting between that
13012 and equivalent gen_rtx_SUBREG code.
13014 2018-02-08 Richard Biener <rguenther@suse.de>
13016 PR tree-optimization/84233
13017 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13018 changed flag instead of boguously re-using phi_inserted.
13020 2018-02-08 Martin Jambor <mjambor@suse.cz>
13022 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13023 static local variables.
13025 2018-02-08 Richard Biener <rguenther@suse.de>
13027 PR tree-optimization/84278
13028 * tree-vect-stmts.c (vectorizable_store): When looking for
13029 smaller vector types to perform grouped strided loads/stores
13030 make sure the mode is supported by the target.
13031 (vectorizable_load): Likewise.
13033 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13035 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13036 Increase LDP/STP opportunities by adding adjacent callee-saves.
13038 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13040 PR rtl-optimization/84068
13041 PR rtl-optimization/83459
13042 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13044 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13046 PR tree-optimization/84224
13047 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13048 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13049 non-zero arguments.
13051 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13054 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13055 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13056 count, remove one USE.
13058 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13060 * doc/install.texi (Configuration): Document the
13061 --with-long-double-format={ibm,ieee} PowerPC configuration
13065 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13066 Convert from define_expand to be define_insn_and_split. Rework
13067 float/double/_Float128 conversions to QI/HI/SImode to work with
13068 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13069 conversions to QI/HImode types did a store and then a load to
13070 truncate the value. For conversions to VSX registers, don't split
13071 the insn, instead emit the code directly. Use the code iterator
13072 any_fix to combine signed and unsigned conversions.
13073 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13074 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13075 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13076 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13077 (fix_<mode>di2_hw): Likewise.
13078 (fixuns_<mode>di2_hw): Likewise.
13079 (fix_<mode>si2_hw): Likewise.
13080 (fixuns_<mode>si2_hw): Likewise.
13081 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13082 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13083 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13084 fix<uns>_trunc<SFDF:mode>si2_p8.
13085 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13086 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13087 (fix<uns>_<mode>_mem): Likewise.
13088 (fctiw<u>z_<mode>_mem): Likewise.
13089 (fix<uns>_<mode>_mem): Likewise.
13090 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13091 the register allocator from doing a direct move to the GPRs to do
13092 a store, and instead use the ISA 3.0 store byte/half-word from
13093 vector register instruction. For IEEE 128-bit floating point,
13094 also optimize stores of 32-bit ints.
13095 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13097 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13099 * genextract.c (push_pathstr_operand): New function to support
13101 (walk_rtx): Call push_pathstr_operand.
13102 (print_path): Support [a-zA-Z].
13104 2018-02-07 Richard Biener <rguenther@suse.de>
13106 PR tree-optimization/84037
13107 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13108 (cse_and_gimplify_to_preheader): Declare.
13109 (vect_get_place_in_interleaving_chain): Likewise.
13110 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13112 (_loop_vec_info::~_loop_vec_info): Delete it.
13113 (cse_and_gimplify_to_preheader): New function.
13114 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13115 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13116 (vectorizable_load): Likewise. For grouped stores always base
13117 the IV on the first element.
13118 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13119 condition before gimplifying.
13121 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13123 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13124 *DIV_EXPR and *MOD_EXPR.
13126 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13129 * config/i386/i386.c (ix86_option_override_internal): Mask out
13130 the CF_SET bit when checking -fcf-protection.
13132 2018-02-07 Tom de Vries <tom@codesourcery.com>
13135 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13138 2018-02-07 Richard Biener <rguenther@suse.de>
13140 PR tree-optimization/84204
13141 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13144 PR tree-optimization/84205
13145 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13146 special-case isl_ast_op_zdiv_r.
13148 PR tree-optimization/84223
13149 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13150 Only add conditions from within the region.
13151 (gather_bbs::after_dom_children): Adjust.
13153 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13156 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13157 * config/avr/avr.md: Only post-reload split REG-REG moves if
13158 either register is GENERAL_REG_P.
13160 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13162 PR tree-optimization/84235
13163 * tree-ssa-scopedtables.c
13164 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13165 if the subtraction is performed in floating point type where NaNs are
13166 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13167 build 1. Formatting fix.
13169 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13172 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13173 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13174 and skip it regardless of bb boundaries. Use CALL_P macro,
13175 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13176 unnecessarily, formatting fix.
13178 2018-02-06 Michael Collison <michael.collison@arm.com>
13180 * config/arm/thumb2.md:
13181 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13182 (*thumb_mov_notscc): Ditto.
13184 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13187 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13188 unsigned_fix, not "s".
13190 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13192 * configure.ac (gcc_fn_eh_frame_ro): New function.
13193 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13194 correct .eh_frame permissions.
13195 * configure: Regenerate.
13197 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13199 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13200 irrelevant options.
13202 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13204 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13205 Display warning message for -mno-speculate-indirect-jumps.
13207 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13209 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13211 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13213 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13215 PR tree-optimization/84225
13216 * tree-eh.c (find_trapping_overflow): Only call
13217 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13219 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13222 * config/i386/i386.c: Reimplement the check of possible options
13223 -mibt/-mshstk conbination. Change error messages.
13224 * doc/invoke.texi: Fix a typo: remove extra '='.
13226 2018-02-06 Marek Polacek <polacek@redhat.com>
13228 PR tree-optimization/84228
13229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13231 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13234 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13235 emitted arch directives.
13236 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13237 __ARM_FEATURE_COPROC before changing architectures.
13239 2018-02-06 Richard Biener <rguenther@suse.de>
13241 * config/i386/i386.c (print_reg): Fix typo.
13242 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13244 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13246 * configure: Regenerate.
13248 2018-02-05 Martin Sebor <msebor@redhat.com>
13250 PR tree-optimization/83369
13251 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13254 2018-02-05 Martin Liska <mliska@suse.cz>
13256 * doc/invoke.texi: Cherry-pick upstream r323995.
13258 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13260 * ira.c (ira_init_register_move_cost): Adjust comment.
13262 2018-02-05 Martin Liska <mliska@suse.cz>
13264 PR gcov-profile/84137
13265 * doc/gcov.texi: Fix typo in documentation.
13267 2018-02-05 Martin Liska <mliska@suse.cz>
13269 PR gcov-profile/83879
13270 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13271 using dlopen functionality.
13273 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13275 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13276 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13277 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13278 _mm_maskz_range_round_ss): New intrinsics.
13279 (__builtin_ia32_rangesd128_round)
13280 (__builtin_ia32_rangess128_round): Remove.
13281 (__builtin_ia32_rangesd128_mask_round,
13282 __builtin_ia32_rangess128_mask_round): New builtins.
13283 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13284 __builtin_ia32_rangess128_round): Remove.
13285 (__builtin_ia32_rangesd128_mask_round,
13286 __builtin_ia32_rangess128_mask_round): New builtins.
13287 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13288 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13289 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13290 "<round_saeonly_constraint>")): Changed to ...
13291 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13292 "<round_saeonly_scalar_constraint>")): ... this.
13293 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13294 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13295 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13296 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13297 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13299 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13301 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13303 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13304 Remove all values except native, 8540 and 8548.
13306 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13308 * config/i386/i386.c (ix86_output_function_return): Pass
13309 INVALID_REGNUM, instead of -1, as invalid register number to
13310 indirect_thunk_name and output_indirect_thunk.
13312 2018-02-02 Julia Koval <julia.koval@intel.com>
13314 * config.gcc: Add -march=icelake.
13315 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13316 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13317 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13318 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13319 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13320 (processor_target_table): Add icelake.
13321 (ix86_option_override_internal): Handle new PTAs.
13322 (get_builtin_code_for_version): Handle icelake.
13323 (M_INTEL_COREI7_ICELAKE): New.
13324 (fold_builtin_cpu): Handle icelake.
13325 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13326 * doc/invoke.texi: Add -march=icelake.
13328 2018-02-02 Julia Koval <julia.koval@intel.com>
13330 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13331 to wide_int_bitmask.
13332 * wide-int-bitmask.h: New.
13334 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13337 * config/i386/i386.md: Replace Pmode with word_mode in
13338 builtin_setjmp_setup and builtin_longjmp to support x32.
13340 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13344 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13346 (rs6000_supported_cpu_names): New static variable.
13347 (linux_cpu_translation_table): Likewise.
13348 (elf_platform) <cpu>: Define new static variable and use it.
13349 Translate kernel AT_PLATFORM name to canonical name if needed.
13350 Error if platform name is unknown.
13352 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13355 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13357 2018-02-01 Jeff Law <law@redhat.com>
13360 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13361 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13362 the scratch if RELEASE_VIA_POP is false.
13363 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13364 If we have to save a temporary register, decrement SIZE appropriately.
13365 Pass new arguments to release_scratch_register_on_entry.
13366 (ix86_adjust_stack_and_probe): Likewise.
13367 (ix86_emit_probe_stack_range): Pass new arguments to
13368 release_scratch_register_on_entry.
13370 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13372 PR rtl-optimization/84157
13373 * combine.c (change_zero_ext): Use REG_P predicate in
13374 front of HARD_REGISTER_P predicate.
13376 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13378 * config/avr/avr.c (avr_option_override): Move disabling of
13379 -fdelete-null-pointer-checks to...
13380 * common/config/avr/avr-common.c (avr_option_optimization_table):
13383 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13385 PR tree-optimization/81635
13386 * tree-data-ref.c (split_constant_offset_1): For types that
13387 wrap on overflow, try to use range info to prove that wrapping
13390 2018-02-01 Renlin Li <renlin.li@arm.com>
13393 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13394 TAILCALL_ADDR_REGS.
13395 (aarch64_register_move_cost): Likewise.
13396 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13397 TAILCALL_ADDR_REGS.
13398 (REG_CLASS_NAMES): Likewise.
13399 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13400 TAILCALL_ADDR_REGS. Remove IP registers.
13401 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13403 2018-02-01 Richard Biener <rguenther@suse.de>
13405 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13406 for specifying RPO order and allow NULL for that.
13407 * domwalk.c (dom_walker::dom_walker): Likewise.
13408 (dom_walker::walk): Handle NULL RPO order.
13409 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13411 (rewrite_update_dom_walker): Likewise.
13412 (mark_def_dom_walker): Likewise.
13414 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13416 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13417 (aarch64_maybe_expand_sve_subreg_move): Declare.
13418 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13419 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13421 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13422 that are semantically a reverse operation.
13423 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13424 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13425 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13427 (aarch64_can_change_mode_class): For big-endian, forbid changes
13428 between two SVE modes if they have different element sizes.
13430 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13432 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13433 the TImode handling for big-endian targets.
13435 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13437 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13438 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13440 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13441 Remove BSWAP handing for big-endian targets and use the form of
13442 LD1RQ appropariate for the mode.
13444 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13446 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13447 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13448 duplicated element.
13450 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13453 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13454 check for operands that need to go through aarch64_sve_reload_be.
13456 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13458 PR tree-optimization/81661
13459 PR tree-optimization/84117
13460 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13461 * tree-eh.c: Include gimplify.h.
13462 (find_trapping_overflow, replace_trapping_overflow,
13463 rewrite_to_non_trapping_overflow): New functions.
13464 * tree-vect-loop.c: Include tree-eh.h.
13465 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13466 * tree-data-ref.c: Include tree-eh.h.
13467 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13469 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13471 PR rtl-optimization/84123
13472 * combine.c (change_zero_ext): Check if hard register satisfies
13473 can_change_dest_mode before calling gen_lowpart_SUBREG.
13475 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13478 * ira.c (ira_init_register_move_cost): Remove assert.
13480 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13482 PR rtl-optimization/84071
13483 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13484 * doc/tm.texi: Regenerate.
13486 2018-01-31 Richard Biener <rguenther@suse.de>
13488 PR tree-optimization/84132
13489 * tree-data-ref.c (analyze_miv_subscript): Properly
13490 check whether evolution_function_is_affine_multivariate_p
13491 before calling gcd_of_steps_may_divide_p.
13493 2018-01-31 Julia Koval <julia.koval@intel.com>
13496 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13497 * config/i386/i386.md (rdpid_rex64) New.
13498 (rdpid): Make 32bit only.
13500 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13503 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13504 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13506 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13509 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13511 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13513 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13515 PR rtl-optimization/84071
13516 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13517 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13519 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13521 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13522 (arc_attribute_table): Add 'aux' attribute.
13523 (arc_in_small_data_p): Consider aux like variables.
13524 (arc_is_aux_reg_p): New function.
13525 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13526 (arc_get_aux_arg): New function.
13527 (prepare_move_operands): Handle aux-register access.
13528 (arc_handle_aux_attribute): New function.
13529 * doc/extend.texi (ARC Variable attributes): Add subsection.
13531 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13533 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13534 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13535 (arc_attribute_table): Add 'uncached' attribute.
13536 (arc_print_operand): Print '.di' flag for uncached memory
13538 (arc_in_small_data_p): Do not consider for small data the uncached
13540 (arc_is_uncached_mem_p): New function.
13541 * config/arc/predicates.md (compact_store_memory_operand): Check
13542 for uncached memory accesses.
13543 (nonvol_nonimm_operand): Likewise.
13544 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13546 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13549 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13550 falign-loops=): Add Optimization flag.
13552 2018-01-30 Jeff Law <law@redhat.com>
13555 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13556 INT_REGISTERS_SAVED. Check it prior to calling
13557 get_scratch_register_on_entry.
13558 (ix86_adjust_stack_and_probe): Similarly.
13559 (ix86_emit_probe_stack_range): Similarly.
13560 (ix86_expand_prologue): Corresponding changes.
13562 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13565 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13566 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13568 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13571 * lra-constraints.c (curr_insn_transform): Process AND in the
13574 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13576 PR rtl-optimization/83986
13577 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13578 dependence against last_pending_memory_flush in addition to
13579 pending_jump_insns.
13581 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13583 PR tree-optimization/81611
13584 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13587 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13590 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13593 2018-01-30 Richard Biener <rguenther@suse.de>
13594 Jakub Jelinek <jakub@redhat.com>
13596 PR tree-optimization/84111
13597 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13598 inner loops added during recursion, as they don't have up-to-date
13601 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13604 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13605 (can_inline_edge_by_limits_p): ... here.
13606 (can_early_inline_edge_p, check_callers,
13607 update_caller_keys, update_callee_keys, recursive_inlining,
13608 add_new_edges_to_heap, speculation_useful_p,
13609 inline_small_functions,
13610 inline_small_functions, flatten_function,
13611 inline_to_all_callers_1): Update.
13613 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13615 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13618 2018-01-30 Richard Biener <rguenther@suse.de>
13620 PR tree-optimization/83008
13621 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13622 invariant and constant vector uses in stmts when they need
13623 more than one stmt.
13625 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13628 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13629 * configure: Regenerate.
13631 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13633 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13635 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13636 Use gen_rtx_REG rather than gen_lowpart.
13638 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13640 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13641 rather than 0 when creating partial subregs.
13643 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13645 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13648 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13651 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13652 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13653 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13654 flags. This restores the settings used before the 2017-07-24.
13655 Turning off pre increment/decrement/modify allows IVOPTS to
13656 optimize DF/SF loops where the index is an int.
13658 2018-01-29 Richard Biener <rguenther@suse.de>
13659 Kelvin Nilsen <kelvin@gcc.gnu.org>
13662 * tree-vect-stmts.c (vectorizable_call): Don't call
13663 targetm.vectorize_builtin_md_vectorized_function if callee is
13666 2018-01-22 Carl Love <cel@us.ibm.com>
13668 * doc/extend.tex: Fix typo in second arg in
13669 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13671 2018-01-29 Richard Biener <rguenther@suse.de>
13673 PR tree-optimization/84086
13674 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13675 (flush_ssaname_freelist): When SSA names were released reset
13676 the SCEV hash table.
13678 2018-01-29 Richard Biener <rguenther@suse.de>
13680 PR tree-optimization/84057
13681 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13682 removed paths when removing edges.
13684 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13686 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13687 -mfunction-return=@var{choice}.
13689 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13691 PR diagnostic/84034
13692 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13693 Handle CR like TAB.
13694 (layout::print_source_line): Likewise.
13695 (test_get_line_width_without_trailing_whitespace): Add test cases.
13697 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13699 PR middle-end/84040
13700 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13703 2018-01-26 Jim Wilson <jimw@sifive.com>
13705 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13707 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13710 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13712 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13713 and CMP + SUB-immediate -> SUBS.
13715 2018-01-26 Martin Sebor <msebor@redhat.com>
13717 PR tree-optimization/83896
13718 * tree-ssa-strlen.c (get_string_len): Rename...
13719 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13720 Avoid assuming length is constant.
13721 (handle_char_store): Use HOST_WIDE_INT for string length.
13723 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13726 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13727 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13729 2018-01-26 Richard Biener <rguenther@suse.de>
13731 PR rtl-optimization/84003
13732 * dse.c (record_store): Only record redundant stores when
13733 the earlier store aliases at least all accesses the later one does.
13735 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13737 PR rtl-optimization/83985
13738 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13739 REG_CFA_RESTORE insns.
13740 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13741 REG_CFA_RESTORE insns here.
13744 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13745 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13747 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13749 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13750 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13751 (arc_init): Likewise.
13752 (arc_override_options): Likewise.
13753 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13755 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13757 * config/arc/arc.h (TARGET_DBNZ): Define.
13758 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13759 properly set the tune attribute.
13760 (dbnz): Use TARGET_DBNZ guard.
13761 * config/arc/arc.opt (mtune): Add core3 option.
13763 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13765 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13766 recognize new pic like addresses.
13767 (arc_delegitimize_address): Clean up.
13769 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13771 * config/arc/arc-arches.def: Option mrf16 valid for all
13773 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13774 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13775 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13776 * config/arc/arc-tables.opt: Regenerate.
13777 * config/arc/arc.c (arc_conditional_register_usage): Handle
13778 reduced register file case.
13779 (arc_file_start): Set must have build attributes.
13780 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13781 mrf16 option value.
13782 * config/arc/arc.opt (mrf16): Add new option.
13783 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13784 * config/arc/genmultilib.awk: Handle new mrf16 option.
13785 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13786 * config/arc/t-multilib: Regenerate.
13787 * doc/invoke.texi (ARC Options): Document mrf16 option.
13789 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13791 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13792 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13793 (arc_attribute_table): Add 'secure_call' attribute.
13794 (arc_print_operand): Print secure call operand.
13795 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13797 (arc_is_secure_call_p): New function. * config/arc/arc.md
13798 (call_i): Add support for sjli instruction.
13799 (call_value_i): Likewise.
13800 * config/arc/constraints.md (Csc): New constraint.
13802 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13803 John Eric Martin <John.Martin@emmicro-us.com>
13805 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13806 * config/arc/arc.c (_arc_jli_section): New struct.
13807 (arc_jli_section): New type.
13808 (rc_jli_sections): New static variable.
13809 (arc_handle_jli_attribute): New function.
13810 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13811 (arc_file_end): New function.
13812 (TARGET_ASM_FILE_END): Define.
13813 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13814 (arc_add_jli_section): New function.
13815 (jli_call_scan): Likewise.
13816 (arc_reorg): Call jli_call_scan.
13817 (arc_output_addsi): Remove 'S' from printing asm operand.
13818 (arc_is_jli_call_p): New function.
13819 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13821 (movhi_insn): Likewise.
13822 (movsi_insn): Likewise.
13823 (movsi_set_cc_insn): Likewise.
13824 (loadqi_update): Likewise.
13825 (load_zeroextendqisi_update): Likewise.
13826 (load_signextendqisi_update): Likewise.
13827 (loadhi_update): Likewise.
13828 (load_zeroextendhisi_update): Likewise.
13829 (load_signextendhisi_update): Likewise.
13830 (loadsi_update): Likewise.
13831 (loadsf_update): Likewise.
13832 (movsicc_insn): Likewise.
13833 (bset_insn): Likewise.
13834 (bxor_insn): Likewise.
13835 (bclr_insn): Likewise.
13836 (bmsk_insn): Likewise.
13837 (bicsi3_insn): Likewise.
13838 (cmpsi_cc_c_insn): Likewise.
13839 (movsi_ne): Likewise.
13840 (movsi_cond_exec): Likewise.
13841 (clrsbsi2): Likewise.
13842 (norm_f): Likewise.
13852 (call_i): Remove 'S' asm letter, add jli instruction.
13853 (call_value_i): Likewise.
13854 * config/arc/arc.op (mjli-always): New option.
13855 * config/arc/constraints.md (Cji): New constraint.
13856 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13858 (subsf3_fpx): Likewise.
13859 (mulsf3_fpx): Likewise.
13860 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13862 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13863 function attrbutes.
13864 * doc/invoke.texi (ARC): Document mjli-always option.
13866 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13868 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13869 avoid addition with 0 and use incw and decw where possible.
13871 2018-01-26 Richard Biener <rguenther@suse.de>
13873 PR tree-optimization/81082
13874 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13875 association if it requires casting to unsigned.
13876 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13877 from fold_plusminus_mult_expr to catch important cases late when
13878 range info is available.
13880 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13882 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13883 * configure.ac (hidden_linkonce): New test.
13884 * configure: Regenerate.
13885 * config.in: Regenerate.
13887 2018-01-26 Julia Koval <julia.koval@intel.com>
13889 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13890 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13891 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13892 _mm_mask_bitshuffle_epi64_mask): Fix type.
13893 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13894 USI_FTYPE_V4DI_V4DI_USI): Remove.
13895 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13896 __builtin_ia32_vpshufbitqmb256_mask,
13897 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13898 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13899 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13901 2018-01-26 Alan Modra <amodra@gmail.com>
13904 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13905 UNSPEC_VBPERMQ. Sort other unspecs.
13907 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13909 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13911 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13913 PR middle-end/83055
13914 * predict.c (drop_profile): Do not push/pop cfun; update also
13916 (handle_missing_profiles): Fix logic looking for zero profiles.
13918 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13920 PR middle-end/83977
13921 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13922 on functions with #pragma omp declare simd or functions with simd
13924 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13925 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13926 Remove trailing \n from warning_at calls.
13928 2018-01-25 Tom de Vries <tom@codesourcery.com>
13931 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13932 for neutered workers.
13934 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13937 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13938 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13940 2018-01-24 Jeff Law <law@redhat.com>
13943 * i386.c (get_probe_interval): Move to earlier point.
13944 (ix86_compute_frame_layout): If -fstack-clash-protection and
13945 the frame is larger than the probe interval, then use pushes
13946 to save registers rather than reg->mem moves.
13947 (ix86_expand_prologue): Remove conditional for int_registers_saved
13950 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13953 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13954 min/max for never referenced object.
13956 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13958 PR middle-end/83977
13959 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13961 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13962 attributes from DECL_ATTRIBUTES (decl) without affecting
13963 DECL_ATTRIBUTES (current_function_decl).
13964 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13965 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13967 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13969 PR tree-optimization/83979
13970 * fold-const.c (fold_comparison): Use constant_boolean_node
13971 instead of boolean_{true,false}_node.
13973 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13975 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13978 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13980 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13981 Simplify the clause that sets the length attribute.
13982 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13983 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13984 clause that sets the length attribute.
13985 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13987 2018-01-24 Tom de Vries <tom@codesourcery.com>
13990 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13991 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13992 Add strict parameter.
13993 (prevent_branch_around_nothing): Insert dummy insn between branch to
13994 label and label with no ptx insn inbetween.
13995 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13997 2018-01-24 Tom de Vries <tom@codesourcery.com>
14000 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14001 for neutered threads in warp.
14002 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14004 2018-01-24 Richard Biener <rguenther@suse.de>
14006 PR tree-optimization/83176
14007 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14010 2018-01-24 Richard Biener <rguenther@suse.de>
14012 PR tree-optimization/82819
14013 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14014 code generating pluses that are no-ops in the target precision.
14016 2018-01-24 Richard Biener <rguenther@suse.de>
14018 PR middle-end/84000
14019 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14021 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14023 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14024 to merge probabilities.
14025 * predict.c (probably_never_executed): Also mark as cold functions
14026 with global 0 profile and guessed local profile.
14027 * profile-count.c (profile_probability::combine_with_count): New
14029 * profile-count.h (profile_probability::operator*,
14030 profile_probability::operator*=, profile_probability::operator/,
14031 profile_probability::operator/=): Reduce precision to adjusted
14032 and set value to guessed on contradictory divisions.
14033 (profile_probability::combine_with_freq): Remove.
14034 (profile_probability::combine_wiht_count): Declare.
14035 (profile_count::force_nonzero):: Set to adjusted.
14036 (profile_count::probability_in):: Set quality to adjusted.
14037 * tree-ssa-tail-merge.c (replace_block_by): Use
14038 combine_with_count.
14040 2018-01-23 Andrew Waterman <andrew@sifive.com>
14041 Jim Wilson <jimw@sifive.com>
14043 * config/riscv/riscv.c (riscv_stack_boundary): New.
14044 (riscv_option_override): Set riscv_stack_boundary. Handle
14045 riscv_preferred_stack_boundary_arg.
14046 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14047 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14048 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14049 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14050 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14051 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14053 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14056 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14057 of struct ix86_frame.
14058 (ix86_expand_epilogue): Likewise. Add a local variable for
14059 the reg_save_offset field in struct ix86_frame.
14061 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14063 PR tree-optimization/82604
14064 * tree-loop-distribution.c (enum partition_kind): New enum item
14065 PKIND_PARTIAL_MEMSET.
14066 (partition_builtin_p): Support above new enum item.
14067 (generate_code_for_partition): Ditto.
14068 (compute_access_range): Differentiate cases that equality can be
14069 proven at all loops, the innermost loops or no loops.
14070 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14071 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14072 (finalize_partitions, distribute_loop): Don't fuse partition of
14073 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14074 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14075 parloop is enabled.
14077 2018-01-23 Martin Liska <mliska@suse.cz>
14079 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14080 order to ignore the predictor.
14081 (PRED_POLYMORPHIC_CALL): Likewise.
14082 (PRED_RECURSIVE_CALL): Likewise.
14084 2018-01-23 Martin Liska <mliska@suse.cz>
14086 * tree-profile.c (tree_profiling): Print function header to
14087 aware reader which function we are working on.
14088 * value-prof.c (gimple_find_values_to_profile): Do not print
14089 not interesting value histograms.
14091 2018-01-23 Martin Liska <mliska@suse.cz>
14093 * profile-count.h (enum profile_quality): Add
14094 profile_uninitialized as the first value. Do not number values
14095 as they are zero based.
14096 (profile_count::verify): Update sanity check.
14097 (profile_probability::verify): Likewise.
14099 2018-01-23 Nathan Sidwell <nathan@acm.org>
14101 * doc/invoke.texi (ffor-scope): Deprecate.
14103 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14105 PR tree-optimization/83510
14106 * domwalk.c (set_all_edges_as_executable): New function.
14107 (dom_walker::dom_walker): Convert bool param
14108 "skip_unreachable_blocks" to enum reachability. Move setup of
14109 edge flags to set_all_edges_as_executable and only do it when
14110 reachability is REACHABLE_BLOCKS.
14111 * domwalk.h (enum dom_walker::reachability): New enum.
14112 (dom_walker::dom_walker): Convert bool param
14113 "skip_unreachable_blocks" to enum reachability.
14114 (set_all_edges_as_executable): New decl.
14115 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14116 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14118 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14119 but converting true to REACHABLE_BLOCKS.
14120 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14122 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14123 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14124 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14126 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14127 if check_all_array_refs will be called.
14129 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14131 * tree.c (selftest::test_location_wrappers): Add more test
14134 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14136 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14137 (selftest::test_bit_in_range): Likewise.
14139 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14142 * doc/sourcebuild.texi (vect_float): Say that the selector
14143 only describes the situation when -funsafe-math-optimizations is on.
14144 (vect_float_strict): Document.
14146 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14148 PR tree-optimization/83965
14149 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14150 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14151 instead of checking only for a reduction.
14152 (vect_recog_widen_sum_pattern): Likewise.
14154 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14156 * predict.c (probably_never_executed): Only use precise profile info.
14157 (compute_function_frequency): Skip after inlining hack since we now
14158 have quality checking.
14160 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14162 * profile-count.h (profile_probability::very_unlikely,
14163 profile_probability::unlikely, profile_probability::even): Set
14164 precision to guessed.
14166 2018-01-23 Richard Biener <rguenther@suse.de>
14168 PR tree-optimization/83963
14169 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14170 Properly terminate dominator walk when crossing the exit edge not
14171 when visiting its source block.
14173 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14176 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14177 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14179 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14181 PR tree-optimization/83957
14182 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14183 semicolon after for body surrounded by braces.
14185 PR tree-optimization/83081
14186 * profile-count.h (profile_probability::split): New method.
14187 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14188 Use profile_probability::split.
14189 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14190 when splitting a single conditional jump into 2.
14192 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14194 PR tree-optimization/69452
14195 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14200 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14201 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14202 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14204 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14206 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14208 * config/rl78/rl78.md (movdi): New define_expand.
14209 * config/rl78/rl78.c (rl78_split_movdi): New function.
14211 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14214 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14216 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14217 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14218 128-bit to produce an UNSPEC move to get the double word with the
14219 signbit and then a shift directly to do signbit.
14220 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14221 implementation with a new version that just does either a direct
14222 move or a regular move. Move memory interface to separate insns.
14223 Move insns so they are next to the expander.
14224 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14225 with signbit move. Split big and little endian case.
14226 (signbit<mode>2_dm_mem_le): Likewise.
14227 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14228 (signbit<mode>2_dm2): Likewise.
14230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14232 * config/rl78/rl78.md (anddi3): New define_expand.
14234 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14236 * config/rl78/rl78.md (umindi3): New define_expand.
14238 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14240 * config/rl78/rl78.md (smindi3): New define_expand.
14242 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14244 * config/rl78/rl78.md (smaxdi3): New define_expand.
14246 2018-01-22 Carl Love <cel@us.ibm.com>
14248 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14249 LVX_V1TI): Add macro expansion.
14250 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14251 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14252 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14253 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14254 Change check to determine if the instruction is a byte reversing
14255 entry. Fix typo in comment.
14256 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14257 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14258 Add def_builtin calls for new builtins.
14259 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14260 Add define_insn expansion.
14262 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14264 * config/rl78/rl78.md (umaxdi3): New define_expand.
14266 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14268 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14269 for non-QImode registers.
14271 2018-01-22 Richard Biener <rguenther@suse.de>
14273 PR tree-optimization/83963
14274 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14275 including the loop exit block.
14276 (scop_detection::merge_sese): Likewise.
14277 (scop_detection::add_scop): Do it here instead.
14279 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14281 * doc/sourcebuild.texi (arm_softfloat): Document.
14283 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14286 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14287 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14288 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14290 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14291 David Edelsohn <dje.gcc@gmail.com>
14294 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14295 Change "crset eq" to "crset 2".
14296 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14297 (*call_indirect_aix<mode>_nospec): Likewise.
14298 (*call_value_indirect_aix<mode>_nospec): Likewise.
14299 (*call_indirect_elfv2<mode>_nospec): Likewise.
14300 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14301 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14302 change assembly output from . to $.
14303 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14304 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14305 (*tablejump<mode>_internal1_nospec): Likewise.
14307 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14310 * config/sh/sh_optimize_sett_clrt.cc:
14311 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14313 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14315 PR tree-optimization/83940
14316 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14317 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14318 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14319 use it to pass back the definition type.
14320 (vect_check_store_rhs): Likewise rhs_dt_out.
14321 (vect_build_gather_load_calls): Add a mask_dt argument and use
14322 it instead of a call to vect_is_simple_use.
14323 (vectorizable_store): Update calls to vect_check_load_store_mask
14324 and vect_check_store_rhs. Use the dt returned by the latter instead
14325 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14326 instead of calls to vect_is_simple_use. Pass the scalar rather
14327 than the vector operand to vect_is_simple_use when handling
14328 second and subsequent copies of an rhs value.
14329 (vectorizable_load): Update calls to vect_check_load_store_mask
14330 and vect_build_gather_load_calls. Use the cached mask_dt and
14331 gs_info.offset_dt instead of calls to vect_is_simple_use.
14333 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14335 PR middle-end/83945
14336 * tree-emutls.c: Include gimplify.h.
14337 (lower_emutls_2): New function.
14338 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14339 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14340 it before further processing.
14343 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14344 UINTVAL (trueop1) instead of INTVAL (op1).
14346 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14350 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14351 INCOMING_FRAME_SP_OFFSET if not defined.
14352 (scan_trace): Add ENTRY argument. If true and
14353 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14354 emit a note to adjust the CFA offset.
14355 (create_cfi_notes): Adjust scan_trace callers.
14356 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14357 INCOMING_FRAME_SP_OFFSET in the CIE.
14358 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14359 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14361 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14362 * doc/tm.texi: Regenerated.
14364 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14366 PR rtl-optimization/83147
14367 * lra-constraints.c (remove_inheritance_pseudos): Use
14368 lra_substitute_pseudo_within_insn.
14370 2018-01-19 Tom de Vries <tom@codesourcery.com>
14371 Cesar Philippidis <cesar@codesourcery.com>
14374 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14376 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14379 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14380 spaces for function labels.
14382 2018-01-19 Martin Liska <mliska@suse.cz>
14384 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14385 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14386 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14387 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14388 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14389 (PRED_CONST_RETURN): Change from 69 to 65.
14390 (PRED_NULL_RETURN): Change from 91 to 71.
14391 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14392 (PRED_LOOP_GUARD): Change from 66 to 73.
14394 2018-01-19 Martin Liska <mliska@suse.cz>
14396 * predict.c (predict_insn_def): Add new assert.
14397 (struct branch_predictor): Change type to signed integer.
14398 (test_prediction_value_range): Amend test to cover
14399 PROB_UNINITIALIZED.
14400 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14401 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14402 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14403 (PRED_LOOP_IV_COMPARE): Likewise.
14404 * predict.h (PROB_UNINITIALIZED): Define new constant.
14406 2018-01-19 Martin Liska <mliska@suse.cz>
14408 * predict.c (dump_prediction): Add new format for
14409 analyze_brprob.py script which is enabled with -details
14411 * profile-count.h (precise_p): New function.
14413 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14415 PR tree-optimization/83922
14416 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14417 there are no statements that need masking.
14418 (vect_active_double_reduction_p): New function.
14419 (vect_analyze_loop_operations): Use it when handling phis that
14420 are not in the loop header.
14422 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14424 PR tree-optimization/83914
14425 * tree-vect-loop.c (vectorizable_induction): Don't convert
14426 init_expr or apply the peeling adjustment for inductions
14427 that are nested within the vectorized loop.
14429 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14431 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14434 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14438 * function.h (gimplify_parameters): Add gimple_seq * argument.
14439 * function.c: Include gimple.h and options.h.
14440 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14441 for the added local temporaries if needed.
14442 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14443 if there are any parameter cleanups, wrap whole body into a
14444 try/finally with the cleanups.
14446 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14449 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14450 Use GET_MODE_CLASS for scalar floating point.
14452 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14456 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14457 Fix call of call_cgraph_insertion_hooks.
14459 2018-01-18 Martin Sebor <msebor@redhat.com>
14461 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14463 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14466 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14469 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14472 * common.opt: (-ffile-prefix-map): New option.
14473 * opts.c (common_handle_option): Defer it.
14474 * opts-global.c (handle_common_deferred_options): Handle it.
14475 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14476 * file-prefix-map.h: New file.
14477 (remap_debug_filename, add_debug_prefix_map): ...here.
14478 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14479 * final.c (debug_prefix_map, add_debug_prefix_map
14480 remap_debug_filename): Move to...
14481 * file-prefix-map.c: New file.
14482 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14483 generalize, get rid of alloca(), use strrchr() instead of strchr().
14484 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14485 Implement in terms of add_prefix_map().
14486 (remap_macro_filename, remap_debug_filename): Implement in term of
14488 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14489 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14490 * dbxout.c: Include file-prefix-map.h.
14491 * varasm.c: Likewise.
14492 * vmsdbgout.c: Likewise.
14493 * xcoffout.c: Likewise.
14494 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14495 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14496 * doc/invoke.texi (-ffile-prefix-map): Document.
14497 (-fdebug-prefix-map): Update description.
14499 2018-01-18 Martin Liska <mliska@suse.cz>
14501 * config/i386/i386.c (indirect_thunk_name): Document that also
14503 (output_indirect_thunk): Document why both instructions
14504 (pause and lfence) are generated.
14506 2018-01-18 Richard Biener <rguenther@suse.de>
14508 PR tree-optimization/83887
14509 * graphite-scop-detection.c
14510 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14511 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14512 (scop_detection::merge_sese): Re-implement with a flood-fill
14513 algorithm that properly finds a SESE region if it exists.
14515 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14518 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14519 pointer_diff optimizations use view_convert instead of convert.
14521 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14523 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14524 Generate different code for -mno-speculate-indirect-jumps.
14525 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14526 (*call_indirect_aix<mode>): Disable for
14527 -mno-speculate-indirect-jumps.
14528 (*call_indirect_aix<mode>_nospec): New define_insn.
14529 (*call_value_indirect_aix<mode>): Disable for
14530 -mno-speculate-indirect-jumps.
14531 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14532 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14533 -mno-speculate-indirect-jumps.
14534 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14536 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14538 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14539 long double type, set the flags for noting the default long double
14540 type, even if we don't pass or return a long double type.
14542 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14545 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14548 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14550 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14551 support for merge[hl].
14552 (fold_mergehl_helper): New helper function.
14553 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14554 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14555 (altivec_vmrglw_direct): Add xxmrglw insn.
14557 2018-01-17 Andrew Waterman <andrew@sifive.com>
14559 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14560 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14562 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14565 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14566 call the lto_location_cache before reading the
14567 DECL_SOURCE_LOCATION of the types.
14569 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14570 Richard Sandiford <richard.sandiford@linaro.org>
14572 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14573 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14574 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14575 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14576 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14578 * config/aarch64/constraints.md (aarch64_movti_operand):
14580 * config/aarch64/predicates.md (Uti): Add new constraint.
14582 2018-01-17 Carl Love <cel@us.ibm.com>
14584 * config/rs6000/vsx.md (define_expand xl_len_r,
14585 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14586 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14588 (define_expand, define_insn): Move the shift left from the
14589 define_insn to the define_expand for lxvl and stxvl instructions.
14590 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14591 and XL_LEN_R definitions to PURE.
14593 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14595 * config/i386/i386.c (indirect_thunk_name): Declare regno
14596 as unsigned int. Compare regno with INVALID_REGNUM.
14597 (output_indirect_thunk): Ditto.
14598 (output_indirect_thunk_function): Ditto.
14599 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14600 in the call to output_indirect_thunk_function.
14602 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14604 PR middle-end/83884
14605 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14606 rather than the size of inner_type to determine the stack slot size
14607 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14609 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14612 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14615 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14617 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14618 endian Linux systems to optionally enable multilibs for selecting
14619 the long double type if the user configured an explicit type.
14620 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14621 have no long double multilibs if not defined.
14622 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14623 warn if the user used -mabi={ieee,ibm}longdouble and we built
14624 multilibs for long double.
14625 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14626 appropriate multilib option.
14627 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14629 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14630 for building long double multilibs.
14631 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14633 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14635 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14638 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14640 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14643 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14646 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14649 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14652 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14654 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14656 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14657 different rtl trees depending on TARGET_64BIT.
14658 (rs6000_gen_lvx): Likewise.
14660 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14662 * config/visium/visium.md (nop): Tweak comment.
14663 (hazard_nop): Likewise.
14665 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14667 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14668 -mspeculate-indirect-jumps.
14669 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14670 for -mno-speculate-indirect-jumps.
14671 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14672 (*call_value_indirect_elfv2<mode>): Disable for
14673 -mno-speculate-indirect-jumps.
14674 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14675 (indirect_jump): Emit different RTL for
14676 -mno-speculate-indirect-jumps.
14677 (*indirect_jump<mode>): Disable for
14678 -mno-speculate-indirect-jumps.
14679 (*indirect_jump<mode>_nospec): New define_insn.
14680 (tablejump): Emit different RTL for
14681 -mno-speculate-indirect-jumps.
14682 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14683 (tablejumpsi_nospec): New define_expand.
14684 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14685 (tablejumpdi_nospec): New define_expand.
14686 (*tablejump<mode>_internal1): Disable for
14687 -mno-speculate-indirect-jumps.
14688 (*tablejump<mode>_internal1_nospec): New define_insn.
14689 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14692 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14694 * caller-save.c (insert_save): Drop unnecessary parameter. All
14697 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14698 Richard Biener <rguenth@suse.de>
14701 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14702 return early, inline manually is_gimple_sizepos. Make sure if we
14703 call gimplify_expr we don't end up with a gimple constant.
14704 * tree.c (variably_modified_type_p): Don't return true for
14705 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14706 * gimplify.h (is_gimple_sizepos): Remove.
14708 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14710 PR tree-optimization/83857
14711 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14712 vectorizable_live_operation for pure SLP statements.
14713 (vectorizable_live_operation): Handle PHIs.
14715 2018-01-16 Richard Biener <rguenther@suse.de>
14717 PR tree-optimization/83867
14718 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14719 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14721 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14724 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14725 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14726 If off is not INTEGER_CST, issue a may not be aligned warning
14727 rather than isn't aligned. Use isn%'t rather than isn't.
14728 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14730 <case MULT_EXPR>: Improve the case when bottom and one of the
14731 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14732 operand, in that case check if the other operand is multiple of
14733 bottom divided by the INTEGER_CST operand.
14735 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14738 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14739 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14740 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14741 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14742 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14743 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14744 (pa_function_arg_size): New function.
14746 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14748 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14749 in a separate statement.
14751 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14753 PR tree-optimization/83847
14754 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14755 group gathers and scatters.
14757 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14759 PR rtl-optimization/86620
14760 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14762 PR rtl-optimization/83213
14763 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14764 to last if both are JUMP_INSNs.
14766 PR tree-optimization/83843
14767 * gimple-ssa-store-merging.c
14768 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14769 store_immediate_info for bswap/nop orig_stores.
14771 2018-01-15 Andrew Waterman <andrew@sifive.com>
14773 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14775 <UDIV>: Increase cost if !TARGET_DIV.
14777 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14779 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14780 (define_attr "cr_logical_3op"): New.
14781 (cceq_ior_compare): Adjust.
14782 (cceq_ior_compare_complement): Adjust.
14783 (*cceq_rev_compare): Adjust.
14784 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14785 (is_cracked_insn): Adjust.
14786 (insn_must_be_first_in_group): Adjust.
14787 * config/rs6000/40x.md: Adjust.
14788 * config/rs6000/440.md: Adjust.
14789 * config/rs6000/476.md: Adjust.
14790 * config/rs6000/601.md: Adjust.
14791 * config/rs6000/603.md: Adjust.
14792 * config/rs6000/6xx.md: Adjust.
14793 * config/rs6000/7450.md: Adjust.
14794 * config/rs6000/7xx.md: Adjust.
14795 * config/rs6000/8540.md: Adjust.
14796 * config/rs6000/cell.md: Adjust.
14797 * config/rs6000/e300c2c3.md: Adjust.
14798 * config/rs6000/e500mc.md: Adjust.
14799 * config/rs6000/e500mc64.md: Adjust.
14800 * config/rs6000/e5500.md: Adjust.
14801 * config/rs6000/e6500.md: Adjust.
14802 * config/rs6000/mpc.md: Adjust.
14803 * config/rs6000/power4.md: Adjust.
14804 * config/rs6000/power5.md: Adjust.
14805 * config/rs6000/power6.md: Adjust.
14806 * config/rs6000/power7.md: Adjust.
14807 * config/rs6000/power8.md: Adjust.
14808 * config/rs6000/power9.md: Adjust.
14809 * config/rs6000/rs64.md: Adjust.
14810 * config/rs6000/titan.md: Adjust.
14812 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14814 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14815 ix86_indirect_branch_register logic.
14817 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14819 * config/i386/constraints.md (Bs): Update
14820 ix86_indirect_branch_register check. Don't check
14821 ix86_indirect_branch_register with GOT_memory_operand.
14823 * config/i386/predicates.md (GOT_memory_operand): Don't check
14824 ix86_indirect_branch_register here.
14825 (GOT32_symbol_operand): Likewise.
14827 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14829 * config/i386/predicates.md (constant_call_address_operand):
14830 Rewrite ix86_indirect_branch_register logic.
14831 (sibcall_insn_operand): Likewise.
14833 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14835 * config/i386/constraints.md (Bs): Replace
14836 ix86_indirect_branch_thunk_register with
14837 ix86_indirect_branch_register.
14839 * config/i386/i386.md (indirect_jump): Likewise.
14840 (tablejump): Likewise.
14841 (*sibcall_memory): Likewise.
14842 (*sibcall_value_memory): Likewise.
14843 Peepholes of indirect call and jump via memory: Likewise.
14844 * config/i386/i386.opt: Likewise.
14845 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14846 (GOT_memory_operand): Likewise.
14847 (call_insn_operand): Likewise.
14848 (sibcall_insn_operand): Likewise.
14849 (GOT32_symbol_operand): Likewise.
14851 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14853 PR middle-end/83837
14854 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14855 type rather than type addr's type points to.
14856 (expand_omp_atomic_mutex): Likewise.
14857 (expand_omp_atomic): Likewise.
14859 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14862 * config/i386/i386.c (output_indirect_thunk_function): Use
14863 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14864 for __x86_return_thunk.
14866 2018-01-15 Richard Biener <rguenther@suse.de>
14868 PR middle-end/83850
14869 * expmed.c (extract_bit_field_1): Fix typo.
14871 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14874 * config/arm/iterators.md (VF): New mode iterator.
14875 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14876 Remove integer-related logic from pattern.
14877 (neon_vabd<mode>_3): Likewise.
14879 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14881 PR middle-end/82694
14882 * common.opt (fstrict-overflow): No longer an alias.
14883 (fwrapv-pointer): New option.
14884 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14885 also for pointer types based on flag_wrapv_pointer.
14886 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14887 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14888 opts->x_flag_wrapv got set.
14889 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14890 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14891 POINTER_TYPE_OVERFLOW_UNDEFINED.
14892 * match.pd: Likewise in address comparison pattern.
14893 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14895 2018-01-15 Richard Biener <rguenther@suse.de>
14898 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14899 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14900 Reset type names to their identifier if their TYPE_DECL doesn't
14901 have linkage (and thus is used for ODR and devirt).
14902 (save_debug_info_for_decl): Remove.
14903 (save_debug_info_for_type): Likewise.
14904 (add_tree_to_fld_list): Adjust.
14905 * tree-pretty-print.c (dump_generic_node): Make dumping of
14906 type names more robust.
14908 2018-01-15 Richard Biener <rguenther@suse.de>
14910 * BASE-VER: Bump to 8.0.1.
14912 2018-01-14 Martin Sebor <msebor@redhat.com>
14915 * builtins.c (check_access): Avoid warning when the no-warning bit
14918 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14920 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14921 * ira-color (allocno_hard_regs_compare): Likewise.
14923 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14926 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14927 Use .pushsection/.popsection.
14929 2018-01-14 Martin Sebor <msebor@redhat.com>
14932 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14934 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14936 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14937 entry from extra_headers.
14938 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14939 extra_headers, make the list bitwise identical to the i?86-*-* one.
14941 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14943 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14944 -mcmodel=large with -mindirect-branch=thunk,
14945 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14946 -mfunction-return=thunk-extern.
14947 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14948 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14949 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14953 * config/i386/i386.c (print_reg): Print the name of the full
14954 integer register without '%'.
14955 (ix86_print_operand): Handle 'V'.
14956 * doc/extend.texi: Document 'V' modifier.
14958 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14960 * config/i386/constraints.md (Bs): Disallow memory operand for
14961 -mindirect-branch-register.
14963 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14964 (GOT_memory_operand): Likewise.
14965 (call_insn_operand): Likewise.
14966 (sibcall_insn_operand): Likewise.
14967 (GOT32_symbol_operand): Likewise.
14968 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14969 for -mindirect-branch-register.
14970 (tablejump): Likewise.
14971 (*sibcall_memory): Likewise.
14972 (*sibcall_value_memory): Likewise.
14973 Disallow peepholes of indirect call and jump via memory for
14974 -mindirect-branch-register.
14975 (*call_pop): Replace m with Bw.
14976 (*call_value_pop): Likewise.
14977 (*sibcall_pop_memory): Replace m with Bs.
14978 * config/i386/i386.opt (mindirect-branch-register): New option.
14979 * doc/invoke.texi: Document -mindirect-branch-register option.
14981 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14983 * config/i386/i386-protos.h (ix86_output_function_return): New.
14984 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14985 set function_return_type.
14986 (indirect_thunk_name): Add ret_p to indicate thunk for function
14988 (output_indirect_thunk_function): Pass false to
14989 indirect_thunk_name.
14990 (ix86_output_indirect_branch_via_reg): Likewise.
14991 (ix86_output_indirect_branch_via_push): Likewise.
14992 (output_indirect_thunk_function): Create alias for function
14993 return thunk if regno < 0.
14994 (ix86_output_function_return): New function.
14995 (ix86_handle_fndecl_attribute): Handle function_return.
14996 (ix86_attribute_table): Add function_return.
14997 * config/i386/i386.h (machine_function): Add
14998 function_return_type.
14999 * config/i386/i386.md (simple_return_internal): Use
15000 ix86_output_function_return.
15001 (simple_return_internal_long): Likewise.
15002 * config/i386/i386.opt (mfunction-return=): New option.
15003 (indirect_branch): Mention -mfunction-return=.
15004 * doc/extend.texi: Document function_return function attribute.
15005 * doc/invoke.texi: Document -mfunction-return= option.
15007 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15009 * config/i386/i386-opts.h (indirect_branch): New.
15010 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15011 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15012 with local indirect jump when converting indirect call and jump.
15013 (ix86_set_indirect_branch_type): New.
15014 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15015 (indirectlabelno): New.
15016 (indirect_thunk_needed): Likewise.
15017 (indirect_thunk_bnd_needed): Likewise.
15018 (indirect_thunks_used): Likewise.
15019 (indirect_thunks_bnd_used): Likewise.
15020 (INDIRECT_LABEL): Likewise.
15021 (indirect_thunk_name): Likewise.
15022 (output_indirect_thunk): Likewise.
15023 (output_indirect_thunk_function): Likewise.
15024 (ix86_output_indirect_branch_via_reg): Likewise.
15025 (ix86_output_indirect_branch_via_push): Likewise.
15026 (ix86_output_indirect_branch): Likewise.
15027 (ix86_output_indirect_jmp): Likewise.
15028 (ix86_code_end): Call output_indirect_thunk_function if needed.
15029 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15031 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15032 (ix86_attribute_table): Add indirect_branch.
15033 * config/i386/i386.h (machine_function): Add indirect_branch_type
15034 and has_local_indirect_jump.
15035 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15037 (tablejump): Likewise.
15038 (*indirect_jump): Use ix86_output_indirect_jmp.
15039 (*tablejump_1): Likewise.
15040 (simple_return_indirect_internal): Likewise.
15041 * config/i386/i386.opt (mindirect-branch=): New option.
15042 (indirect_branch): New.
15045 (thunk-inline): Likewise.
15046 (thunk-extern): Likewise.
15047 * doc/extend.texi: Document indirect_branch function attribute.
15048 * doc/invoke.texi: Document -mindirect-branch= option.
15050 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15053 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15055 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15057 * ipa-inline.c (want_inline_small_function_p): Return false if
15058 inlining has already failed with CIF_FINAL_ERROR.
15059 (update_caller_keys): Call want_inline_small_function_p before
15061 (update_callee_keys): Likewise.
15063 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15065 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15067 (rs6000_quadword_masked_address_p): Likewise.
15068 (quad_aligned_load_p): Likewise.
15069 (quad_aligned_store_p): Likewise.
15070 (const_load_sequence_p): Add comment to describe the outer-most loop.
15071 (mimic_memory_attributes_and_flags): New function.
15072 (rs6000_gen_stvx): Likewise.
15073 (replace_swapped_aligned_store): Likewise.
15074 (rs6000_gen_lvx): Likewise.
15075 (replace_swapped_aligned_load): Likewise.
15076 (replace_swapped_load_constant): Capitalize argument name in
15077 comment describing this function.
15078 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15079 and stores that access quad-word aligned addresses and replace
15080 with stvx or lvx instructions when appropriate.
15081 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15082 New function prototype.
15083 (rs6000_quadword_masked_address_p): Likewise.
15084 (rs6000_gen_lvx): Likewise.
15085 (rs6000_gen_stvx): Likewise.
15086 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15087 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15088 when memory address is aligned.
15089 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15090 this split to select lvx instruction when memory address is aligned.
15091 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15092 instruction when memory address is aligned.
15093 (*vsx_le_perm_load_v16qi): Likewise.
15094 (four unnamed splitters): Modify to select the stvx instruction
15095 when memory is aligned.
15097 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15099 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15100 which appears in the queue multiple times.
15102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15103 Alan Hayward <alan.hayward@arm.com>
15104 David Sherwood <david.sherwood@arm.com>
15106 * tree-vectorizer.h (vec_lower_bound): New structure.
15107 (_loop_vec_info): Add check_nonzero and lower_bounds.
15108 (LOOP_VINFO_CHECK_NONZERO): New macro.
15109 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15110 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15111 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15112 fields. Make seg_len the distance travelled, not including the
15114 (dr_direction_indicator): Declare.
15115 (dr_zero_step_indicator): Likewise.
15116 (dr_known_forward_stride_p): Likewise.
15117 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15119 (runtime_alias_check_p): Allow runtime alias checks with
15121 (operator ==): Compare access_size and align.
15122 (prune_runtime_alias_test_list): Rework for new distinction between
15123 the access_size and seg_len.
15124 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15126 (get_segment_min_max): New function.
15127 (create_intersect_range_checks): Use it.
15128 (dr_step_indicator): New function.
15129 (dr_direction_indicator): Likewise.
15130 (dr_zero_step_indicator): Likewise.
15131 (dr_known_forward_stride_p): Likewise.
15132 * tree-loop-distribution.c (data_ref_segment_size): Return
15133 DR_STEP * (niters - 1).
15134 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15136 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15137 (vect_preserves_scalar_order_p): New function, split out from...
15138 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15139 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15140 (vect_vfa_access_size): New function.
15141 (vect_vfa_align): Likewise.
15142 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15143 (dump_lower_bound): New function.
15144 (vect_check_lower_bound): Likewise.
15145 (vect_small_gap_p): Likewise.
15146 (vectorizable_with_step_bound_p): Likewise.
15147 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15148 depencies if the vectorization factor is 1. Convert the checks
15149 for nonzero steps into checks on the bounds of DR_STEP. Try using
15150 a bunds check for variable steps if the minimum required step is
15151 relatively small. Update calls to the dr_with_seg_len
15152 constructor and to vect_compile_time_alias.
15153 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15155 (vect_loop_versioning): Call it.
15156 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15158 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15161 Alan Hayward <alan.hayward@arm.com>
15162 David Sherwood <david.sherwood@arm.com>
15164 * doc/sourcebuild.texi (vect_scatter_store): Document.
15165 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15167 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15169 * genopinit.c (main): Add supports_vec_scatter_store and
15170 supports_vec_scatter_store_cached to target_optabs.
15171 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15172 IFN_MASK_SCATTER_STORE.
15173 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15175 * internal-fn.h (internal_store_fn_p): Declare.
15176 (internal_fn_stored_value_index): Likewise.
15177 * internal-fn.c (scatter_store_direct): New macro.
15178 (expand_scatter_store_optab_fn): New function.
15179 (direct_scatter_store_optab_supported_p): New macro.
15180 (internal_store_fn_p): New function.
15181 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15182 IFN_MASK_SCATTER_STORE.
15183 (internal_fn_mask_index): Likewise.
15184 (internal_fn_stored_value_index): New function.
15185 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15186 for scatter stores.
15187 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15188 * optabs-query.c (supports_vec_scatter_store_p): New function.
15189 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15190 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15191 true for scatter stores.
15192 (vect_gather_scatter_fn_p): Handle scatter stores too.
15193 (vect_check_gather_scatter): Consider using scatter stores if
15194 supports_vec_scatter_store_p.
15195 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15196 scatter stores too.
15197 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15198 internal_fn_stored_value_index.
15199 (check_load_store_masking): Handle scatter stores too.
15200 (vect_get_store_rhs): Make public.
15201 (vectorizable_call): Use internal_store_fn_p.
15202 (vectorizable_store): Handle scatter store internal functions.
15203 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15204 when deciding whether the end of the group has been reached.
15205 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15206 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15207 (mask_scatter_store<mode>): New insns.
15209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15210 Alan Hayward <alan.hayward@arm.com>
15211 David Sherwood <david.sherwood@arm.com>
15213 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15214 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15215 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15217 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15218 Use vect_truncate_gather_scatter_offset if we can't treat the
15219 operation as a normal gather load or scatter store.
15220 (get_group_load_store_type): Take the gather_scatter_info
15221 as argument. Try using a gather load or scatter store for
15222 single-element groups.
15223 (get_load_store_type): Update calls to get_group_load_store_type
15224 and vect_use_strided_gather_scatters_p.
15226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15227 Alan Hayward <alan.hayward@arm.com>
15228 David Sherwood <david.sherwood@arm.com>
15230 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15231 optional tree argument.
15232 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15234 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15235 but continue to use the current value as a fallback.
15236 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15237 to compare the updates.
15238 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15239 (get_load_store_type): Use it when handling a strided access.
15240 (vect_get_strided_load_store_ops): New function.
15241 (vect_get_data_ptr_increment): Likewise.
15242 (vectorizable_load): Handle strided gather loads. Always pass
15243 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15246 Alan Hayward <alan.hayward@arm.com>
15247 David Sherwood <david.sherwood@arm.com>
15249 * doc/md.texi (gather_load@var{m}): Document.
15250 (mask_gather_load@var{m}): Likewise.
15251 * genopinit.c (main): Add supports_vec_gather_load and
15252 supports_vec_gather_load_cached to target_optabs.
15253 * optabs-tree.c (init_tree_optimization_optabs): Use
15254 ggc_cleared_alloc to allocate target_optabs.
15255 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15256 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15258 * internal-fn.h (internal_load_fn_p): Declare.
15259 (internal_gather_scatter_fn_p): Likewise.
15260 (internal_fn_mask_index): Likewise.
15261 (internal_gather_scatter_fn_supported_p): Likewise.
15262 * internal-fn.c (gather_load_direct): New macro.
15263 (expand_gather_load_optab_fn): New function.
15264 (direct_gather_load_optab_supported_p): New macro.
15265 (direct_internal_fn_optab): New function.
15266 (internal_load_fn_p): Likewise.
15267 (internal_gather_scatter_fn_p): Likewise.
15268 (internal_fn_mask_index): Likewise.
15269 (internal_gather_scatter_fn_supported_p): Likewise.
15270 * optabs-query.c (supports_at_least_one_mode_p): New function.
15271 (supports_vec_gather_load_p): Likewise.
15272 * optabs-query.h (supports_vec_gather_load_p): Declare.
15273 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15274 and memory_type field.
15275 (NUM_PATTERNS): Bump to 15.
15276 * tree-vect-data-refs.c: Include internal-fn.h.
15277 (vect_gather_scatter_fn_p): New function.
15278 (vect_describe_gather_scatter_call): Likewise.
15279 (vect_check_gather_scatter): Try using internal functions for
15280 gather loads. Recognize existing calls to a gather load function.
15281 (vect_analyze_data_refs): Consider using gather loads if
15282 supports_vec_gather_load_p.
15283 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15284 (vect_get_gather_scatter_offset_type): Likewise.
15285 (vect_convert_mask_for_vectype): Likewise.
15286 (vect_add_conversion_to_patterm): Likewise.
15287 (vect_try_gather_scatter_pattern): Likewise.
15288 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15289 (vect_vect_recog_func_ptrs): Add it.
15290 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15291 internal_fn_mask_index and internal_gather_scatter_fn_p.
15292 (check_load_store_masking): Take the gather_scatter_info as an
15293 argument and handle gather loads.
15294 (vect_get_gather_scatter_ops): New function.
15295 (vectorizable_call): Check internal_load_fn_p.
15296 (vectorizable_load): Likewise. Handle gather load internal
15298 (vectorizable_store): Update call to check_load_store_masking.
15299 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15300 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15301 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15302 (aarch64_gather_scale_operand_d): New predicates.
15303 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15304 (mask_gather_load<mode>): New insns.
15306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15307 Alan Hayward <alan.hayward@arm.com>
15308 David Sherwood <david.sherwood@arm.com>
15310 * optabs.def (fold_left_plus_optab): New optab.
15311 * doc/md.texi (fold_left_plus_@var{m}): Document.
15312 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15313 * internal-fn.c (fold_left_direct): Define.
15314 (expand_fold_left_optab_fn): Likewise.
15315 (direct_fold_left_optab_supported_p): Likewise.
15316 * fold-const-call.c (fold_const_fold_left): New function.
15317 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15318 * tree-parloops.c (valid_reduction_p): New function.
15319 (gather_scalar_reductions): Use it.
15320 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15321 (vect_finish_replace_stmt): Declare.
15322 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15323 (needs_fold_left_reduction_p): New function, split out from...
15324 (vect_is_simple_reduction): ...here. Accept reductions that
15325 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15326 (vect_force_simple_reduction): Also store the reduction type in
15327 the assignment's STMT_VINFO_REDUC_TYPE.
15328 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15329 (merge_with_identity): New function.
15330 (vect_expand_fold_left): Likewise.
15331 (vectorize_fold_left_reduction): Likewise.
15332 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15333 scalar phi in place for it. Check for target support and reject
15334 cases that would reassociate the operation. Defer the transform
15335 phase to vectorize_fold_left_reduction.
15336 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15337 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15338 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15342 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15343 call to ifc_temp_var.
15345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15346 Alan Hayward <alan.hayward@arm.com>
15347 David Sherwood <david.sherwood@arm.com>
15349 * target.def (legitimize_address_displacement): Take the original
15350 offset as a poly_int.
15351 * targhooks.h (default_legitimize_address_displacement): Update
15353 * targhooks.c (default_legitimize_address_displacement): Likewise.
15354 * doc/tm.texi: Regenerate.
15355 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15356 as an argument, moving assert of ad->disp == ad->disp_term to...
15357 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15358 Try calling targetm.legitimize_address_displacement before expanding
15359 the address rather than afterwards, and adjust for the new interface.
15360 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15361 Match the new hook interface. Handle SVE addresses.
15362 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15363 new hook interface.
15365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15367 * Makefile.in (OBJS): Add early-remat.o.
15368 * target.def (select_early_remat_modes): New hook.
15369 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15370 * doc/tm.texi: Regenerate.
15371 * targhooks.h (default_select_early_remat_modes): Declare.
15372 * targhooks.c (default_select_early_remat_modes): New function.
15373 * timevar.def (TV_EARLY_REMAT): New timevar.
15374 * passes.def (pass_early_remat): New pass.
15375 * tree-pass.h (make_pass_early_remat): Declare.
15376 * early-remat.c: New file.
15377 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15379 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15382 Alan Hayward <alan.hayward@arm.com>
15383 David Sherwood <david.sherwood@arm.com>
15385 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15386 vfm1 with a bound_epilog parameter.
15387 (vect_do_peeling): Update calls accordingly, and move the prologue
15388 call earlier in the function. Treat the base bound_epilog as 0 for
15389 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15390 this base when peeling for gaps.
15391 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15392 with fully-masked loops.
15393 (vect_estimate_min_profitable_iters): Handle the single peeled
15394 iteration in that case.
15396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15397 Alan Hayward <alan.hayward@arm.com>
15398 David Sherwood <david.sherwood@arm.com>
15400 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15401 single-element interleaving even if the size is not a power of 2.
15402 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15403 accesses for single-element interleaving if the group size is
15406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15407 Alan Hayward <alan.hayward@arm.com>
15408 David Sherwood <david.sherwood@arm.com>
15410 * doc/md.texi (fold_extract_last_@var{m}): Document.
15411 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15412 * optabs.def (fold_extract_last_optab): New optab.
15413 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15414 * internal-fn.c (fold_extract_direct): New macro.
15415 (expand_fold_extract_optab_fn): Likewise.
15416 (direct_fold_extract_optab_supported_p): Likewise.
15417 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15418 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15419 EXTRACT_LAST_REDUCTION.
15420 (get_initial_def_for_reduction): Do not create an initial vector
15421 for EXTRACT_LAST_REDUCTION reductions.
15422 (vectorizable_reduction): Leave the scalar phi in place for
15423 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15424 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15425 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15426 transform phase to vectorizable_condition.
15427 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15429 (vect_finish_stmt_generation): ...here.
15430 (vect_finish_replace_stmt): New function.
15431 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15432 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15434 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15437 Alan Hayward <alan.hayward@arm.com>
15438 David Sherwood <david.sherwood@arm.com>
15440 * doc/md.texi (extract_last_@var{m}): Document.
15441 * optabs.def (extract_last_optab): New optab.
15442 * internal-fn.def (EXTRACT_LAST): New internal function.
15443 * internal-fn.c (cond_unary_direct): New macro.
15444 (expand_cond_unary_optab_fn): Likewise.
15445 (direct_cond_unary_optab_supported_p): Likewise.
15446 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15447 loops using EXTRACT_LAST.
15448 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15449 (extract_last_<mode>): ...this optab.
15450 (vec_extract<mode><Vel>): Update accordingly.
15452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15453 Alan Hayward <alan.hayward@arm.com>
15454 David Sherwood <david.sherwood@arm.com>
15456 * target.def (empty_mask_is_expensive): New hook.
15457 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15458 * doc/tm.texi: Regenerate.
15459 * targhooks.h (default_empty_mask_is_expensive): Declare.
15460 * targhooks.c (default_empty_mask_is_expensive): New function.
15461 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15462 if the target says that empty masks are expensive.
15463 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15465 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15468 Alan Hayward <alan.hayward@arm.com>
15469 David Sherwood <david.sherwood@arm.com>
15471 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15472 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15473 (vect_use_loop_mask_for_alignment_p): New function.
15474 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15475 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15476 niters_skip argument. Make sure that the first niters_skip elements
15477 of the first iteration are inactive.
15478 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15479 Update call to vect_set_loop_masks_directly.
15480 (get_misalign_in_elems): New function, split out from...
15481 (vect_gen_prolog_loop_niters): ...here.
15482 (vect_update_init_of_dr): Take a code argument that specifies whether
15483 the adjustment should be added or subtracted.
15484 (vect_update_init_of_drs): Likewise.
15485 (vect_prepare_for_masked_peels): New function.
15486 (vect_do_peeling): Skip prologue peeling if we're using a mask
15487 instead. Update call to vect_update_inits_of_drs.
15488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15490 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15491 alignment. Do not include the number of peeled iterations in
15492 the minimum threshold in that case.
15493 (vectorizable_induction): Adjust the start value down by
15494 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15495 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15496 Take the number of skipped iterations into account when calculating
15498 * tree-vect-stmts.c (vect_gen_while_not): New function.
15500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15501 Alan Hayward <alan.hayward@arm.com>
15502 David Sherwood <david.sherwood@arm.com>
15504 * doc/sourcebuild.texi (vect_fully_masked): Document.
15505 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15506 default value to 0.
15507 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15509 (vect_analyze_loop_2): ...here. Don't check the vectorization
15510 factor against the number of loop iterations if the loop is
15513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15514 Alan Hayward <alan.hayward@arm.com>
15515 David Sherwood <david.sherwood@arm.com>
15517 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15518 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15519 (dump_groups): Update accordingly.
15520 (iv_use::mem_type): New member variable.
15521 (address_p): New function.
15522 (record_use): Add a mem_type argument and initialize the new
15524 (record_group_use): Add a mem_type argument. Use address_p.
15525 Remove obsolete null checks of base_object. Update call to record_use.
15526 (find_interesting_uses_op): Update call to record_group_use.
15527 (find_interesting_uses_cond): Likewise.
15528 (find_interesting_uses_address): Likewise.
15529 (get_mem_type_for_internal_fn): New function.
15530 (find_address_like_use): Likewise.
15531 (find_interesting_uses_stmt): Try find_address_like_use before
15532 calling find_interesting_uses_op.
15533 (addr_offset_valid_p): Use the iv mem_type field as the type
15534 of the addressed memory.
15535 (add_autoinc_candidates): Likewise.
15536 (get_address_cost): Likewise.
15537 (split_small_address_groups_p): Use address_p.
15538 (split_address_groups): Likewise.
15539 (add_iv_candidate_for_use): Likewise.
15540 (autoinc_possible_for_pair): Likewise.
15541 (rewrite_groups): Likewise.
15542 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15543 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15544 (get_alias_ptr_type_for_ptr_address): New function.
15545 (rewrite_use_address): Rewrite address uses in calls that were
15546 identified by find_address_like_use.
15548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15549 Alan Hayward <alan.hayward@arm.com>
15550 David Sherwood <david.sherwood@arm.com>
15552 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15554 * gimple-expr.h (is_gimple_addressable: Likewise.
15555 * gimple-expr.c (is_gimple_address): Likewise.
15556 * internal-fn.c (expand_call_mem_ref): New function.
15557 (expand_mask_load_optab_fn): Use it.
15558 (expand_mask_store_optab_fn): Likewise.
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 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15565 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15566 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15567 (cond_umax@var{mode}): Document.
15568 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15569 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15570 (cond_umin_optab, cond_umax_optab): New optabs.
15571 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15572 (COND_IOR, COND_XOR): New internal functions.
15573 * internal-fn.h (get_conditional_internal_fn): Declare.
15574 * internal-fn.c (cond_binary_direct): New macro.
15575 (expand_cond_binary_optab_fn): Likewise.
15576 (direct_cond_binary_optab_supported_p): Likewise.
15577 (get_conditional_internal_fn): New function.
15578 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15579 Cope with reduction statements that are vectorized as calls rather
15581 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15582 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15583 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15584 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15585 (UNSPEC_COND_EOR): New unspecs.
15586 (optab): Add mappings for them.
15587 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15588 (sve_int_op, sve_fp_op): New int attributes.
15590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15591 Alan Hayward <alan.hayward@arm.com>
15592 David Sherwood <david.sherwood@arm.com>
15594 * optabs.def (while_ult_optab): New optab.
15595 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15596 * internal-fn.def (WHILE_ULT): New internal function.
15597 * internal-fn.h (direct_internal_fn_supported_p): New override
15598 that takes two types as argument.
15599 * internal-fn.c (while_direct): New macro.
15600 (expand_while_optab_fn): New function.
15601 (convert_optab_supported_p): Likewise.
15602 (direct_while_optab_supported_p): New macro.
15603 * wide-int.h (wi::udiv_ceil): New function.
15604 * tree-vectorizer.h (rgroup_masks): New structure.
15605 (vec_loop_masks): New typedef.
15606 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15607 and fully_masked_p.
15608 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15609 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15610 (vect_max_vf): New function.
15611 (slpeel_make_loop_iterate_ntimes): Delete.
15612 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15613 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15614 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15615 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15616 internal-fn.h, stor-layout.h and optabs-query.h.
15617 (vect_set_loop_mask): New function.
15618 (add_preheader_seq): Likewise.
15619 (add_header_seq): Likewise.
15620 (interleave_supported_p): Likewise.
15621 (vect_maybe_permute_loop_masks): Likewise.
15622 (vect_set_loop_masks_directly): Likewise.
15623 (vect_set_loop_condition_masked): Likewise.
15624 (vect_set_loop_condition_unmasked): New function, split out from
15625 slpeel_make_loop_iterate_ntimes.
15626 (slpeel_make_loop_iterate_ntimes): Rename to..
15627 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15628 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15629 (vect_do_peeling): Update call accordingly.
15630 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15632 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15633 mask_compare_type, can_fully_mask_p and fully_masked_p.
15634 (release_vec_loop_masks): New function.
15635 (_loop_vec_info): Use it to free the loop masks.
15636 (can_produce_all_loop_masks_p): New function.
15637 (vect_get_max_nscalars_per_iter): Likewise.
15638 (vect_verify_full_masking): Likewise.
15639 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15640 retries, and free the mask rgroups before retrying. Check loop-wide
15641 reasons for disallowing fully-masked loops. Make the final decision
15642 about whether use a fully-masked loop or not.
15643 (vect_estimate_min_profitable_iters): Do not assume that peeling
15644 for the number of iterations will be needed for fully-masked loops.
15645 (vectorizable_reduction): Disable fully-masked loops.
15646 (vectorizable_live_operation): Likewise.
15647 (vect_halve_mask_nunits): New function.
15648 (vect_double_mask_nunits): Likewise.
15649 (vect_record_loop_mask): Likewise.
15650 (vect_get_loop_mask): Likewise.
15651 (vect_transform_loop): Handle the case in which the final loop
15652 iteration might handle a partial vector. Call vect_set_loop_condition
15653 instead of slpeel_make_loop_iterate_ntimes.
15654 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15655 (check_load_store_masking): New function.
15656 (prepare_load_store_mask): Likewise.
15657 (vectorizable_store): Handle fully-masked loops.
15658 (vectorizable_load): Likewise.
15659 (supportable_widening_operation): Use vect_halve_mask_nunits for
15661 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15662 (vect_gen_while): New function.
15663 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15664 (aarch64_uqdec<mode>): New insn.
15666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15667 Alan Hayward <alan.hayward@arm.com>
15668 David Sherwood <david.sherwood@arm.com>
15670 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15671 (reduc_xor_scal_optab): New optabs.
15672 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15673 (reduc_xor_scal_@var{m}): Document.
15674 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15675 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15676 internal functions.
15677 * fold-const-call.c (fold_const_call): Handle them.
15678 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15679 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15680 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15681 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15682 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15683 (UNSPEC_XORV): New unspecs.
15684 (optab): Add entries for them.
15685 (BITWISEV): New int iterator.
15686 (bit_reduc_op): New int attributes.
15688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15689 Alan Hayward <alan.hayward@arm.com>
15690 David Sherwood <david.sherwood@arm.com>
15692 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15693 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15694 * optabs.def (vec_shl_insert_optab): New optab.
15695 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15696 (duplicate_and_interleave): Likewise.
15697 * tree-vect-loop.c: Include internal-fn.h.
15698 (neutral_op_for_slp_reduction): New function, split out from
15699 get_initial_defs_for_reduction.
15700 (get_initial_def_for_reduction): Handle option 2 for variable-length
15701 vectors by loading the neutral value into a vector and then shifting
15702 the initial value into element 0.
15703 (get_initial_defs_for_reduction): Replace the code argument with
15704 the neutral value calculated by neutral_op_for_slp_reduction.
15705 Use gimple_build_vector for constant-length vectors.
15706 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15707 but the first group_size elements have a neutral value.
15708 Use duplicate_and_interleave otherwise.
15709 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15710 Update call to get_initial_defs_for_reduction. Handle SLP
15711 reductions for variable-length vectors by creating one vector
15712 result for each scalar result, with the elements associated
15713 with other scalar results stubbed out with the neutral value.
15714 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15715 Require IFN_VEC_SHL_INSERT for double reductions on
15716 variable-length vectors, or SLP reductions that have
15717 a neutral value. Require can_duplicate_and_interleave_p
15718 support for variable-length unchained SLP reductions if there
15719 is no neutral value, such as for MIN/MAX reductions. Also require
15720 the number of vector elements to be a multiple of the number of
15721 SLP statements when doing variable-length unchained SLP reductions.
15722 Update call to vect_create_epilog_for_reduction.
15723 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15724 and remove initial values.
15725 (duplicate_and_interleave): Make public.
15726 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15727 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15730 Alan Hayward <alan.hayward@arm.com>
15731 David Sherwood <david.sherwood@arm.com>
15733 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15734 (can_duplicate_and_interleave_p): New function.
15735 (vect_get_and_check_slp_defs): Take the vector of statements
15736 rather than just the current one. Remove excess parentheses.
15737 Restriction rejectinon of vect_constant_def and vect_external_def
15738 for variable-length vectors to boolean types, or types for which
15739 can_duplicate_and_interleave_p is false.
15740 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15741 (duplicate_and_interleave): New function.
15742 (vect_get_constant_vectors): Use gimple_build_vector for
15743 constant-length vectors and suitable variable-length constant
15744 vectors. Use duplicate_and_interleave for other variable-length
15745 vectors. Don't defer the update when inserting new statements.
15747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15748 Alan Hayward <alan.hayward@arm.com>
15749 David Sherwood <david.sherwood@arm.com>
15751 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15752 min_profitable_iters doesn't go negative.
15754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15755 Alan Hayward <alan.hayward@arm.com>
15756 David Sherwood <david.sherwood@arm.com>
15758 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15759 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15760 * optabs.def (vec_mask_load_lanes_optab): New optab.
15761 (vec_mask_store_lanes_optab): Likewise.
15762 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15763 (MASK_STORE_LANES): Likewise.
15764 * internal-fn.c (mask_load_lanes_direct): New macro.
15765 (mask_store_lanes_direct): Likewise.
15766 (expand_mask_load_optab_fn): Handle masked operations.
15767 (expand_mask_load_lanes_optab_fn): New macro.
15768 (expand_mask_store_optab_fn): Handle masked operations.
15769 (expand_mask_store_lanes_optab_fn): New macro.
15770 (direct_mask_load_lanes_optab_supported_p): Likewise.
15771 (direct_mask_store_lanes_optab_supported_p): Likewise.
15772 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15774 (vect_load_lanes_supported): Likewise.
15775 * tree-vect-data-refs.c (strip_conversion): New function.
15776 (can_group_stmts_p): Likewise.
15777 (vect_analyze_data_ref_accesses): Use it instead of checking
15778 for a pair of assignments.
15779 (vect_store_lanes_supported): Take a masked_p parameter.
15780 (vect_load_lanes_supported): Likewise.
15781 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15782 vect_store_lanes_supported and vect_load_lanes_supported.
15783 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15784 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15785 parameter. Don't allow gaps for masked accesses.
15786 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15787 and vect_load_lanes_supported.
15788 (get_load_store_type): Take a masked_p parameter and update
15789 call to get_group_load_store_type.
15790 (vectorizable_store): Update call to get_load_store_type.
15791 Handle IFN_MASK_STORE_LANES.
15792 (vectorizable_load): Update call to get_load_store_type.
15793 Handle IFN_MASK_LOAD_LANES.
15795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15796 Alan Hayward <alan.hayward@arm.com>
15797 David Sherwood <david.sherwood@arm.com>
15799 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15801 * config/aarch64/aarch64-protos.h
15802 (aarch64_sve_struct_memory_operand_p): Declare.
15803 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15804 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15805 (VPRED, vpred): Handle SVE structure modes.
15806 * config/aarch64/constraints.md (Utx): New constraint.
15807 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15808 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15809 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15810 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15811 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15812 structure modes. Split into pieces after RA.
15813 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15814 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15816 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15817 SVE structure modes.
15818 (aarch64_classify_address): Likewise.
15819 (sizetochar): Move earlier in file.
15820 (aarch64_print_operand): Handle SVE register lists.
15821 (aarch64_array_mode): New function.
15822 (aarch64_sve_struct_memory_operand_p): Likewise.
15823 (TARGET_ARRAY_MODE): Redefine.
15825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15826 Alan Hayward <alan.hayward@arm.com>
15827 David Sherwood <david.sherwood@arm.com>
15829 * target.def (array_mode): New target hook.
15830 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15831 * doc/tm.texi: Regenerate.
15832 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15833 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15834 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15835 targetm.array_mode.
15836 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15840 Alan Hayward <alan.hayward@arm.com>
15841 David Sherwood <david.sherwood@arm.com>
15843 * fold-const.c (fold_binary_loc): Check the argument types
15844 rather than the result type when testing for a vector operation.
15846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15848 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15849 * doc/tm.texi: Regenerate.
15851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15852 Alan Hayward <alan.hayward@arm.com>
15853 David Sherwood <david.sherwood@arm.com>
15855 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15856 (sve): Document new AArch64 extension.
15857 * doc/md.texi (w): Extend the description of the AArch64
15858 constraint to include SVE vectors.
15859 (Upl, Upa): Document new AArch64 predicate constraints.
15860 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15862 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15863 (msve-vector-bits=): New option.
15864 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15865 SVE when these are disabled.
15866 (sve): New extension.
15867 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15868 modes. Adjust their number of units based on aarch64_sve_vg.
15869 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15870 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15871 aarch64_addr_query_type.
15872 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15873 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15874 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15875 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15876 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15877 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15878 (aarch64_simd_imm_zero_p): Delete.
15879 (aarch64_check_zero_based_sve_index_immediate): Declare.
15880 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15881 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15882 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15883 (aarch64_sve_float_mul_immediate_p): Likewise.
15884 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15885 rather than an rtx.
15886 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15887 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15888 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15889 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15890 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15891 (aarch64_regmode_natural_size): Likewise.
15892 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15893 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15895 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15896 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15897 for VG and the SVE predicate registers.
15898 (V_ALIASES): Add a "z"-prefixed alias.
15899 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15900 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15901 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15902 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15903 (REG_CLASS_NAMES): Add entries for them.
15904 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15905 and the predicate registers.
15906 (aarch64_sve_vg): Declare.
15907 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15908 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15909 (REGMODE_NATURAL_SIZE): Define.
15910 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15912 * config/aarch64/aarch64.c: Include cfgrtl.h.
15913 (simd_immediate_info): Add a constructor for series vectors,
15914 and an associated step field.
15915 (aarch64_sve_vg): New variable.
15916 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15917 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15918 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15919 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15920 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15921 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15922 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15923 (aarch64_get_mask_mode): New functions.
15924 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15925 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15926 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15927 predicate modes and predicate registers. Explicitly restrict
15928 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15929 to store a vector mode if it is recognized by
15930 aarch64_classify_vector_mode.
15931 (aarch64_regmode_natural_size): New function.
15932 (aarch64_hard_regno_caller_save_mode): Return the original mode
15934 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15935 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15936 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15937 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15939 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15940 does not overlap dest if the function is frame-related. Handle
15942 (aarch64_split_add_offset): New function.
15943 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15944 them aarch64_add_offset.
15945 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15946 and update call to aarch64_sub_sp.
15947 (aarch64_add_cfa_expression): New function.
15948 (aarch64_expand_prologue): Pass extra temporary registers to the
15949 functions above. Handle the case in which we need to emit new
15950 DW_CFA_expressions for registers that were originally saved
15951 relative to the stack pointer, but now have to be expressed
15952 relative to the frame pointer.
15953 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15955 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15956 IP0 and IP1 values for SVE frames.
15957 (aarch64_expand_vec_series): New function.
15958 (aarch64_expand_sve_widened_duplicate): Likewise.
15959 (aarch64_expand_sve_const_vector): Likewise.
15960 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15961 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15962 into the register, rather than emitting a SET directly.
15963 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15964 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15965 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15966 (offset_9bit_signed_scaled_p): New functions.
15967 (aarch64_replicate_bitmask_imm): New function.
15968 (aarch64_bitmask_imm): Use it.
15969 (aarch64_cannot_force_const_mem): Reject expressions involving
15970 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15971 (aarch64_classify_index): Handle SVE indices, by requiring
15972 a plain register index with a scale that matches the element size.
15973 (aarch64_classify_address): Handle SVE addresses. Assert that
15974 the mode of the address is VOIDmode or an integer mode.
15975 Update call to aarch64_classify_symbol.
15976 (aarch64_classify_symbolic_expression): Update call to
15977 aarch64_classify_symbol.
15978 (aarch64_const_vec_all_in_range_p): New function.
15979 (aarch64_print_vector_float_operand): Likewise.
15980 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15981 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15982 and the FP immediates 1.0 and 0.5.
15983 (aarch64_print_address_internal): Handle SVE addresses.
15984 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15985 (aarch64_regno_regclass): Handle predicate registers.
15986 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15988 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15989 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15990 (aarch64_convert_sve_vector_bits): New function.
15991 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15992 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15993 rather than an rtx.
15994 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15995 Handle SVE vector and predicate modes. Accept VL-based constants
15996 that need only one temporary register, and VL offsets that require
15997 no temporary registers.
15998 (aarch64_conditional_register_usage): Mark the predicate registers
15999 as fixed if SVE isn't available.
16000 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16001 Return true for SVE vector and predicate modes.
16002 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16003 rather than an unsigned int. Handle SVE modes.
16004 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16006 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16008 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16009 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16010 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16011 (aarch64_sve_float_mul_immediate_p): New functions.
16012 (aarch64_sve_valid_immediate): New function.
16013 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16014 Explicitly reject structure modes. Check for INDEX constants.
16015 Handle PTRUE and PFALSE constants.
16016 (aarch64_check_zero_based_sve_index_immediate): New function.
16017 (aarch64_simd_imm_zero_p): Delete.
16018 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16019 vector modes. Accept constants in the range of CNT[BHWD].
16020 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16021 ask for an Advanced SIMD mode.
16022 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16023 (aarch64_simd_vector_alignment): Handle SVE predicates.
16024 (aarch64_vectorize_preferred_vector_alignment): New function.
16025 (aarch64_simd_vector_alignment_reachable): Use it instead of
16027 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16028 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16030 (MAX_VECT_LEN): Delete.
16031 (expand_vec_perm_d): Add a vec_flags field.
16032 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16033 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16034 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16036 (aarch64_evpc_rev): Rename to...
16037 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16038 (aarch64_evpc_rev_global): New function.
16039 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16040 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16042 (aarch64_evpc_sve_tbl): New function.
16043 (aarch64_expand_vec_perm_const_1): Update after rename of
16044 aarch64_evpc_rev. Handle SVE permutes too, trying
16045 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16046 than aarch64_evpc_tbl.
16047 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16048 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16049 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16050 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16051 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16052 (aarch64_expand_sve_vcond): New functions.
16053 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16054 of aarch64_vector_mode_p.
16055 (aarch64_dwarf_poly_indeterminate_value): New function.
16056 (aarch64_compute_pressure_classes): Likewise.
16057 (aarch64_can_change_mode_class): Likewise.
16058 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16059 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16060 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16061 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16062 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16063 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16064 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16065 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16067 (Dn, Dl, Dr): Accept const as well as const_vector.
16068 (Dz): Likewise. Compare against CONST0_RTX.
16069 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16070 of "vector" where appropriate.
16071 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16072 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16073 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16074 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16075 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16076 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16077 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16078 (v_int_equiv): Extend to SVE modes.
16079 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16081 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16082 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16083 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16084 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16085 (SVE_COND_FP_CMP): New int iterators.
16086 (perm_hilo): Handle the new unpack unspecs.
16087 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16089 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16090 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16091 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16092 (aarch64_equality_operator, aarch64_constant_vector_operand)
16093 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16094 (aarch64_sve_nonimmediate_operand): Likewise.
16095 (aarch64_sve_general_operand): Likewise.
16096 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16097 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16098 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16099 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16100 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16101 (aarch64_sve_float_arith_immediate): Likewise.
16102 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16103 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16104 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16105 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16106 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16107 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16108 (aarch64_sve_float_arith_operand): Likewise.
16109 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16110 (aarch64_sve_float_mul_operand): Likewise.
16111 (aarch64_sve_vec_perm_operand): Likewise.
16112 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16113 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16114 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16115 as well as const_vector.
16116 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16117 in file. Use CONST0_RTX and CONSTM1_RTX.
16118 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16119 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16120 Use aarch64_simd_imm_zero.
16121 * config/aarch64/aarch64-sve.md: New file.
16122 * config/aarch64/aarch64.md: Include it.
16123 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16124 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16125 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16126 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16127 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16128 (sve): New attribute.
16129 (enabled): Disable instructions with the sve attribute unless
16131 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16132 aarch64_expand_mov_immediate.
16133 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16134 CNT[BHSD] immediates.
16135 (movti): Split CONST_POLY_INT moves into two halves.
16136 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16137 Split additions that need a temporary here if the destination
16138 is the stack pointer.
16139 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16140 (*add<mode>3_poly_1): New instruction.
16141 (set_clobber_cc): New expander.
16143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16145 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16146 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16147 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16148 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16149 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16150 Change innermode from fixed_mode_size to machine_mode.
16151 (simplify_subreg): Update call accordingly. Handle a constant-sized
16152 subreg of a variable-length CONST_VECTOR.
16154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16155 Alan Hayward <alan.hayward@arm.com>
16156 David Sherwood <david.sherwood@arm.com>
16158 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16159 (add_offset_to_base): New function, split out from...
16160 (create_mem_ref): ...here. When handling a scale other than 1,
16161 check first whether the address is valid without the offset.
16162 Add it into the base if so, leaving the index and scale as-is.
16164 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16167 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16168 fold_for_warn before checking if arg2 is INTEGER_CST.
16170 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16172 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16173 (store_multiple_operation): Delete.
16174 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16175 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16176 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16177 guarded by TARGET_STRING.
16178 (rs6000_output_load_multiple): Delete.
16179 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16180 OPTION_MASK_STRING / TARGET_STRING handling.
16181 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16182 (const rs6000_opt_masks) <"string">: Change mask to 0.
16183 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16184 (MASK_STRING): Delete.
16185 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16187 (load_multiple): Delete.
16194 (store_multiple): Delete.
16201 (movmemsi_8reg): Delete.
16202 (corresponding unnamed define_insn): Delete.
16203 (movmemsi_6reg): Delete.
16204 (corresponding unnamed define_insn): Delete.
16205 (movmemsi_4reg): Delete.
16206 (corresponding unnamed define_insn): Delete.
16207 (movmemsi_2reg): Delete.
16208 (corresponding unnamed define_insn): Delete.
16209 (movmemsi_1reg): Delete.
16210 (corresponding unnamed define_insn): Delete.
16211 * config/rs6000/rs6000.opt (mno-string): New.
16212 (mstring): Replace by deprecation warning stub.
16213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16215 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16217 * regrename.c (regrename_do_replace): If replacing the same
16218 reg multiple times, try to reuse last created gen_raw_REG.
16221 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16222 main to workaround a bug in GDB.
16224 2018-01-12 Tom de Vries <tom@codesourcery.com>
16227 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16229 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16231 PR rtl-optimization/80481
16232 * ira-color.c (get_cap_member): New function.
16233 (allocnos_conflict_by_live_ranges_p): Use it.
16234 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16235 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16237 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16240 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16241 (*saddl_se_1): Ditto.
16242 (*ssubsi_1): Ditto.
16243 (*ssubl_se_1): Ditto.
16245 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16247 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16248 rather than wi::to_widest for DR_INITs.
16249 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16250 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16251 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16253 (vect_analyze_group_access_1): Note that here.
16255 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16257 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16258 polynomial type sizes.
16260 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16262 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16263 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16264 (gimple_add_tmp_var): Likewise.
16266 2018-01-12 Martin Liska <mliska@suse.cz>
16268 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16269 (gimple_alloc_sizes): Likewise.
16270 (dump_gimple_statistics): Use PRIu64 in printf format.
16271 * gimple.h: Change uint64_t to int.
16273 2018-01-12 Martin Liska <mliska@suse.cz>
16275 * tree-core.h: Use uint64_t instead of int.
16276 * tree.c (tree_node_counts): Likewise.
16277 (tree_node_sizes): Likewise.
16278 (dump_tree_statistics): Use PRIu64 in printf format.
16280 2018-01-12 Martin Liska <mliska@suse.cz>
16282 * Makefile.in: As qsort_chk is implemented in vec.c, add
16283 vec.o to linkage of gencfn-macros.
16284 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16285 passing the info to record_node_allocation_statistics.
16286 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16288 * ggc-common.c (struct ggc_usage): Add operator== and use
16289 it in operator< and compare function.
16290 * mem-stats.h (struct mem_usage): Likewise.
16291 * vec.c (struct vec_usage): Remove operator< and compare
16292 function. Can be simply inherited.
16294 2018-01-12 Martin Jambor <mjambor@suse.cz>
16297 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16298 * tree-ssa-math-opts.c: Include domwalk.h.
16299 (convert_mult_to_fma_1): New function.
16300 (fma_transformation_info): New type.
16301 (fma_deferring_state): Likewise.
16302 (cancel_fma_deferring): New function.
16303 (result_of_phi): Likewise.
16304 (last_fma_candidate_feeds_initial_phi): Likewise.
16305 (convert_mult_to_fma): Added deferring logic, split actual
16306 transformation to convert_mult_to_fma_1.
16307 (math_opts_dom_walker): New type.
16308 (math_opts_dom_walker::after_dom_children): New method, body moved
16309 here from pass_optimize_widening_mul::execute, added deferring logic
16311 (pass_optimize_widening_mul::execute): Moved most of code to
16312 math_opts_dom_walker::after_dom_children.
16313 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16314 * config/i386/i386.c (ix86_option_override_internal): Added
16315 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16317 2018-01-12 Richard Biener <rguenther@suse.de>
16320 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16321 inline instance vars.
16323 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16326 * config/rx/rx.c (rx_is_restricted_memory_address):
16327 Handle SUBREG case.
16329 2018-01-12 Richard Biener <rguenther@suse.de>
16331 PR tree-optimization/80846
16332 * target.def (split_reduction): New target hook.
16333 * targhooks.c (default_split_reduction): New function.
16334 * targhooks.h (default_split_reduction): Declare.
16335 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16336 target requests first reduce vectors by combining low and high
16338 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16339 (get_vectype_for_scalar_type_and_size): Export.
16340 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16341 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16342 * doc/tm.texi: Regenerate.
16343 * config/i386/i386.c (ix86_split_reduction): Implement
16344 TARGET_VECTORIZE_SPLIT_REDUCTION.
16346 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16349 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16350 in PIC mode except for TARGET_VXWORKS_RTP.
16351 * config/sparc/sparc.c: Include cfgrtl.h.
16352 (TARGET_INIT_PIC_REG): Define.
16353 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16354 (sparc_pic_register_p): New predicate.
16355 (sparc_legitimate_address_p): Use it.
16356 (sparc_legitimize_pic_address): Likewise.
16357 (sparc_delegitimize_address): Likewise.
16358 (sparc_mode_dependent_address_p): Likewise.
16359 (gen_load_pcrel_sym): Remove 4th parameter.
16360 (load_got_register): Adjust call to above. Remove obsolete stuff.
16361 (sparc_expand_prologue): Do not call load_got_register here.
16362 (sparc_flat_expand_prologue): Likewise.
16363 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16364 (sparc_use_pseudo_pic_reg): New function.
16365 (sparc_init_pic_reg): Likewise.
16366 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16367 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16369 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16371 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16372 Add item for branch_cost.
16374 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16376 PR rtl-optimization/83565
16377 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16378 not extend the result to a larger mode for rotate operations.
16379 (num_sign_bit_copies1): Likewise.
16381 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16384 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16386 Use values-Xc.o for -pedantic.
16387 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16389 2018-01-12 Martin Liska <mliska@suse.cz>
16392 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16394 (possible_polymorphic_call_targets): Use it.
16395 (ipa_devirt): Likewise.
16397 2018-01-12 Martin Liska <mliska@suse.cz>
16399 * profile-count.h (enum profile_quality): Use 0 as invalid
16400 enum value of profile_quality.
16402 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16404 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16405 -mext-string options.
16407 2018-01-12 Richard Biener <rguenther@suse.de>
16409 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16410 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16411 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16413 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16415 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16417 * configure.ac (--with-long-double-format): Add support for the
16418 configuration option to change the default long double format on
16420 * config.gcc (powerpc*-linux*-*): Likewise.
16421 * configure: Regenerate.
16422 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16423 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16424 used without modification.
16426 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16428 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16429 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16430 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16431 MISC_BUILTIN_SPEC_BARRIER.
16432 (rs6000_init_builtins): Likewise.
16433 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16435 (speculation_barrier): New define_insn.
16436 * doc/extend.texi: Document __builtin_speculation_barrier.
16438 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16441 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16442 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16443 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16445 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16446 integral modes instead of "ss" and "sd".
16447 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16448 vectors with 32-bit and 64-bit elements.
16449 (vecdupssescalarmodesuffix): New mode attribute.
16450 (vec_dup<mode>): Use it.
16452 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16455 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16456 frame if argument is passed on stack.
16458 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16461 * ree.c (combine_reaching_defs): Optimize also
16462 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16463 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16465 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16467 PR middle-end/83189
16468 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16470 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16472 PR middle-end/83718
16473 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16474 after they are computed.
16476 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16478 PR tree-optimization/83695
16479 * gimple-loop-linterchange.cc
16480 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16481 reset cached scev information after interchange.
16482 (pass_linterchange::execute): Remove call to scev_reset_htab.
16484 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16486 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16487 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16488 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16489 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16490 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16491 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16492 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16493 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16494 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16495 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16496 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16497 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16498 (V_lane_reg): Likewise.
16499 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16501 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16502 (vfmal_lane_low<mode>_intrinsic,
16503 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16504 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16505 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16506 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16507 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16508 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16510 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16512 * config/arm/arm-cpus.in (fp16fml): New feature.
16513 (ALL_SIMD): Add fp16fml.
16514 (armv8.2-a): Add fp16fml as an option.
16515 (armv8.3-a): Likewise.
16516 (armv8.4-a): Add fp16fml as part of fp16.
16517 * config/arm/arm.h (TARGET_FP16FML): Define.
16518 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16520 * config/arm/arm-modes.def (V2HF): Define.
16521 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16522 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16523 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16524 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16525 vfmsl_low, vfmsl_high): New set of builtins.
16526 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16527 (vfml_op): New code attribute.
16528 (VFMLHALVES): New int iterator.
16529 (VFML, VFMLSEL): New mode attributes.
16530 (V_reg): Define mapping for V2HF.
16531 (V_hi, V_lo): New mode attributes.
16532 (VF_constraint): Likewise.
16533 (vfml_half, vfml_half_selector): New int attributes.
16534 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16536 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16537 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16539 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16540 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16541 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16542 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16544 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16545 Document new effective target and option set.
16547 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16549 * config/arm/arm-cpus.in (armv8_4): New feature.
16550 (ARMv8_4a): New fgroup.
16551 (armv8.4-a): New arch.
16552 * config/arm/arm-tables.opt: Regenerate.
16553 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16554 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16555 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16556 Add matching rules for -march=armv8.4-a and extensions.
16557 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16559 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16562 * config/rx/rx.md (BW): New mode attribute.
16563 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16565 2018-01-11 Richard Biener <rguenther@suse.de>
16567 PR tree-optimization/83435
16568 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16569 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16570 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16573 Alan Hayward <alan.hayward@arm.com>
16574 David Sherwood <david.sherwood@arm.com>
16576 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16578 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16579 (aarch64_print_address_internal): Use it to check for a zero offset.
16581 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16582 Alan Hayward <alan.hayward@arm.com>
16583 David Sherwood <david.sherwood@arm.com>
16585 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16586 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16587 Return a poly_int64 rather than a HOST_WIDE_INT.
16588 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16589 rather than a HOST_WIDE_INT.
16590 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16591 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16592 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16593 final_offset from HOST_WIDE_INT to poly_int64.
16594 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16595 to_constant when getting the number of units in an Advanced SIMD
16597 (aarch64_builtin_vectorized_function): Check for a constant number
16599 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16601 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16602 attribute instead of GET_MODE_NUNITS.
16603 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16604 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16605 GET_MODE_SIZE for fixed-size registers.
16606 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16607 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16608 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16609 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16610 (aarch64_print_operand, aarch64_print_address_internal)
16611 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16612 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16613 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16614 Handle polynomial GET_MODE_SIZE.
16615 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16616 wider than SImode without modification.
16617 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16618 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16619 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16620 passing and returning SVE modes.
16621 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16622 rather than GEN_INT.
16623 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16624 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16625 (aarch64_allocate_and_probe_stack_space): Likewise.
16626 (aarch64_layout_frame): Cope with polynomial offsets.
16627 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16628 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16629 polynomial offsets.
16630 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16631 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16632 poly_int64 rather than a HOST_WIDE_INT.
16633 (aarch64_get_separate_components, aarch64_process_components)
16634 (aarch64_expand_prologue, aarch64_expand_epilogue)
16635 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16636 (aarch64_anchor_offset): New function, split out from...
16637 (aarch64_legitimize_address): ...here.
16638 (aarch64_builtin_vectorization_cost): Handle polynomial
16639 TYPE_VECTOR_SUBPARTS.
16640 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16642 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16643 number of elements from the PARALLEL rather than the mode.
16644 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16645 rather than GET_MODE_BITSIZE.
16646 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16647 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16648 (aarch64_expand_vec_perm_const_1): Handle polynomial
16649 d->perm.length () and d->perm elements.
16650 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16651 Apply to_constant to d->perm elements.
16652 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16653 polynomial CONST_VECTOR_NUNITS.
16654 (aarch64_move_pointer): Take amount as a poly_int64 rather
16656 (aarch64_progress_pointer): Avoid temporary variable.
16657 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16658 the mode attribute instead of GET_MODE.
16660 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16661 Alan Hayward <alan.hayward@arm.com>
16662 David Sherwood <david.sherwood@arm.com>
16664 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16665 x exists before using it.
16666 (aarch64_add_constant_internal): Rename to...
16667 (aarch64_add_offset_1): ...this. Replace regnum with separate
16668 src and dest rtxes. Handle the case in which they're different,
16669 including when the offset is zero. Replace scratchreg with an rtx.
16670 Use 2 additions if there is no spare register into which we can
16671 move a 16-bit constant.
16672 (aarch64_add_constant): Delete.
16673 (aarch64_add_offset): Replace reg with separate src and dest
16674 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16675 Use aarch64_add_offset_1.
16676 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16677 an rtx rather than an int. Take the delta as a poly_int64
16678 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16679 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16680 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16681 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16682 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16683 and aarch64_add_sp.
16684 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16685 aarch64_add_constant.
16687 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16689 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16690 Use scalar_float_mode.
16692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16694 * config/aarch64/aarch64-simd.md
16695 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16696 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16697 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16698 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16699 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16700 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16701 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16702 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16703 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16704 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16706 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16709 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16710 targ_options->x_arm_arch_string is non NULL.
16712 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16714 * config/aarch64/aarch64.h
16715 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16717 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16720 * expmed.c (emit_store_flag_force): Swap if const op0
16721 and change VOIDmode to mode of op0.
16723 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16725 PR rtl-optimization/83761
16726 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16727 than bytes to mode_for_size.
16729 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16731 PR middle-end/83189
16732 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16733 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16736 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16738 PR middle-end/83575
16739 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16740 when in layout mode.
16741 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16742 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16745 2018-01-10 Michael Collison <michael.collison@arm.com>
16747 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16748 * config/aarch64/aarch64-option-extension.def: Add
16749 AARCH64_OPT_EXTENSION of 'fp16fml'.
16750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16751 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16752 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16753 * config/aarch64/constraints.md (Ui7): New constraint.
16754 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16755 (VFMLA_SEL_W): Ditto.
16758 (VFMLA16_LOW): New int iterator.
16759 (VFMLA16_HIGH): Ditto.
16760 (UNSPEC_FMLAL): New unspec.
16761 (UNSPEC_FMLSL): Ditto.
16762 (UNSPEC_FMLAL2): Ditto.
16763 (UNSPEC_FMLSL2): Ditto.
16764 (f16mac): New code attribute.
16765 * config/aarch64/aarch64-simd-builtins.def
16766 (aarch64_fmlal_lowv2sf): Ditto.
16767 (aarch64_fmlsl_lowv2sf): Ditto.
16768 (aarch64_fmlalq_lowv4sf): Ditto.
16769 (aarch64_fmlslq_lowv4sf): Ditto.
16770 (aarch64_fmlal_highv2sf): Ditto.
16771 (aarch64_fmlsl_highv2sf): Ditto.
16772 (aarch64_fmlalq_highv4sf): Ditto.
16773 (aarch64_fmlslq_highv4sf): Ditto.
16774 (aarch64_fmlal_lane_lowv2sf): Ditto.
16775 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16776 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16777 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16778 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16779 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16780 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16781 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16782 (aarch64_fmlal_lane_highv2sf): Ditto.
16783 (aarch64_fmlsl_lane_highv2sf): Ditto.
16784 (aarch64_fmlal_laneq_highv2sf): Ditto.
16785 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16786 (aarch64_fmlalq_lane_highv4sf): Ditto.
16787 (aarch64_fmlsl_lane_highv4sf): Ditto.
16788 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16789 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16790 * config/aarch64/aarch64-simd.md:
16791 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16792 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16793 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16794 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16795 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16796 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16797 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16798 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16799 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16800 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16801 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16802 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16803 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16804 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16805 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16806 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16807 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16808 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16809 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16810 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16811 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16812 (vfmlsl_low_u32): Ditto.
16813 (vfmlalq_low_u32): Ditto.
16814 (vfmlslq_low_u32): Ditto.
16815 (vfmlal_high_u32): Ditto.
16816 (vfmlsl_high_u32): Ditto.
16817 (vfmlalq_high_u32): Ditto.
16818 (vfmlslq_high_u32): Ditto.
16819 (vfmlal_lane_low_u32): Ditto.
16820 (vfmlsl_lane_low_u32): Ditto.
16821 (vfmlal_laneq_low_u32): Ditto.
16822 (vfmlsl_laneq_low_u32): Ditto.
16823 (vfmlalq_lane_low_u32): Ditto.
16824 (vfmlslq_lane_low_u32): Ditto.
16825 (vfmlalq_laneq_low_u32): Ditto.
16826 (vfmlslq_laneq_low_u32): Ditto.
16827 (vfmlal_lane_high_u32): Ditto.
16828 (vfmlsl_lane_high_u32): Ditto.
16829 (vfmlal_laneq_high_u32): Ditto.
16830 (vfmlsl_laneq_high_u32): Ditto.
16831 (vfmlalq_lane_high_u32): Ditto.
16832 (vfmlslq_lane_high_u32): Ditto.
16833 (vfmlalq_laneq_high_u32): Ditto.
16834 (vfmlslq_laneq_high_u32): Ditto.
16835 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16836 (AARCH64_FL_FOR_ARCH8_4): New.
16837 (AARCH64_ISA_F16FML): New ISA flag.
16838 (TARGET_F16FML): New feature flag for fp16fml.
16839 (doc/invoke.texi): Document new fp16fml option.
16841 2018-01-10 Michael Collison <michael.collison@arm.com>
16843 * config/aarch64/aarch64-builtins.c:
16844 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16845 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16846 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16847 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16848 (AARCH64_ISA_SHA3): New ISA flag.
16849 (TARGET_SHA3): New feature flag for sha3.
16850 * config/aarch64/iterators.md (sha512_op): New int attribute.
16851 (CRYPTO_SHA512): New int iterator.
16852 (UNSPEC_SHA512H): New unspec.
16853 (UNSPEC_SHA512H2): Ditto.
16854 (UNSPEC_SHA512SU0): Ditto.
16855 (UNSPEC_SHA512SU1): Ditto.
16856 * config/aarch64/aarch64-simd-builtins.def
16857 (aarch64_crypto_sha512hqv2di): New builtin.
16858 (aarch64_crypto_sha512h2qv2di): Ditto.
16859 (aarch64_crypto_sha512su0qv2di): Ditto.
16860 (aarch64_crypto_sha512su1qv2di): Ditto.
16861 (aarch64_eor3qv8hi): Ditto.
16862 (aarch64_rax1qv2di): Ditto.
16863 (aarch64_xarqv2di): Ditto.
16864 (aarch64_bcaxqv8hi): Ditto.
16865 * config/aarch64/aarch64-simd.md:
16866 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16867 (aarch64_crypto_sha512su0qv2di): Ditto.
16868 (aarch64_crypto_sha512su1qv2di): Ditto.
16869 (aarch64_eor3qv8hi): Ditto.
16870 (aarch64_rax1qv2di): Ditto.
16871 (aarch64_xarqv2di): Ditto.
16872 (aarch64_bcaxqv8hi): Ditto.
16873 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16874 (vsha512h2q_u64): Ditto.
16875 (vsha512su0q_u64): Ditto.
16876 (vsha512su1q_u64): Ditto.
16877 (veor3q_u16): Ditto.
16878 (vrax1q_u64): Ditto.
16879 (vxarq_u64): Ditto.
16880 (vbcaxq_u16): Ditto.
16881 * config/arm/types.md (crypto_sha512): New type attribute.
16882 (crypto_sha3): Ditto.
16883 (doc/invoke.texi): Document new sha3 option.
16885 2018-01-10 Michael Collison <michael.collison@arm.com>
16887 * config/aarch64/aarch64-builtins.c:
16888 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16889 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16890 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16891 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16892 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16893 (AARCH64_ISA_SM4): New ISA flag.
16894 (TARGET_SM4): New feature flag for sm4.
16895 * config/aarch64/aarch64-simd-builtins.def
16896 (aarch64_sm3ss1qv4si): Ditto.
16897 (aarch64_sm3tt1aq4si): Ditto.
16898 (aarch64_sm3tt1bq4si): Ditto.
16899 (aarch64_sm3tt2aq4si): Ditto.
16900 (aarch64_sm3tt2bq4si): Ditto.
16901 (aarch64_sm3partw1qv4si): Ditto.
16902 (aarch64_sm3partw2qv4si): Ditto.
16903 (aarch64_sm4eqv4si): Ditto.
16904 (aarch64_sm4ekeyqv4si): Ditto.
16905 * config/aarch64/aarch64-simd.md:
16906 (aarch64_sm3ss1qv4si): Ditto.
16907 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16908 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16909 (aarch64_sm4eqv4si): Ditto.
16910 (aarch64_sm4ekeyqv4si): Ditto.
16911 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16912 (sm3part_op): Ditto.
16913 (CRYPTO_SM3TT): Ditto.
16914 (CRYPTO_SM3PART): Ditto.
16915 (UNSPEC_SM3SS1): New unspec.
16916 (UNSPEC_SM3TT1A): Ditto.
16917 (UNSPEC_SM3TT1B): Ditto.
16918 (UNSPEC_SM3TT2A): Ditto.
16919 (UNSPEC_SM3TT2B): Ditto.
16920 (UNSPEC_SM3PARTW1): Ditto.
16921 (UNSPEC_SM3PARTW2): Ditto.
16922 (UNSPEC_SM4E): Ditto.
16923 (UNSPEC_SM4EKEY): Ditto.
16924 * config/aarch64/constraints.md (Ui2): New constraint.
16925 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16926 * config/arm/types.md (crypto_sm3): New type attribute.
16927 (crypto_sm4): Ditto.
16928 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16929 (vsm3tt1aq_u32): Ditto.
16930 (vsm3tt1bq_u32): Ditto.
16931 (vsm3tt2aq_u32): Ditto.
16932 (vsm3tt2bq_u32): Ditto.
16933 (vsm3partw1q_u32): Ditto.
16934 (vsm3partw2q_u32): Ditto.
16935 (vsm4eq_u32): Ditto.
16936 (vsm4ekeyq_u32): Ditto.
16937 (doc/invoke.texi): Document new sm4 option.
16939 2018-01-10 Michael Collison <michael.collison@arm.com>
16941 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16942 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16943 (AARCH64_FL_FOR_ARCH8_4): New.
16944 (AARCH64_FL_V8_4): New flag.
16945 (doc/invoke.texi): Document new armv8.4-a option.
16947 2018-01-10 Michael Collison <michael.collison@arm.com>
16949 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16950 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16951 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16952 * config/aarch64/aarch64-option-extension.def: Add
16953 AARCH64_OPT_EXTENSION of 'sha2'.
16954 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16955 (crypto): Disable sha2 and aes if crypto disabled.
16956 (crypto): Enable aes and sha2 if enabled.
16957 (simd): Disable sha2 and aes if simd disabled.
16958 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16960 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16961 (TARGET_SHA2): New feature flag for sha2.
16962 (TARGET_AES): New feature flag for aes.
16963 * config/aarch64/aarch64-simd.md:
16964 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16965 conditional on TARGET_AES.
16966 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16967 (aarch64_crypto_sha1hsi): Make pattern conditional
16969 (aarch64_crypto_sha1hv4si): Ditto.
16970 (aarch64_be_crypto_sha1hv4si): Ditto.
16971 (aarch64_crypto_sha1su1v4si): Ditto.
16972 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16973 (aarch64_crypto_sha1su0v4si): Ditto.
16974 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16975 (aarch64_crypto_sha256su0v4si): Ditto.
16976 (aarch64_crypto_sha256su1v4si): Ditto.
16977 (doc/invoke.texi): Document new aes and sha2 options.
16979 2018-01-10 Martin Sebor <msebor@redhat.com>
16981 PR tree-optimization/83781
16982 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16985 2018-01-11 Martin Sebor <msebor@gmail.com>
16986 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16988 PR tree-optimization/83501
16989 PR tree-optimization/81703
16991 * tree-ssa-strlen.c (get_string_cst): Rename...
16992 (get_string_len): ...to this. Handle global constants.
16993 (handle_char_store): Adjust.
16995 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16996 Jim Wilson <jimw@sifive.com>
16998 * config/riscv/riscv-protos.h (riscv_output_return): New.
16999 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17000 (riscv_attribute_table, riscv_output_return),
17001 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17002 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17003 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17005 (riscv_expand_prologue): Add early return for naked function.
17006 (riscv_expand_epilogue): Likewise.
17007 (riscv_function_ok_for_sibcall): Return false for naked function.
17008 (riscv_set_current_function): New.
17009 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17010 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17011 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17012 * doc/extend.texi (RISC-V Function Attributes): New.
17014 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17016 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17017 check for 128-bit long double before checking TCmode.
17018 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17019 128-bit long doubles before checking TFmode or TCmode.
17020 (FLOAT128_IBM_P): Likewise.
17022 2018-01-10 Martin Sebor <msebor@redhat.com>
17024 PR tree-optimization/83671
17025 * builtins.c (c_strlen): Unconditionally return zero for the empty
17027 Use -Warray-bounds for warnings.
17028 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17029 for non-constant array indices with COMPONENT_REF, arrays of
17030 arrays, and pointers to arrays.
17031 (gimple_fold_builtin_strlen): Determine and set length range for
17032 non-constant character arrays.
17034 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17036 PR middle-end/81897
17037 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17040 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17042 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17044 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17047 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17048 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17049 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17050 indexed_or_indirect_operand predicate.
17051 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17052 (*vsx_le_perm_load_v8hi): Likewise.
17053 (*vsx_le_perm_load_v16qi): Likewise.
17054 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17055 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17056 (*vsx_le_perm_store_v8hi): Likewise.
17057 (*vsx_le_perm_store_v16qi): Likewise.
17058 (eight unnamed splitters): Likewise.
17060 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17062 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17063 * config/rs6000/emmintrin.h: Likewise.
17064 * config/rs6000/mmintrin.h: Likewise.
17065 * config/rs6000/xmmintrin.h: Likewise.
17067 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17070 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17072 * tree.c (tree_nop_conversion): Return true for location wrapper
17074 (maybe_wrap_with_location): New function.
17075 (selftest::check_strip_nops): New function.
17076 (selftest::test_location_wrappers): New function.
17077 (selftest::tree_c_tests): Call it.
17078 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17079 (maybe_wrap_with_location): New decl.
17080 (EXPR_LOCATION_WRAPPER_P): New macro.
17081 (location_wrapper_p): New inline function.
17082 (tree_strip_any_location_wrapper): New inline function.
17084 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17087 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17088 stack_realign_offset for the largest alignment of stack slot
17090 (ix86_find_max_used_stack_alignment): New function.
17091 (ix86_finalize_stack_frame_flags): Use it. Set
17092 max_used_stack_alignment if we don't realign stack.
17093 * config/i386/i386.h (machine_function): Add
17094 max_used_stack_alignment.
17096 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17098 * config/arm/arm.opt (-mbranch-cost): New option.
17099 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17102 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17105 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17106 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17108 2018-01-10 Richard Biener <rguenther@suse.de>
17111 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17112 early out so it also covers the case where we have a non-NULL
17115 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17117 PR tree-optimization/83753
17118 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17119 for non-strided grouped accesses if the number of elements is 1.
17121 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17124 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17125 * i386.h (TARGET_USE_GATHER): Define.
17126 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17128 2018-01-10 Martin Liska <mliska@suse.cz>
17131 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17132 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17134 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17135 CLEANUP_NO_PARTITIONING is not set.
17137 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17139 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17140 for vectors, as a partial revert of r254296.
17141 * rtl.h (const_vec_p): Delete.
17142 (const_vec_duplicate_p): Don't test for vector CONSTs.
17143 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17144 * expmed.c (make_tree): Likewise.
17147 * common.md (E, F): Use CONSTANT_P instead of checking for
17149 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17150 checking for CONST_VECTOR.
17152 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17154 PR middle-end/83575
17155 * predict.c (force_edge_cold): Handle in more sane way edges
17156 with no prediction.
17158 2018-01-09 Carl Love <cel@us.ibm.com>
17160 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17162 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17163 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17164 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17165 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17166 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17167 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17168 * config/rs6000/rs6000-protos.h: Add extern defition for
17169 rs6000_generate_float2_double_code.
17170 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17172 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17173 (float2_v2df): Add define_expand.
17175 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17178 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17179 op_mode in the force_to_mode call.
17181 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17183 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17184 instead of checking each element individually.
17185 (aarch64_evpc_uzp): Likewise.
17186 (aarch64_evpc_zip): Likewise.
17187 (aarch64_evpc_ext): Likewise.
17188 (aarch64_evpc_rev): Likewise.
17189 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17190 instead of checking each element individually. Return true without
17192 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17193 whether all selected elements come from the same input, instead of
17194 checking each element individually. Remove calls to gen_rtx_REG,
17195 start_sequence and end_sequence and instead assert that no rtl is
17198 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17200 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17201 order of HIGH and CONST checks.
17203 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17205 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17206 if the destination isn't an SSA_NAME.
17208 2018-01-09 Richard Biener <rguenther@suse.de>
17210 PR tree-optimization/83668
17211 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17213 (canonicalize_loop_form): ... here, renamed from ...
17214 (canonicalize_loop_closed_ssa_form): ... this and amended to
17215 swap successor edges for loop exit blocks to make us use
17216 the RPO order we need for initial schedule generation.
17218 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17220 PR tree-optimization/64811
17221 * match.pd: When optimizing comparisons with Inf, avoid
17222 introducing or losing exceptions from comparisons with NaN.
17224 2018-01-09 Martin Liska <mliska@suse.cz>
17227 * asan.c (shadow_mem_size): Add gcc_assert.
17229 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17231 Don't save registers in main().
17234 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17235 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17236 * config/avr/avr.c (avr_set_current_function): Don't error if
17237 naked, OS_task or OS_main are specified at the same time.
17238 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17240 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17242 * common/config/avr/avr-common.c (avr_option_optimization_table):
17243 Switch on -mmain-is-OS_task for optimizing compilations.
17245 2018-01-09 Richard Biener <rguenther@suse.de>
17247 PR tree-optimization/83572
17248 * graphite.c: Include cfganal.h.
17249 (graphite_transform_loops): Connect infinite loops to exit
17250 and remove fake edges at the end.
17252 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17254 * ipa-inline.c (edge_badness): Revert accidental checkin.
17256 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17259 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17260 symbols; not inline clones.
17262 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17265 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17266 hard registers. Formatting fixes.
17268 PR preprocessor/83722
17269 * gcc.c (try_generate_repro): Pass
17270 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17271 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17274 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17275 Kito Cheng <kito.cheng@gmail.com>
17277 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17278 (riscv_leaf_function_p): Delete.
17279 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17281 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17283 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17285 (do_ifelse): New function.
17286 (do_isel): New function.
17287 (do_sub3): New function.
17288 (do_add3): New function.
17289 (do_load_mask_compare): New function.
17290 (do_overlap_load_compare): New function.
17291 (expand_compare_loop): New function.
17292 (expand_block_compare): Call expand_compare_loop() when appropriate.
17293 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17294 option description.
17295 (-mblock-compare-inline-loop-limit): New option.
17297 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17300 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17301 Reverse order of second and third operands in first alternative.
17302 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17303 of first and second elements in UNSPEC_VPERMR vector.
17304 (altivec_expand_vec_perm_le): Likewise.
17306 2018-01-08 Jeff Law <law@redhat.com>
17308 PR rtl-optimizatin/81308
17309 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17310 (process_switch): If group_case_labels makes a change, then set
17312 (pass_convert_switch::execute): If a switch is converted, then
17313 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17315 PR rtl-optimization/81308
17316 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17319 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17321 PR target/83663 - Revert r255946
17322 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17323 generation for cases where splatting a value is not useful.
17324 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17325 across a vec_duplicate and a paradoxical subreg forming a vector
17326 mode to a vec_concat.
17328 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17330 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17331 -march=armv8.3-a variants.
17332 * config/arm/t-multilib: Likewise.
17333 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17335 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17337 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17339 (cceq_ior_compare_complement): Give it a name so I can use it, and
17340 change boolean_or_operator predicate to boolean_operator so it can
17341 be used to generate a crand.
17342 (eqne): New code iterator.
17343 (bd/bd_neg): New code_attrs.
17344 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17345 a single define_insn.
17346 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17347 decrement (bdnzt/bdnzf/bdzt/bdzf).
17348 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17349 with the new names of the branch decrement patterns, and added the
17350 names of the branch decrement conditional patterns.
17352 2018-01-08 Richard Biener <rguenther@suse.de>
17354 PR tree-optimization/83563
17355 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17358 2018-01-08 Richard Biener <rguenther@suse.de>
17360 PR middle-end/83713
17361 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17363 2018-01-08 Richard Biener <rguenther@suse.de>
17365 PR tree-optimization/83685
17366 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17367 references to abnormals.
17369 2018-01-08 Richard Biener <rguenther@suse.de>
17372 * dwarf2out.c (output_indirect_strings): Handle empty
17373 skeleton_debug_str_hash.
17374 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17376 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17378 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17379 (emit_store_direct): Likewise.
17380 (arc_trampoline_adjust_address): Likewise.
17381 (arc_asm_trampoline_template): New function.
17382 (arc_initialize_trampoline): Use asm_trampoline_template.
17383 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17384 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17385 * config/arc/arc.md (flush_icache): Delete pattern.
17387 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17389 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17390 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17393 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17396 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17397 by not USED_FOR_TARGET.
17398 (make_pass_resolve_sw_modes): Likewise.
17400 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17402 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17405 2018-01-08 Richard Biener <rguenther@suse.de>
17407 PR middle-end/83580
17408 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17410 2018-01-08 Richard Biener <rguenther@suse.de>
17412 PR middle-end/83517
17413 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17415 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17417 PR middle-end/81897
17418 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17419 basic blocks with a small number of successors.
17420 (convert_control_dep_chain_into_preds): Improve handling of
17422 (dump_predicates): Split apart into...
17423 (dump_pred_chain): ...here...
17424 (dump_pred_info): ...and here.
17425 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17426 (can_chain_union_be_invalidated_p): Improve check for invalidation
17428 (uninit_uses_cannot_happen): Avoid unnecessary if
17429 convert_control_dep_chain_into_preds yielded nothing.
17431 2018-01-06 Martin Sebor <msebor@redhat.com>
17433 PR tree-optimization/83640
17434 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17435 subtracting negative offset from size.
17436 (builtin_access::overlap): Adjust offset bounds of the access to fall
17437 within the size of the object if possible.
17439 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17441 PR rtl-optimization/83699
17442 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17443 extract_bit_field_as_subreg to cases in which the extracted
17444 value is also a vector.
17446 * lra-constraints.c (process_alt_operands): Test for the equivalence
17447 substitutions when detecting a possible reload cycle.
17449 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17452 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17453 by default if flag_selective_schedling{,2}. Formatting fixes.
17455 PR rtl-optimization/83682
17456 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17457 if it has non-VECTOR_MODE element mode.
17458 (vec_duplicate_p): Likewise.
17460 PR middle-end/83694
17461 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17462 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17464 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17467 * config/i386/i386-builtin.def
17468 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17469 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17470 Require also OPTION_MASK_ISA_AVX512F in addition to
17471 OPTION_MASK_ISA_GFNI.
17472 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17473 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17474 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17475 to OPTION_MASK_ISA_GFNI.
17476 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17477 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17478 OPTION_MASK_ISA_AVX512BW.
17479 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17480 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17481 addition to OPTION_MASK_ISA_GFNI.
17482 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17483 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17484 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17485 to OPTION_MASK_ISA_GFNI.
17486 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17487 a requirement for all ISAs rather than any of them with a few
17489 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17491 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17492 bitmasks to be enabled with 3 exceptions, instead of requiring any
17493 enabled ISA with lots of exceptions.
17494 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17495 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17496 Change avx512bw in isa attribute to avx512f.
17497 * config/i386/sgxintrin.h: Add license boilerplate.
17498 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17499 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17500 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17501 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17503 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17504 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17505 temporarily sse2 rather than sse if not enabled already.
17508 * config/i386/sse.md (VI248_VLBW): Rename to ...
17509 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17510 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17511 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17512 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17513 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17514 mode iterator instead of VI248_VLBW.
17516 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17518 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17519 (record_modified): Skip clobbers; add debug output.
17520 (param_change_prob): Use sreal frequencies.
17522 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17524 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17525 punt for user-aligned variables.
17527 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17529 * tree-chrec.c (chrec_contains_symbols): Return true for
17532 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17535 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17536 of (x|y) == x for BICS pattern.
17538 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17540 PR tree-optimization/83605
17541 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17542 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17545 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17547 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17548 * config/epiphany/rtems.h: New file.
17550 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17551 Uros Bizjak <ubizjak@gmail.com>
17554 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17555 QIreg_operand instead of register_operand predicate.
17556 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17557 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17558 comments instead of -fmitigate[-_]rop.
17560 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17563 * cgraphunit.c (symbol_table::compile): Switch to text_section
17564 before calling assembly_start debug hook.
17565 * run-rtl-passes.c (run_rtl_passes): Likewise.
17568 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17570 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17571 range_int_cst_p rather than !symbolic_range_p before calling
17572 extract_range_from_multiplicative_op_1.
17574 2018-01-04 Jeff Law <law@redhat.com>
17576 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17577 redundant test in assertion.
17579 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17581 * doc/rtl.texi: Document machine_mode wrapper classes.
17583 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17585 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17586 using tree_to_uhwi.
17588 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17590 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17591 the VEC_PERM_EXPR fold to fail.
17593 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17596 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17597 to switched_sections.
17599 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17602 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17603 test for d.testing.
17605 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17608 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17609 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17611 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17614 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17615 is BLKmode and bitpos not zero or mode change is needed.
17617 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17620 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17623 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17626 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17627 instead of MULT rtx. Update all corresponding splitters.
17628 (*saddl_se): Ditto.
17629 (*ssub<modesuffix>): Ditto.
17630 (*ssubl_se): Ditto.
17631 (*cmp_sadd_di): Update split patterns.
17632 (*cmp_sadd_si): Ditto.
17633 (*cmp_sadd_sidi): Ditto.
17634 (*cmp_ssub_di): Ditto.
17635 (*cmp_ssub_si): Ditto.
17636 (*cmp_ssub_sidi): Ditto.
17637 * config/alpha/predicates.md (const23_operand): New predicate.
17638 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17639 Look for ASHIFT, not MULT inner operand.
17640 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17642 2018-01-04 Martin Liska <mliska@suse.cz>
17644 PR gcov-profile/83669
17645 * gcov.c (output_intermediate_file): Add version to intermediate
17647 * doc/gcov.texi: Document new field 'version' in intermediate
17648 file format. Fix location of '-k' option of gcov command.
17650 2018-01-04 Martin Liska <mliska@suse.cz>
17653 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17655 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17657 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17659 2018-01-03 Martin Sebor <msebor@redhat.com>
17661 PR tree-optimization/83655
17662 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17663 checking calls with invalid arguments.
17665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17667 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17668 (vectorizable_mask_load_store): Delete.
17669 (vectorizable_call): Return false for masked loads and stores.
17670 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17671 instead of gimple_assign_rhs1.
17672 (vectorizable_load): Handle IFN_MASK_LOAD.
17673 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17677 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17679 (vectorizable_mask_load_store): ...here.
17680 (vectorizable_load): ...and here.
17682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17684 * tree-vect-stmts.c (vect_build_all_ones_mask)
17685 (vect_build_zero_merge_argument): New functions, split out from...
17686 (vectorizable_load): ...here.
17688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17690 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17692 (vectorizable_mask_load_store): ...here.
17693 (vectorizable_store): ...and here.
17695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17697 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17699 (vectorizable_mask_load_store): ...here.
17701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17703 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17704 (vect_model_store_cost): Take a vec_load_store_type instead of a
17706 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17707 (vect_model_store_cost): Take a vec_load_store_type instead of a
17709 (vectorizable_mask_load_store): Update accordingly.
17710 (vectorizable_store): Likewise.
17711 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17715 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17716 IFN_MASK_LOAD calls here rather than...
17717 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17720 Alan Hayward <alan.hayward@arm.com>
17721 David Sherwood <david.sherwood@arm.com>
17723 * expmed.c (extract_bit_field_1): For vector extracts,
17724 fall back to extract_bit_field_as_subreg if vec_extract
17727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17728 Alan Hayward <alan.hayward@arm.com>
17729 David Sherwood <david.sherwood@arm.com>
17731 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17732 they are variable or constant sized.
17733 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17734 slots for constant-sized data.
17736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17737 Alan Hayward <alan.hayward@arm.com>
17738 David Sherwood <david.sherwood@arm.com>
17740 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17741 handling COND_EXPRs with boolean comparisons, try to find a better
17742 basis for the mask type than the boolean itself.
17744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17746 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17747 is calculated and how it can be overridden.
17748 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17749 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17751 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17755 Alan Hayward <alan.hayward@arm.com>
17756 David Sherwood <david.sherwood@arm.com>
17758 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17759 Remove the mode argument.
17760 (aarch64_simd_valid_immediate): Remove the mode and inverse
17762 * config/aarch64/iterators.md (bitsize): New iterator.
17763 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17764 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17765 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17766 aarch64_simd_valid_immediate.
17767 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17768 (aarch64_reg_or_bic_imm): Likewise.
17769 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17770 with an insn_type enum and msl with a modifier_type enum.
17771 Replace element_width with a scalar_mode. Change the shift
17772 to unsigned int. Add constructors for scalar_float_mode and
17773 scalar_int_mode elements.
17774 (aarch64_vect_float_const_representable_p): Delete.
17775 (aarch64_can_const_movi_rtx_p)
17776 (aarch64_simd_scalar_immediate_valid_for_move)
17777 (aarch64_simd_make_constant): Update call to
17778 aarch64_simd_valid_immediate.
17779 (aarch64_advsimd_valid_immediate_hs): New function.
17780 (aarch64_advsimd_valid_immediate): Likewise.
17781 (aarch64_simd_valid_immediate): Remove mode and inverse
17782 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17783 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17784 and aarch64_float_const_representable_p on the result.
17785 (aarch64_output_simd_mov_immediate): Remove mode argument.
17786 Update call to aarch64_simd_valid_immediate and use of
17787 simd_immediate_info.
17788 (aarch64_output_scalar_simd_mov_immediate): Update call
17791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17792 Alan Hayward <alan.hayward@arm.com>
17793 David Sherwood <david.sherwood@arm.com>
17795 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17796 (mode_nunits): Likewise CONST_MODE_NUNITS.
17797 * machmode.def (ADJUST_NUNITS): Document.
17798 * genmodes.c (mode_data::need_nunits_adj): New field.
17799 (blank_mode): Update accordingly.
17800 (adj_nunits): New variable.
17801 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17803 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17804 listed in adj_nunits.
17805 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17806 listed in adj_nunits. Don't emit case statements for such modes.
17807 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17808 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17809 nothing if adj_nunits is nonnull.
17810 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17811 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17812 (emit_mode_fbit): Update use of print_maybe_const_decl.
17813 (emit_move_size): Likewise. Treat the array as non-const
17815 (emit_mode_adjustments): Handle adj_nunits.
17817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17819 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17820 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17821 (VECTOR_MODES): Use it.
17822 (make_vector_modes): Take the prefix as an argument.
17824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17825 Alan Hayward <alan.hayward@arm.com>
17826 David Sherwood <david.sherwood@arm.com>
17828 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17829 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17830 for MODE_VECTOR_BOOL.
17831 * machmode.def (VECTOR_BOOL_MODE): Document.
17832 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17833 (make_vector_bool_mode): New function.
17834 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17836 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17837 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17839 * stor-layout.c (int_mode_for_mode): Likewise.
17840 * tree.c (build_vector_type_for_mode): Likewise.
17841 * varasm.c (output_constant_pool_2): Likewise.
17842 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17843 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17844 for MODE_VECTOR_BOOL.
17845 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17846 of mode class checks.
17847 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17848 instead of a list of mode class checks.
17849 (expand_vector_scalar_condition): Likewise.
17850 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17853 Alan Hayward <alan.hayward@arm.com>
17854 David Sherwood <david.sherwood@arm.com>
17856 * machmode.h (mode_size): Change from unsigned short to
17858 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17859 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17860 or if measurement_type is not polynomial.
17861 (fixed_size_mode::includes_p): Check for constant-sized modes.
17862 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17863 return a poly_uint16 rather than an unsigned short.
17864 (emit_mode_size): Change the type of mode_size from unsigned short
17865 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17866 (emit_mode_adjustments): Cope with polynomial vector sizes.
17867 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17869 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17871 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17872 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17873 * caller-save.c (setup_save_areas): Likewise.
17874 (replace_reg_with_saved_mem): Likewise.
17875 * calls.c (emit_library_call_value_1): Likewise.
17876 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17877 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17878 (gen_lowpart_for_combine): Likewise.
17879 * convert.c (convert_to_integer_1): Likewise.
17880 * cse.c (equiv_constant, cse_insn): Likewise.
17881 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17882 (cselib_subst_to_values): Likewise.
17883 * dce.c (word_dce_process_block): Likewise.
17884 * df-problems.c (df_word_lr_mark_ref): Likewise.
17885 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17886 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17887 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17888 (rtl_for_decl_location): Likewise.
17889 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17890 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17891 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17892 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17893 (expand_expr_real_1): Likewise.
17894 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17895 (pad_below): Likewise.
17896 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17897 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17898 * ira.c (get_subreg_tracking_sizes): Likewise.
17899 * ira-build.c (ira_create_allocno_objects): Likewise.
17900 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17901 (ira_sort_regnos_for_alter_reg): Likewise.
17902 * ira-costs.c (record_operand_costs): Likewise.
17903 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17904 (resolve_simple_move): Likewise.
17905 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17906 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17907 (lra_constraints): Likewise.
17908 (CONST_POOL_OK_P): Reject variable-sized modes.
17909 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17910 (add_pseudo_to_slot, lra_spill): Likewise.
17911 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17912 * optabs-query.c (get_best_extraction_insn): Likewise.
17913 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17914 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17915 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17916 * recog.c (offsettable_address_addr_space_p): Likewise.
17917 * regcprop.c (maybe_mode_change): Likewise.
17918 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17919 * regrename.c (build_def_use): Likewise.
17920 * regstat.c (dump_reg_info): Likewise.
17921 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17922 (find_reloads, find_reloads_subreg_address): Likewise.
17923 * reload1.c (eliminate_regs_1): Likewise.
17924 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17925 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17926 (simplify_binary_operation_1, simplify_subreg): Likewise.
17927 * targhooks.c (default_function_arg_padding): Likewise.
17928 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17929 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17930 (verify_gimple_assign_ternary): Likewise.
17931 * tree-inline.c (estimate_move_cost): Likewise.
17932 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17933 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17934 (get_address_cost_ainc): Likewise.
17935 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17936 (vect_supportable_dr_alignment): Likewise.
17937 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17938 (vectorizable_reduction): Likewise.
17939 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17940 (vectorizable_operation, vectorizable_load): Likewise.
17941 * tree.c (build_same_sized_truth_vector_type): Likewise.
17942 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17943 * var-tracking.c (emit_note_insn_var_location): Likewise.
17944 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17945 (ADDR_VEC_ALIGN): Likewise.
17947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17948 Alan Hayward <alan.hayward@arm.com>
17949 David Sherwood <david.sherwood@arm.com>
17951 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17953 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17954 or if measurement_type is polynomial.
17955 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17956 * combine.c (make_extraction): Likewise.
17957 * dse.c (find_shift_sequence): Likewise.
17958 * dwarf2out.c (mem_loc_descriptor): Likewise.
17959 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17960 (extract_bit_field, extract_low_bits): Likewise.
17961 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17962 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17963 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17964 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17965 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17966 * reload.c (find_reloads): Likewise.
17967 * reload1.c (alter_reg): Likewise.
17968 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17969 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17970 * tree-if-conv.c (predicate_mem_writes): Likewise.
17971 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17972 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17973 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17974 * valtrack.c (dead_debug_insert_temp): Likewise.
17975 * varasm.c (mergeable_constant_section): Likewise.
17976 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17979 Alan Hayward <alan.hayward@arm.com>
17980 David Sherwood <david.sherwood@arm.com>
17982 * expr.c (expand_assignment): Cope with polynomial mode sizes
17983 when assigning to a CONCAT.
17985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17986 Alan Hayward <alan.hayward@arm.com>
17987 David Sherwood <david.sherwood@arm.com>
17989 * machmode.h (mode_precision): Change from unsigned short to
17991 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17993 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17994 or if measurement_type is not polynomial.
17995 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17996 in which the mode is already known to be a scalar_int_mode.
17997 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17998 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18000 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18001 for GET_MODE_PRECISION.
18002 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18003 for GET_MODE_PRECISION.
18004 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18006 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18007 (expand_field_assignment, make_extraction): Likewise.
18008 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18009 (get_last_value): Likewise.
18010 * convert.c (convert_to_integer_1): Likewise.
18011 * cse.c (cse_insn): Likewise.
18012 * expr.c (expand_expr_real_1): Likewise.
18013 * lra-constraints.c (simplify_operand_subreg): Likewise.
18014 * optabs-query.c (can_atomic_load_p): Likewise.
18015 * optabs.c (expand_atomic_load): Likewise.
18016 (expand_atomic_store): Likewise.
18017 * ree.c (combine_reaching_defs): Likewise.
18018 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18019 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18020 * tree.h (type_has_mode_precision_p): Likewise.
18021 * ubsan.c (instrument_si_overflow): Likewise.
18023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18024 Alan Hayward <alan.hayward@arm.com>
18025 David Sherwood <david.sherwood@arm.com>
18027 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18028 polynomial numbers of units.
18029 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18030 (valid_vector_subparts_p): New function.
18031 (build_vector_type): Remove temporary shim and take the number
18032 of units as a poly_uint64 rather than an int.
18033 (build_opaque_vector_type): Take the number of units as a
18034 poly_uint64 rather than an int.
18035 * tree.c (build_vector_from_ctor): Handle polynomial
18036 TYPE_VECTOR_SUBPARTS.
18037 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18038 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18039 (build_vector_from_val): If the number of units is variable,
18040 use build_vec_duplicate_cst for constant operands and
18041 VEC_DUPLICATE_EXPR otherwise.
18042 (make_vector_type): Remove temporary is_constant ().
18043 (build_vector_type, build_opaque_vector_type): Take the number of
18044 units as a poly_uint64 rather than an int.
18045 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18047 * cfgexpand.c (expand_debug_expr): Likewise.
18048 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18049 (store_constructor, expand_expr_real_1): Likewise.
18050 (const_scalar_mask_from_tree): Likewise.
18051 * fold-const-call.c (fold_const_reduction): Likewise.
18052 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18053 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18054 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18055 (fold_relational_const): Likewise.
18056 (native_interpret_vector): Likewise. Change the size from an
18057 int to an unsigned int.
18058 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18059 TYPE_VECTOR_SUBPARTS.
18060 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18061 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18062 duplicating a non-constant operand into a variable-length vector.
18063 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18064 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18065 * ipa-icf.c (sem_variable::equals): Likewise.
18066 * match.pd: Likewise.
18067 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18068 * print-tree.c (print_node): Likewise.
18069 * stor-layout.c (layout_type): Likewise.
18070 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18071 * tree-cfg.c (verify_gimple_comparison): Likewise.
18072 (verify_gimple_assign_binary): Likewise.
18073 (verify_gimple_assign_ternary): Likewise.
18074 (verify_gimple_assign_single): Likewise.
18075 * tree-pretty-print.c (dump_generic_node): Likewise.
18076 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18077 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18078 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18079 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18080 (vect_shift_permute_load_chain): Likewise.
18081 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18082 (expand_vector_condition, optimize_vector_constructor): Likewise.
18083 (lower_vec_perm, get_compute_type): Likewise.
18084 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18085 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18086 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18087 (vect_recog_mask_conversion_pattern): Likewise.
18088 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18089 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18090 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18091 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18092 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18093 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18094 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18095 (supportable_widening_operation): Likewise.
18096 (supportable_narrowing_operation): Likewise.
18097 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18099 * varasm.c (output_constant): Likewise.
18101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18102 Alan Hayward <alan.hayward@arm.com>
18103 David Sherwood <david.sherwood@arm.com>
18105 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18106 so that both the length == 3 and length != 3 cases set up their
18107 own permute vectors. Add comments explaining why we know the
18108 number of elements is constant.
18109 (vect_permute_load_chain): 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_nunits): Change from unsigned char to
18117 (ONLY_FIXED_SIZE_MODES): New macro.
18118 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18119 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18120 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18122 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18123 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18124 or if measurement_type is not polynomial.
18125 * genmodes.c (ZERO_COEFFS): New macro.
18126 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18128 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18129 Use ZERO_COEFFS when emitting initializers.
18130 * data-streamer.h (bp_pack_poly_value): New function.
18131 (bp_unpack_poly_value): Likewise.
18132 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18133 for GET_MODE_NUNITS.
18134 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18135 for GET_MODE_NUNITS.
18136 * tree.c (make_vector_type): Remove temporary shim and make
18137 the real function take the number of units as a poly_uint64
18138 rather than an int.
18139 (build_vector_type_for_mode): Handle polynomial nunits.
18140 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18141 * emit-rtl.c (const_vec_series_p_1): Likewise.
18142 (gen_rtx_CONST_VECTOR): Likewise.
18143 * fold-const.c (test_vec_duplicate_folding): Likewise.
18144 * genrecog.c (validate_pattern): Likewise.
18145 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18146 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18147 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18148 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18149 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18150 * rtlanal.c (subreg_get_info): Likewise.
18151 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18152 (vect_grouped_load_supported): Likewise.
18153 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18154 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18155 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18156 (simplify_const_unary_operation, simplify_binary_operation_1)
18157 (simplify_const_binary_operation, simplify_ternary_operation)
18158 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18159 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18160 instead of CONST_VECTOR_NUNITS.
18161 * varasm.c (output_constant_pool_2): Likewise.
18162 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18163 explicit-encoded elements in the XVEC for variable-length vectors.
18165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18167 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18170 Alan Hayward <alan.hayward@arm.com>
18171 David Sherwood <david.sherwood@arm.com>
18173 * coretypes.h (fixed_size_mode): Declare.
18174 (fixed_size_mode_pod): New typedef.
18175 * builtins.h (target_builtins::x_apply_args_mode)
18176 (target_builtins::x_apply_result_mode): Change type to
18177 fixed_size_mode_pod.
18178 * builtins.c (apply_args_size, apply_result_size, result_vector)
18179 (expand_builtin_apply_args_1, expand_builtin_apply)
18180 (expand_builtin_return): Update accordingly.
18182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18184 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18185 * cselib.c (cselib_hash_rtx): Likewise.
18186 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18187 CONST_VECTOR encoding.
18189 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18190 Jeff Law <law@redhat.com>
18193 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18194 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18195 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18196 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18199 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18200 explicitly probe *sp in a noreturn function if there were any callee
18201 register saves or frame pointer is needed.
18203 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18206 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18207 BLKmode for ternary, binary or unary expressions.
18210 * var-tracking.c (delete_vta_debug_insn): New inline function.
18211 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18212 insns from get_insns () to NULL instead of each bb separately.
18213 Use delete_vta_debug_insn. No longer static.
18214 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18215 delete_vta_debug_insns callers.
18216 * rtl.h (delete_vta_debug_insns): Declare.
18217 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18218 instead of variable_tracking_main.
18220 2018-01-03 Martin Sebor <msebor@redhat.com>
18222 PR tree-optimization/83603
18223 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18224 arguments past the endof the argument list in functions declared
18225 without a prototype.
18226 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18227 Avoid checking when arguments are null.
18229 2018-01-03 Martin Sebor <msebor@redhat.com>
18232 * doc/extend.texi (attribute const): Fix a typo.
18233 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18234 issuing -Wsuggest-attribute for void functions.
18236 2018-01-03 Martin Sebor <msebor@redhat.com>
18238 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18239 offset_int::from instead of wide_int::to_shwi.
18240 (maybe_diag_overlap): Remove assertion.
18241 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18242 * gimple-ssa-sprintf.c (format_directive): Same.
18243 (parse_directive): Same.
18244 (sprintf_dom_walker::compute_format_length): Same.
18245 (try_substitute_return_value): Same.
18247 2018-01-03 Jeff Law <law@redhat.com>
18249 PR middle-end/83654
18250 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18251 non-constant residual for zero at runtime and avoid probing in
18252 that case. Reorganize code for trailing problem to mirror handling
18255 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18257 PR tree-optimization/83501
18258 * tree-ssa-strlen.c (get_string_cst): New.
18259 (handle_char_store): Call get_string_cst.
18261 2018-01-03 Martin Liska <mliska@suse.cz>
18263 PR tree-optimization/83593
18264 * tree-ssa-strlen.c: Include tree-cfg.h.
18265 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18266 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18267 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18269 (strlen_dom_walker::before_dom_children): Call
18270 gimple_purge_dead_eh_edges. Dump tranformation with details
18272 (strlen_dom_walker::before_dom_children): Update call by adding
18273 new argument cleanup_eh.
18274 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18276 2018-01-03 Martin Liska <mliska@suse.cz>
18279 * cif-code.def (VARIADIC_THUNK): New enum value.
18280 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18283 2018-01-03 Jan Beulich <jbeulich@suse.com>
18285 * sse.md (mov<mode>_internal): Tighten condition for when to use
18286 vmovdqu<ssescalarsize> for TI and OI modes.
18288 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18290 Update copyright years.
18292 2018-01-03 Martin Liska <mliska@suse.cz>
18295 * ipa-visibility.c (function_and_variable_visibility): Skip
18296 functions with noipa attribure.
18298 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18300 * gcc.c (process_command): Update copyright notice dates.
18301 * gcov-dump.c (print_version): Ditto.
18302 * gcov.c (print_version): Ditto.
18303 * gcov-tool.c (print_version): Ditto.
18304 * gengtype.c (create_file): Ditto.
18305 * doc/cpp.texi: Bump @copying's copyright year.
18306 * doc/cppinternals.texi: Ditto.
18307 * doc/gcc.texi: Ditto.
18308 * doc/gccint.texi: Ditto.
18309 * doc/gcov.texi: Ditto.
18310 * doc/install.texi: Ditto.
18311 * doc/invoke.texi: Ditto.
18313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18315 * vector-builder.h (vector_builder::m_full_nelts): Change from
18316 unsigned int to poly_uint64.
18317 (vector_builder::full_nelts): Update prototype accordingly.
18318 (vector_builder::new_vector): Likewise.
18319 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18320 (vector_builder::operator ==): Likewise.
18321 (vector_builder::finalize): Likewise.
18322 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18323 Take the number of elements as a poly_uint64 rather than an
18325 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18326 from unsigned int to poly_uint64.
18327 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18328 (vec_perm_indices::new_vector): Likewise.
18329 (vec_perm_indices::length): Likewise.
18330 (vec_perm_indices::nelts_per_input): Likewise.
18331 (vec_perm_indices::input_nelts): Likewise.
18332 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18333 number of elements per input as a poly_uint64 rather than an
18334 unsigned int. Use the original encoding for variable-length
18335 vectors, rather than clamping each individual element.
18336 For the second and subsequent elements in each pattern,
18337 clamp the step and base before clamping their sum.
18338 (vec_perm_indices::series_p): Handle polynomial element counts.
18339 (vec_perm_indices::all_in_range_p): Likewise.
18340 (vec_perm_indices_to_tree): Likewise.
18341 (vec_perm_indices_to_rtx): Likewise.
18342 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18343 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18344 (tree_vector_builder::new_binary_operation): Handle polynomial
18345 element counts. Return false if we need to know the number
18346 of elements at compile time.
18347 * fold-const.c (fold_vec_perm): Punt if the number of elements
18348 isn't known at compile time.
18350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18352 * vec-perm-indices.h (vec_perm_builder): Change element type
18353 from HOST_WIDE_INT to poly_int64.
18354 (vec_perm_indices::element_type): Update accordingly.
18355 (vec_perm_indices::clamp): Handle polynomial element_types.
18356 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18357 (vec_perm_indices::all_in_range_p): Likewise.
18358 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18360 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18361 polynomial vec_perm_indices element types.
18362 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18363 * fold-const.c (fold_vec_perm): Likewise.
18364 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18365 * tree-vect-generic.c (lower_vec_perm): Likewise.
18366 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18367 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18368 element type to HOST_WIDE_INT.
18370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18371 Alan Hayward <alan.hayward@arm.com>
18372 David Sherwood <david.sherwood@arm.com>
18374 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18375 rather than an int. Use plus_constant.
18376 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18377 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18380 Alan Hayward <alan.hayward@arm.com>
18381 David Sherwood <david.sherwood@arm.com>
18383 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18384 a HOST_WIDE_INT to a poly_int64.
18386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18387 Alan Hayward <alan.hayward@arm.com>
18388 David Sherwood <david.sherwood@arm.com>
18390 * calls.c (load_register_parameters): Cope with polynomial
18391 mode sizes. Require a constant size for BLKmode parameters
18392 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18393 forces a parameter to be padded at the lsb end in order to
18394 fill a complete number of words, require the parameter size
18395 to be ordered wrt UNITS_PER_WORD.
18397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18398 Alan Hayward <alan.hayward@arm.com>
18399 David Sherwood <david.sherwood@arm.com>
18401 * reload1.c (spill_stack_slot_width): Change element type
18402 from unsigned int to poly_uint64_pod.
18403 (alter_reg): Treat mode sizes as polynomial.
18405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18406 Alan Hayward <alan.hayward@arm.com>
18407 David Sherwood <david.sherwood@arm.com>
18409 * reload.c (complex_word_subreg_p): New function.
18410 (reload_inner_reg_of_subreg, push_reload): Use it.
18412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18413 Alan Hayward <alan.hayward@arm.com>
18414 David Sherwood <david.sherwood@arm.com>
18416 * lra-constraints.c (process_alt_operands): Reject matched
18417 operands whose sizes aren't ordered.
18418 (match_reload): Refer to this check here.
18420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18421 Alan Hayward <alan.hayward@arm.com>
18422 David Sherwood <david.sherwood@arm.com>
18424 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18425 that the mode size is in the set {1, 2, 4, 8, 16}.
18427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18428 Alan Hayward <alan.hayward@arm.com>
18429 David Sherwood <david.sherwood@arm.com>
18431 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18432 Use plus_constant instead of gen_rtx_PLUS.
18434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18435 Alan Hayward <alan.hayward@arm.com>
18436 David Sherwood <david.sherwood@arm.com>
18438 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18439 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18440 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18441 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18442 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18443 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18444 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18445 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18446 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18447 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18449 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18450 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18451 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18452 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18453 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18454 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18455 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18456 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18457 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18458 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18460 * expr.c (emit_move_resolve_push): Treat the input and result
18461 of PUSH_ROUNDING as a poly_int64.
18462 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18463 (emit_push_insn): Likewise.
18464 * lra-eliminations.c (mark_not_eliminable): Likewise.
18465 * recog.c (push_operand): Likewise.
18466 * reload1.c (elimination_effects): Likewise.
18467 * rtlanal.c (nonzero_bits1): Likewise.
18468 * calls.c (store_one_arg): Likewise. Require the padding to be
18469 known at compile time.
18471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18472 Alan Hayward <alan.hayward@arm.com>
18473 David Sherwood <david.sherwood@arm.com>
18475 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18476 Use plus_constant instead of gen_rtx_PLUS.
18478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18479 Alan Hayward <alan.hayward@arm.com>
18480 David Sherwood <david.sherwood@arm.com>
18482 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18483 rather than an int.
18485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18486 Alan Hayward <alan.hayward@arm.com>
18487 David Sherwood <david.sherwood@arm.com>
18489 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18490 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18491 via stack temporaries. Treat the mode size as polynomial too.
18493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18494 Alan Hayward <alan.hayward@arm.com>
18495 David Sherwood <david.sherwood@arm.com>
18497 * expr.c (expand_expr_real_2): When handling conversions involving
18498 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18499 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18500 as a poly_uint64 too.
18502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18503 Alan Hayward <alan.hayward@arm.com>
18504 David Sherwood <david.sherwood@arm.com>
18506 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18509 Alan Hayward <alan.hayward@arm.com>
18510 David Sherwood <david.sherwood@arm.com>
18512 * combine.c (can_change_dest_mode): Handle polynomial
18513 REGMODE_NATURAL_SIZE.
18514 * expmed.c (store_bit_field_1): Likewise.
18515 * expr.c (store_constructor): Likewise.
18516 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18517 and polynomial REGMODE_NATURAL_SIZE.
18518 (gen_lowpart_common): Likewise.
18519 * reginfo.c (record_subregs_of_mode): Likewise.
18520 * rtlanal.c (read_modify_subreg_p): Likewise.
18522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18523 Alan Hayward <alan.hayward@arm.com>
18524 David Sherwood <david.sherwood@arm.com>
18526 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18527 numbers of elements.
18529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18530 Alan Hayward <alan.hayward@arm.com>
18531 David Sherwood <david.sherwood@arm.com>
18533 * match.pd: Cope with polynomial numbers of vector elements.
18535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18536 Alan Hayward <alan.hayward@arm.com>
18537 David Sherwood <david.sherwood@arm.com>
18539 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18540 in a POINTER_PLUS_EXPR.
18542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18543 Alan Hayward <alan.hayward@arm.com>
18544 David Sherwood <david.sherwood@arm.com>
18546 * omp-simd-clone.c (simd_clone_subparts): New function.
18547 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18548 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18555 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18556 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18557 (expand_vector_condition, vector_element): Likewise.
18558 (subparts_gt): New function.
18559 (get_compute_type): Use subparts_gt.
18560 (count_type_subparts): Delete.
18561 (expand_vector_operations_1): Use subparts_gt instead of
18562 count_type_subparts.
18564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18565 Alan Hayward <alan.hayward@arm.com>
18566 David Sherwood <david.sherwood@arm.com>
18568 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18569 (vect_compile_time_alias): ...this new function. Do the calculation
18570 on poly_ints rather than trees.
18571 (vect_prune_runtime_alias_test_list): Update call accordingly.
18573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18574 Alan Hayward <alan.hayward@arm.com>
18575 David Sherwood <david.sherwood@arm.com>
18577 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18579 (vect_schedule_slp_instance): Likewise.
18581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18582 Alan Hayward <alan.hayward@arm.com>
18583 David Sherwood <david.sherwood@arm.com>
18585 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18586 constant and extern definitions for variable-length vectors.
18587 (vect_get_constant_vectors): Note that the number of units
18588 is known to be constant.
18590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18591 Alan Hayward <alan.hayward@arm.com>
18592 David Sherwood <david.sherwood@arm.com>
18594 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18595 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
18603 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18606 Alan Hayward <alan.hayward@arm.com>
18607 David Sherwood <david.sherwood@arm.com>
18609 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18610 vectors as polynomial. Use build_index_vector for
18611 IFN_GOMP_SIMD_LANE.
18613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18614 Alan Hayward <alan.hayward@arm.com>
18615 David Sherwood <david.sherwood@arm.com>
18617 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18618 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18619 for variable-length vectors.
18620 (vectorizable_mask_load_store): Treat the number of units as
18621 polynomial, asserting that it is constant if the condition has
18622 already been enforced.
18623 (vectorizable_store, vectorizable_load): Likewise.
18625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18626 Alan Hayward <alan.hayward@arm.com>
18627 David Sherwood <david.sherwood@arm.com>
18629 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18630 of units as polynomial. Punt if we can't tell at compile time
18631 which vector contains the final result.
18633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18634 Alan Hayward <alan.hayward@arm.com>
18635 David Sherwood <david.sherwood@arm.com>
18637 * tree-vect-loop.c (vectorizable_induction): Treat the number
18638 of units as polynomial. Punt on SLP inductions. Use an integer
18639 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18640 cast of such a series for variable-length floating-point
18643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18644 Alan Hayward <alan.hayward@arm.com>
18645 David Sherwood <david.sherwood@arm.com>
18647 * tree.h (build_index_vector): Declare.
18648 * tree.c (build_index_vector): New function.
18649 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18650 of units as polynomial, forcibly converting it to a constant if
18651 vectorizable_reduction has already enforced the condition.
18652 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18653 to create a {1,2,3,...} vector.
18654 (vectorizable_reduction): Treat the number of units as polynomial.
18655 Choose vectype_in based on the largest scalar element size rather
18656 than the smallest number of units. Enforce the restrictions
18659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18660 Alan Hayward <alan.hayward@arm.com>
18661 David Sherwood <david.sherwood@arm.com>
18663 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18664 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18671 * target.def (autovectorize_vector_sizes): Return the vector sizes
18672 by pointer, using vector_sizes rather than a bitmask.
18673 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18674 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18675 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18677 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18678 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18679 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18680 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18681 * omp-general.c (omp_max_vf): Likewise.
18682 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18683 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18684 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18685 * tree-vect-slp.c (vect_slp_bb): Likewise.
18686 * doc/tm.texi: Regenerate.
18687 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18689 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18690 the vector size as a poly_uint64 rather than an unsigned int.
18691 (current_vector_size): Change from an unsigned int to a poly_uint64.
18692 (get_vectype_for_scalar_type): Update accordingly.
18693 * tree.h (build_truth_vector_type): Take the size and number of
18694 units as a poly_uint64 rather than an unsigned int.
18695 (build_vector_type): Add a temporary overload that takes
18696 the number of units as a poly_uint64 rather than an unsigned int.
18697 * tree.c (make_vector_type): Likewise.
18698 (build_truth_vector_type): Take the number of units as a poly_uint64
18699 rather than an unsigned int.
18701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18702 Alan Hayward <alan.hayward@arm.com>
18703 David Sherwood <david.sherwood@arm.com>
18705 * target.def (get_mask_mode): Take the number of units and length
18706 as poly_uint64s rather than unsigned ints.
18707 * targhooks.h (default_get_mask_mode): Update accordingly.
18708 * targhooks.c (default_get_mask_mode): Likewise.
18709 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18710 * doc/tm.texi: Regenerate.
18712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18713 Alan Hayward <alan.hayward@arm.com>
18714 David Sherwood <david.sherwood@arm.com>
18716 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18717 * omp-general.c (omp_max_vf): Likewise.
18718 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18719 (expand_omp_simd): Handle polynomial safelen.
18720 * omp-low.c (omplow_simd_context): Add a default constructor.
18721 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18722 (lower_rec_simd_input_clauses): Update accordingly.
18723 (lower_rec_input_clauses): Likewise.
18725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18726 Alan Hayward <alan.hayward@arm.com>
18727 David Sherwood <david.sherwood@arm.com>
18729 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18730 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18731 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18732 (vect_analyze_slp_cost): Likewise.
18733 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18734 (vect_model_load_cost): Likewise.
18736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18737 Alan Hayward <alan.hayward@arm.com>
18738 David Sherwood <david.sherwood@arm.com>
18740 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18741 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18742 from an unsigned int * to a poly_uint64_pod *.
18743 (calculate_unrolling_factor): New function.
18744 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18747 Alan Hayward <alan.hayward@arm.com>
18748 David Sherwood <david.sherwood@arm.com>
18750 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18751 from an unsigned int to a poly_uint64.
18752 (_loop_vec_info::slp_unrolling_factor): Likewise.
18753 (_loop_vec_info::vectorization_factor): Change from an int
18755 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18756 (vect_get_num_vectors): New function.
18757 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18758 (vect_get_num_copies): Use vect_get_num_vectors.
18759 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18760 to an unsigned int *.
18761 (vect_analyze_data_refs): Change min_vf from an int * to a
18763 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18764 than an unsigned HOST_WIDE_INT.
18765 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18766 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18767 to an unsigned int *.
18768 (vect_analyze_data_ref_dependences): Likewise.
18769 (vect_compute_data_ref_alignment): Handle polynomial vf.
18770 (vect_enhance_data_refs_alignment): Likewise.
18771 (vect_prune_runtime_alias_test_list): Likewise.
18772 (vect_shift_permute_load_chain): Likewise.
18773 (vect_supportable_dr_alignment): Likewise.
18774 (dependence_distance_ge_vf): Take the vectorization factor as a
18775 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18776 (vect_analyze_data_refs): Change min_vf from an int * to a
18778 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18779 vfm1 as a poly_uint64 rather than an int. Make the same change
18780 for the returned bound_scalar.
18781 (vect_gen_vector_loop_niters): Handle polynomial vf.
18782 (vect_do_peeling): Likewise. Update call to
18783 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18784 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18786 * tree-vect-loop.c (vect_determine_vectorization_factor)
18787 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18788 (vect_get_known_peeling_cost): Likewise.
18789 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18790 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18791 (vect_transform_loop): Likewise. Use the lowest possible VF when
18792 updating the upper bounds of the loop.
18793 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18794 rather than an int.
18795 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18796 polynomial unroll factors.
18797 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18798 (vect_make_slp_decision): Likewise.
18799 (vect_supported_load_permutation_p): Likewise, and polynomial
18801 (vect_analyze_slp_cost): Handle polynomial vf.
18802 (vect_slp_analyze_node_operations): Likewise.
18803 (vect_slp_analyze_bb_1): Likewise.
18804 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18805 than an unsigned HOST_WIDE_INT.
18806 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18807 (vectorizable_load): Handle polynomial vf.
18808 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18810 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18813 Alan Hayward <alan.hayward@arm.com>
18814 David Sherwood <david.sherwood@arm.com>
18816 * match.pd: Handle bit operations involving three constants
18817 and try to fold one pair.
18819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18821 * tree-vect-loop-manip.c: Include gimple-fold.h.
18822 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18823 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18824 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18825 Add a path that uses a step of VF instead of 1, but disable it
18827 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18828 and niters_no_overflow parameters. Update calls to
18829 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18830 Create a new SSA name if the latter choses to use a ste other
18831 than zero, and return it via niters_vector_mult_vf_var.
18832 * tree-vect-loop.c (vect_transform_loop): Update calls to
18833 vect_do_peeling, vect_gen_vector_loop_niters and
18834 slpeel_make_loop_iterate_ntimes.
18835 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18836 (vect_gen_vector_loop_niters): Update declarations after above changes.
18838 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18840 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18841 128-bit round to integer instructions.
18842 (ceil<mode>2): Likewise.
18843 (btrunc<mode>2): Likewise.
18844 (round<mode>2): Likewise.
18846 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18848 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18849 unaligned VSX load/store on P8/P9.
18850 (expand_block_clear): Allow the use of unaligned VSX
18851 load/store on P8/P9.
18853 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18855 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18857 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18858 swap associated with both a load and a store.
18860 2018-01-02 Andrew Waterman <andrew@sifive.com>
18862 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18863 * config/riscv/riscv.md (clear_cache): Use it.
18865 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18867 * web.c: Remove out-of-date comment.
18869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18871 * expr.c (fixup_args_size_notes): Check that any existing
18872 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18873 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18874 (emit_single_push_insn): ...here.
18876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18878 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18879 (const_vector_encoded_nelts): New function.
18880 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18881 (const_vector_int_elt, const_vector_elt): Declare.
18882 * emit-rtl.c (const_vector_int_elt_1): New function.
18883 (const_vector_elt): Likewise.
18884 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18885 of CONST_VECTOR_ELT.
18887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18889 * expr.c: Include rtx-vector-builder.h.
18890 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18891 directly on the tree encoding.
18892 (const_vector_from_tree): Likewise.
18893 * optabs.c: Include rtx-vector-builder.h.
18894 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18895 sequence of "u" values.
18896 * vec-perm-indices.c: Include rtx-vector-builder.h.
18897 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18898 directly on the vec_perm_indices encoding.
18900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18902 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18903 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18904 * rtx-vector-builder.h: New file.
18905 * rtx-vector-builder.c: Likewise.
18906 * rtl.h (rtx_def::u2): Add a const_vector field.
18907 (CONST_VECTOR_NPATTERNS): New macro.
18908 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18909 (CONST_VECTOR_DUPLICATE_P): Likewise.
18910 (CONST_VECTOR_STEPPED_P): Likewise.
18911 (CONST_VECTOR_ENCODED_ELT): Likewise.
18912 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18913 (unwrap_const_vec_duplicate): Likewise.
18914 (const_vec_series_p): Check for a non-duplicated vector encoding.
18915 Say that the function only returns true for integer vectors.
18916 * emit-rtl.c: Include rtx-vector-builder.h.
18917 (gen_const_vec_duplicate_1): Delete.
18918 (gen_const_vector): Call gen_const_vec_duplicate instead of
18919 gen_const_vec_duplicate_1.
18920 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18921 (gen_const_vec_duplicate): Use rtx_vector_builder.
18922 (gen_const_vec_series): Likewise.
18923 (gen_rtx_CONST_VECTOR): Likewise.
18924 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18925 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18926 Build a new vector rather than modifying a CONST_VECTOR in-place.
18927 (handle_special_swappables): Update call accordingly.
18928 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18929 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18930 Build a new vector rather than modifying a CONST_VECTOR in-place.
18931 (handle_special_swappables): Update call accordingly.
18933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18935 * simplify-rtx.c (simplify_const_binary_operation): Use
18936 CONST_VECTOR_ELT instead of XVECEXP.
18938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18940 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18941 the selector elements to be different from the data elements
18942 if the selector is a VECTOR_CST.
18943 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18944 ssizetype for the selector.
18946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18948 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18949 before testing each element individually.
18950 * tree-vect-generic.c (lower_vec_perm): Likewise.
18952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18954 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18955 * selftest-run-tests.c (selftest::run_tests): Call it.
18956 * vector-builder.h (vector_builder::operator ==): New function.
18957 (vector_builder::operator !=): Likewise.
18958 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18959 (vec_perm_indices::all_from_input_p): New function.
18960 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18961 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18962 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18963 instead of reading the VECTOR_CST directly. Detect whether both
18964 vector inputs are the same before constructing the vec_perm_indices,
18965 and update the number of inputs argument accordingly. Use the
18966 utility functions added above. Only construct sel2 if we need to.
18968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18970 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18971 the broadcast of the low byte.
18972 (expand_mult_highpart): Use an explicit encoding for the permutes.
18973 * optabs-query.c (can_mult_highpart_p): Likewise.
18974 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18975 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18976 (vectorizable_bswap): Likewise.
18977 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18978 explicit encoding for the power-of-2 permutes.
18979 (vect_permute_store_chain): Likewise.
18980 (vect_grouped_load_supported): Likewise.
18981 (vect_permute_load_chain): Likewise.
18983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18985 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18986 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18987 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18988 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18989 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18990 (vect_gen_perm_mask_any): Likewise.
18992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18994 * int-vector-builder.h: New file.
18995 * vec-perm-indices.h: Include int-vector-builder.h.
18996 (vec_perm_indices): Redefine as an int_vector_builder.
18997 (auto_vec_perm_indices): Delete.
18998 (vec_perm_builder): Redefine as a stand-alone class.
18999 (vec_perm_indices::vec_perm_indices): New function.
19000 (vec_perm_indices::clamp): Likewise.
19001 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19002 (vec_perm_indices::new_vector): New function.
19003 (vec_perm_indices::new_expanded_vector): Update for new
19004 vec_perm_indices class.
19005 (vec_perm_indices::rotate_inputs): New function.
19006 (vec_perm_indices::all_in_range_p): Operate directly on the
19007 encoded form, without computing elided elements.
19008 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19009 encoding. Update for new vec_perm_indices class.
19010 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19011 the given vec_perm_builder.
19012 (expand_vec_perm_var): Update vec_perm_builder constructor.
19013 (expand_mult_highpart): Use vec_perm_builder instead of
19014 auto_vec_perm_indices.
19015 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19016 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19017 or double series encoding as appropriate.
19018 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19019 vec_perm_indices instead of auto_vec_perm_indices.
19020 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19021 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19022 (vect_permute_store_chain): Likewise.
19023 (vect_grouped_load_supported): Likewise.
19024 (vect_permute_load_chain): Likewise.
19025 (vect_shift_permute_load_chain): Likewise.
19026 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19027 (vect_transform_slp_perm_load): Likewise.
19028 (vect_schedule_slp_instance): Likewise.
19029 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19030 (vectorizable_mask_load_store): Likewise.
19031 (vectorizable_bswap): Likewise.
19032 (vectorizable_store): Likewise.
19033 (vectorizable_load): Likewise.
19034 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19035 vec_perm_indices instead of auto_vec_perm_indices. Use
19036 tree_to_vec_perm_builder to read the vector from a tree.
19037 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19038 vec_perm_builder instead of a vec_perm_indices.
19039 (have_whole_vector_shift): Use vec_perm_builder and
19040 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19041 truncation to calc_vec_perm_mask_for_shift.
19042 (vect_create_epilog_for_reduction): Likewise.
19043 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19044 from auto_vec_perm_indices to vec_perm_indices.
19045 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19046 instead of changing individual elements.
19047 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19048 the vector in d.perm.
19049 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19050 from auto_vec_perm_indices to vec_perm_indices.
19051 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19052 instead of changing individual elements.
19053 (arm_vectorize_vec_perm_const): Use new_vector to install
19054 the vector in d.perm.
19055 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19056 Update vec_perm_builder constructor.
19057 (rs6000_expand_interleave): Likewise.
19058 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19059 (rs6000_expand_interleave): Likewise.
19061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19063 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19064 to qimode could truncate the indices.
19065 * optabs.c (expand_vec_perm_var): Likewise.
19067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19069 * Makefile.in (OBJS): Add vec-perm-indices.o.
19070 * vec-perm-indices.h: New file.
19071 * vec-perm-indices.c: Likewise.
19072 * target.h (vec_perm_indices): Replace with a forward class
19074 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19075 * optabs.h: Include vec-perm-indices.h.
19076 (expand_vec_perm): Delete.
19077 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19078 (expand_vec_perm_const): Declare.
19079 * target.def (vec_perm_const_ok): Replace with...
19080 (vec_perm_const): ...this new hook.
19081 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19082 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19083 * doc/tm.texi: Regenerate.
19084 * optabs.def (vec_perm_const): Delete.
19085 * doc/md.texi (vec_perm_const): Likewise.
19086 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19087 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19088 expand_vec_perm for constant permutation vectors. Assert that
19089 the mode of variable permutation vectors is the integer equivalent
19090 of the mode that is being permuted.
19091 * optabs-query.h (selector_fits_mode_p): Declare.
19092 * optabs-query.c: Include vec-perm-indices.h.
19093 (selector_fits_mode_p): New function.
19094 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19095 is defined, instead of checking whether the vec_perm_const_optab
19096 exists. Use targetm.vectorize.vec_perm_const instead of
19097 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19098 fit in the vector mode before using a variable permute.
19099 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19100 vec_perm_indices instead of an rtx.
19101 (expand_vec_perm): Replace with...
19102 (expand_vec_perm_const): ...this new function. Take the selector
19103 as a vec_perm_indices rather than an rtx. Also take the mode of
19104 the selector. Update call to shift_amt_for_vec_perm_mask.
19105 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19106 Use vec_perm_indices::new_expanded_vector to expand the original
19107 selector into bytes. Check whether the indices fit in the vector
19108 mode before using a variable permute.
19109 (expand_vec_perm_var): Make global.
19110 (expand_mult_highpart): Use expand_vec_perm_const.
19111 * fold-const.c: Includes vec-perm-indices.h.
19112 * tree-ssa-forwprop.c: Likewise.
19113 * tree-vect-data-refs.c: Likewise.
19114 * tree-vect-generic.c: Likewise.
19115 * tree-vect-loop.c: Likewise.
19116 * tree-vect-slp.c: Likewise.
19117 * tree-vect-stmts.c: Likewise.
19118 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19120 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19121 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19122 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19123 (aarch64_vectorize_vec_perm_const): ...this new function.
19124 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19125 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19126 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19127 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19128 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19129 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19130 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19132 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19133 check for NEON modes.
19134 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19135 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19136 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19137 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19139 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19140 the old VEC_PERM_CONST conditions.
19141 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19142 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19143 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19144 (ia64_vectorize_vec_perm_const_ok): Merge into...
19145 (ia64_vectorize_vec_perm_const): ...this new function.
19146 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19147 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19148 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19149 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19150 * config/mips/mips.c (mips_expand_vec_perm_const)
19151 (mips_vectorize_vec_perm_const_ok): Merge into...
19152 (mips_vectorize_vec_perm_const): ...this new function.
19153 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19154 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19155 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19156 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19157 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19158 (rs6000_expand_vec_perm_const): Delete.
19159 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19161 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19162 (altivec_expand_vec_perm_const_le): Take each operand individually.
19163 Operate on constant selectors rather than rtxes.
19164 (altivec_expand_vec_perm_const): Likewise. Update call to
19165 altivec_expand_vec_perm_const_le.
19166 (rs6000_expand_vec_perm_const): Delete.
19167 (rs6000_vectorize_vec_perm_const_ok): Delete.
19168 (rs6000_vectorize_vec_perm_const): New function.
19169 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19170 an element count and rtx array.
19171 (rs6000_expand_extract_even): Update call accordingly.
19172 (rs6000_expand_interleave): Likewise.
19173 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19174 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19175 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19176 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19177 (rs6000_expand_vec_perm_const): Delete.
19178 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19179 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19180 (altivec_expand_vec_perm_const_le): Take each operand individually.
19181 Operate on constant selectors rather than rtxes.
19182 (altivec_expand_vec_perm_const): Likewise. Update call to
19183 altivec_expand_vec_perm_const_le.
19184 (rs6000_expand_vec_perm_const): Delete.
19185 (rs6000_vectorize_vec_perm_const_ok): Delete.
19186 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19187 reference to the SPE evmerge intructions.
19188 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19189 an element count and rtx array.
19190 (rs6000_expand_extract_even): Update call accordingly.
19191 (rs6000_expand_interleave): Likewise.
19192 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19193 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19195 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19199 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19200 vector mode and that that mode matches the mode of the data
19202 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19203 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19204 directly using expand_vec_perm_1 when forcing selectors into
19206 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19210 * optabs-query.h (can_vec_perm_p): Delete.
19211 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19212 * optabs-query.c (can_vec_perm_p): Split into...
19213 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19214 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19215 particular selector is valid.
19216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19217 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19218 (vect_grouped_load_supported): Likewise.
19219 (vect_shift_permute_load_chain): Likewise.
19220 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19221 (vect_transform_slp_perm_load): Likewise.
19222 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19223 (vectorizable_bswap): Likewise.
19224 (vect_gen_perm_mask_checked): Likewise.
19225 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19226 implementations of variable permutation vectors into account
19227 when deciding which selector to use.
19228 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19229 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19230 with a false third argument.
19231 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19232 to test whether the constant selector is valid and can_vec_perm_var_p
19233 to test whether a variable selector is valid.
19235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19237 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19238 * optabs-query.c (can_vec_perm_p): Likewise.
19239 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19240 instead of vec_perm_indices.
19241 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19242 (vect_gen_perm_mask_checked): Likewise,
19243 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19244 (vect_gen_perm_mask_checked): Likewise,
19246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19248 * optabs-query.h (qimode_for_vec_perm): Declare.
19249 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19250 (qimode_for_vec_perm): ...this new function.
19251 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19253 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19255 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19256 does not have a conditional at the top.
19258 2018-01-02 Richard Biener <rguenther@suse.de>
19260 * ipa-inline.c (big_speedup_p): Fix expression.
19262 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19265 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19268 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19272 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19273 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19274 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19275 cond_taken_branch_cost 3->4.
19277 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19279 PR tree-optimization/83581
19280 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19281 TODO_cleanup_cfg if any changes have been made.
19283 PR middle-end/83608
19284 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19285 convert_modes if target mode has the right side, but different mode
19288 PR middle-end/83609
19289 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19290 last argument when extracting from CONCAT. If either from_real or
19291 from_imag is NULL, use expansion through memory. If result is not
19292 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19293 the parts directly to inner mode, if even that fails, use expansion
19296 PR middle-end/83623
19297 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19298 check for bswap in mode rather than HImode and use that in expand_unop
19301 Copyright (C) 2018 Free Software Foundation, Inc.
19303 Copying and distribution of this file, with or without modification,
19304 are permitted in any medium without royalty provided the copyright
19305 notice and this notice are preserved.